I'm not sure I follow. So you failed to measure software productivity in lines of code, therefore it follows that "There's No Such Thing as Software Productivity"? Don't you think that giving up after n=1 attempts at measuring software productivity might be a tad too fast to draw a generalized claim of impossibility? I might argue the real lesson learned is "Lines of Code are Not a Measure of Productivity in an Isolated, Toy Example".
I suspect this sort of thing gets promulgated because it kind of massages our ego, like yes, they can measure other sorts of productivity, but not ours, oh no, we're too complex and intelligent, there's no way to measure the deep sorts of work that we do! Which, yes, OK, we're not exactly bricklayers, but surely, if you had to, you could do better.
> it kind of massages our ego, like yes, they can measure other sorts of productivity, but not ours, oh no, we're too complex and intelligent, there's no way to measure the deep sorts of work that we do! Which, yes, OK, we're not exactly bricklayers, but surely, if you had to, you could do better.
Productivity is not a cut-n-dry stat for any kind of knowledge work.
For an admin, we could say they were more productive if they handled more cases than last month. But this does not account for tricky cases or ambiguous cases, which might have taken 2x time to sort out but on rote numbers it looks they were less productive.
For us, there are no specific goals(except arbitrary imaginary deadlines by clueless management and sprint points) to correctly measure productivity. A junior engineer may spend 4h each day blasting out many lines of codes, a mid/senior may spend more hours in reviews or meetings. A 5 line PR with test coverage is faster to review than a 800+ LOC PR with additional tests and have significant risk of breaking something, so review number is also not a good indicator. I need to coordinate with min. 12 people to get any new credentials or access to specific env owned by another team, which is not countable as productive but unavoidable. How about that garbage meeting where I was just spending time yawning because some new tech lead who believes that a modular monolith system is better if it was rewritten as event based. Productivity measurement is hard.
To take your envy(I sense it on the quoted remark) to another place, think about how you go about measuring productivity of a CEO, PO, PM, Scrum Master, Agilist/Agile-Specialist etc?
Sure, productivity it's not cut-n-dry stat, but it's not completely impossible to measure either. For a team, there are very clear goals: to make product that customers like, and to keep improving it for a long time. The last part is especially hard to measure, so people do all sorts of approximations, some very bad.
For individual, productivity is measured on multiple axis - new long-term features, throw-away and prototypes, code maintenance, fire-fighting, code reviews, high-level design, outside-of-team communications, intra-team communications (for example helping teammates), archeology, specific parts of your project, probably others. In ideal team, everyone is strong on at least one or two axes, so no person is worse than others. In real teams, I've seen people who were weak on every single axis and did not seem to improve with time. I've also seen people who are great at everything, and it's extremely nice when such people exist - but they are very rare.
(Btw, re your specific example: if you weren't productive because you were waiting for others to approve your access then you were not productive. There is nothing complex about this, there were plenty of times when I've said, "I did nothing for project S last week because they still could not give me access")
> To take your envy(I sense it on the quoted remark) to another place, think about how you go about measuring productivity of a CEO, PO, PM, Scrum Master, Agilist/Agile-Specialist etc?
I have no envy; I'm a software engineer as well. I just don't seem to struggle, as others claim to, in measuring productivity. I find it fairly straightforward to see that some people are more productive than others, at least in my workplace where we can hold most things constant (e.g. meeting count, managerial ability, etc, etc). Yes, there may be external factors, and that's unfortunate. Yes, it's not fair that some people are more productive than others despite putting in half the effort. But I'm not going to stick my head in the sand and pretend it's invisible.
> I just don't seem to struggle, as others claim to, in measuring productivity.
Because you are measuring at a very broad and basic level.
Steve is more productive than Susan.
Great. How much more productive? Can you turn it into a number?
Can you still do it consistently when Steve and Susan are in different teams in different parts of the organisation trying to achieve different goals?
I've done DB upgrades that took 10 minutes and I've done DB upgrades that took 3-4 months. What changed was not my productivity but the nature of the problem. Yet from the outside they were both just DB upgrades.
If Susan had done the DB upgrade in 12 weeks could we confidently claim that Steve could have done it in 11 weeks? Steve hasn't even done a DB upgrade since he joined the company. Perhaps Steve could have done it in 10 minutes?
I don't think anyone can get numbers, but partial ordering is much easier.
If Steve and Susan are in different part of organization, the answer is "cannot compare". If they are doing different job, the answer is the same.
But every once in a while there is a scenarios when you can compare people easily.
There has a weekly rotation to be an support person for other team. During his week, John always answers questions quickly and to the team's satisfaction. Meanwhile James struggles to answer them and cannot troubleshoot product his team is writing. This has been going on for multiple months and hundreds of questions for each, so it's not "bad week" unlucky or fluke. We now know who is better at answering questions about product.
John and James are doing DB migrations, they did many dozens of them. The migrations are assigned randomly. But John is usually finishing his migrations with no problems, while James often caused outages or missing data. A few times James took over two months to migrate, so the task was taken from him and given to John. John had to discard everything James did and migrate everything from scratch. Now there is a migration for a very important client and CEO is fed up with random assignment.. who is he going to choose?
What your scenario doesn't address is that while John finished his migrations on time, James has designed the flagship order processing pipeline something that John could never pull off.
Or maybe while John is technically adept, he's also a huge jerk and belittles people at standup, while James is the quintessential communicator with jr devs, etc.
Real life is messy. I've seen more people get replaced due to attitude or teamfit issues than specifically due to technical incompetence.
"What changed was not my productivity but the nature of the problem"
I think that's the source of the problem: it's impossible to measure the "work" required to solve most software problems. If you tell me to carry a stone up a hill, I can put that in a formula and know exactly how much work I'll have to do. But if you give me a ticket to do a DB upgrade, I can, at best, make an educated guess.
So by the time I close the ticket, how much work have I done, and how do I know whether the time I've spent it proportional to the work I've done?
> Great. How much more productive? Can you turn it into a number?
This is moving goalposts. OP's argument was "There's No Such Thing as Software Productivity", not "You Can't Convert Software Productivity into a Floating Point Number With 3 Decimals of Accuracy."
By that measure there also isn't a real dependable/reproducible scalar that measures athleticism. Nonetheless some people are clearly more athletic than others. Also within a single sport we can easily see that some players are better than others. Here too you could object that people who play offense should not be compared to players in defense. Or that it's not individual players that matters but the team as a whole. And yet, we can still figure out easily who the star players are.
The ability to look productive is a different skill. The ability to create any impression is a skill some have. Creating fires and putting them out like a hero is another skill.
What are your eyes measuring? Are you being fooled?
If you are IC, it's pretty easy to see if your teammates are "looking productive" vs "are really productive", as you know all the code and how it works.
For other teams yeah, it's harder, or maybe even impossible if the teams are isolated.
> If you are IC, it's pretty easy to see if your teammates
> are "looking productive" vs "are really productive"
I'm not sure how true this really is. We all like to think that we're good at estimating other people's ability, particularly if we work closely with them, but how can you validate that impression?
Something that happened to me at one point in my career was that I joined a team that was half new hires with little experience, and half long-term employees with a lot of experience. Because the long-term employees were typically extremely busy, I ended up answering a lot of questions from the entry-level devs. So I spent time documenting the system, doing trainings, doing 1-on-1 code reviews, and so on.
During the performance review, feedback from the entry-level devs was that I was highly productive, and feedback from the long-term employees was that I was a slacker.
I've been on all sides of this issue and I guarantee that as an individual contributor you lack the perspective and visibility into all of the factors that make up net productivity for your colleagues. Once you get a little more maturity and experience you'll probably understand this better.
No offense but reading this comment with the utmost attempt at interpreting it in good faith, it kind of sounds like you're just going off of vibes and can't actually quantify productivity any more than anyone else can. Otherwise... How exactly are you quantifying it? If you had to show your work, could you?
I dunno, the opposite side seems equally absurd to me. Do you work at a software company? Do you have coworkers? Are you honestly telling me that, gun to your head, you couldn't say that some are more productive and some are less? That if you had to choose a co-founder for your next startup, that you would have NO idea where to begin, that any of them would be equally as good as the next?
I mean look, some people seem to be extremely productive, and sure, I could say that those people are almost surely more productive than others. It's still genuinely very hard to quantify if those people are actually vastly more productive, or if it just looks like that because they produce more obvious artifacts of their productivity. Hell, what is productivity, is it more productive to fix 1000 bugs or to be the tech lead on a product that reaches 1 million ARR? Does it matter what bugs?
I stand by what I said, whether people like it or not.
Including the part about going off vibes. Please explain how this "you can just kind of tell" mentality is not literally just going off of vibes?
(And I wouldn't choose a cofounder only based on productivity, anyways. I'm not even sure that would be the main criteria.)
How would you choose a cofounder? I would call whatever metric you use to determine "would that person be a good cofounder" productivity. (OK, sure, let's narrow it down and say that you were tasked with choosing a purely technical cofounder; perhaps in this hypothetical you're helping your non-technical friend find one?) Do you still believe you couldn't do better than chance?
Well, a lot of the factors explicitly have nothing to do with general productivity at all, such as alignment. The goal with regards to selecting for productivity would be "someone who seems productive enough" but getting hung up on exactly how productive feels like a mistake, as aside from being unquantifiable, that isn't going to be the only or probably even main factor that decides the fate of the endeavor. Productivity is not the most challenging problem in software development. In my opinion, as far as factors for success for a technology endeavor goes, even overall technical competence winds up factoring in fairly low at the end of the day. What's important is meeting the threshold you need, not getting as high of a score as possible.
I should clarify that I actually agree with you if what you're saying is that the best measure of productivity we have is literally just going off your gut, but my argument is that this gut feeling is terrible. This is in large part because humans are biased, our gut feelings are swayed by things that simply shouldn't factor in. We tend to have more positive opinions of people that we think are like us and we sometimes wind up having negative opinions of someone based on stupid things like disagreeing with them on something that is ultimately irrelevant to whether or not they are productive.
And to me, the ultimate nail in the coffin is really in the question of what really constitutes productivity. Productivity is supposed to measure the efficiency of producing something, which already has pitfalls in and of itself when dealing with things that do have discrete, measurable indicators of progress, but programming doesn't, it's not even always obvious if progress is forward or backward sometimes. What's most important, performance optimizations, disaster planning, features, robustness, minimizing resource utilization? The best answer you can generally say is "It depends," and moreover, everyone will have a different set of competencies they're best at, so a person doesn't really have some single "productivity value" you could summarize them with. Yet, all of those things are pretty important for any serious technology organization, so you would want people with a variety of different affinities.
At that point, it starts to beg the question of whether or not attempting to directly measure software productivity is a worthwhile endeavor. I'd argue not.
I suppose the question is: can you supply a word you would use when deciding on who to hire for a technical role? Does anyone off the street contribute the same amount of that word, or would you be selective?
> Don't you think that giving up after n=1 attempts at measuring software productivity might be a tad too fast to draw a generalized claim of impossibility?
Software developers should look at anyone claiming they can measure software productivity as a snake oil salesman.
We have seen hundreds of attempts over the years and they have all "failed". More accurately they all have large error bars and biases.
Researchers can and should continue looking into how to measure software development productivity. It is likely over the next few decades we will start to understand how to measure it appropriately.
I agree. I know from looking at a story that if Alice picks it up it’ll take her about a day and be decent quality, but if Bob picks it up it’ll take him about a week and be worse quality. And I know that this will be pretty consistent week-in, week-out.
Developer productivity exists and I am totally comfortable describing Alice as being more productive than Bob. The fact that there isn’t a good, generic way to distill this into particular values does not mean that the phenomenon doesn’t exist.
We’ve all worked with Alice and Bobs. Claiming there’s no such thing as developer productivity is denying what we’ve all experienced. We aren’t special snowflakes whose work is beyond the ken of mortal men. Our work has value and sometimes we produce a lot of value in a particular time period and sometimes we product a little. This is productivity.
This is not an isolated toy example, it's reductio ad absurdum (https://en.wikipedia.org/wiki/Reductio_ad_absurdum) - and until today I haven't seen a way to measure software productivity (might also apply to other knowledge work) that is resistant to that...
The text gives an example to the core problem, and to argue differently requires thinking around it.
In practice. I’ve seen many attempts at measuring productivity, but once you dig into them, you see they are just abstraction mechanisms above something that is similar to lines of code.
I have yet to see an idea that sidesteps the core issue described in this post. Also, it applies to many types of work, and software is not unique in any way.
All other measures are a proxy for happy customers.
Actually, happy customers is also a proxy (the real measure is profits) but measuring profits directly (in the short term) can lead to decisions that have adverse long term effects. It's too easy to increase profits in the short term by avoiding long-term expenses.
So, if you're in the business of software, the goal is happy customers. (And I use the word Customers carefully here. Not just Users who pay nothing, but Customers who spend money.)
In a business context, it's really the only thing that matters. But, of course, it can be hard to measure (are they Happy?) and relies on multiple disciplines. Production (coding), Marketing, Sales, Support, Documentation, Training- all need to be working well to make it work.
Ultimately if the big picture doesn't lead to Happy Customers (again, I stress, in a businesses context) then no-one is "being productive."
When a measure becomes the objective, it ceases to be a good measure.
Even if you could perfectly measure customer happiness (very hard, as you note) - it's relatively easy to make customers happy by giving them more value than what they pay for. Sure, that may cost your business more money than what it makes with said customers, but hey, who cares, "profit" was not the metric...
(and as you note, if you make "profit" the metric, that has its own set of challenges - e.g. the optimization towards short-term profit in detriment of the long-term sanity, which is what we observe in a lot of corporations).
That's very tricky thing to quantify, especially with "unsung heroes". If my work is in preventing problems, the guy that fixes problems will be seen as the one that contributes more to profit, since impact is directly observed/measured.
This is something that one of the orgs I worked for eventually realized. The people f'ing up, and then fixing their mistakes, were the ones getting promotions/bonuses/raises, because they were the ones interacting with all the execs.
I think people get hung up on wanting to collapse developer productivity into a single dimension, usually for stack-ranking purposes. This, I think, is always going to punish good engineers and reward bad ones to some degree.
Measuring developer productivity should, in my opinion, have one dimension for speed, one for quality, and one for user impact. LOC can be fine as a measurement for speed, you just don’t want to look at it in isolation. You would want to also measure, for example, escape rate and usage for the features the developer worked on, and be willing to change or refine these if circumstances require it.
You also need to look for different profiles based on the developer’s level of seniority. A senior dev probably shouldn’t be writing as much code as a contributor, but their user impact should be high, and escape rate low. Analyzing differences between teams is important, as well. A team that has a lot of escapes or little user impact probably has issues that need management attention, and may not have anything at all to do with individual developer productivity or ability.
In brief, the numbers are there to help you make better management decisions, not to relieve you of having to make them.
Bricklayers have measures that are usable by people who know little to nothing about the profession who are incapable of doing the work. This is what people want. An objective measure that can be applied by people functionally incapable of actually doing the work.
Wherein the work isn't repeating existing work no such measure should be expected to exist.
Any creative work is going to suffer from the same problem.
If you asked a brick layer to develop a better workflow for his fellows to work more effectively in a particular sitiation it would be the same. Are you going to measure words per minute?
The problem is that corporate types don't see programming as creative work. Type faster!!!! If you push the buttons faster the product gets build faster!
Also imagine every other week as the bricklayer had half a wall built a PM comes out and says, "tear this all down and move it 3 feet to the north", and then the week after, "now move it east 1 foot, and why is it taking you so long?"
Function point analysis works well enough for measuring software productivity in most domains, provided that quality is held constant. Like other software productivity metrics it's only meaningful at the level of a complete product team and worse than useless for evaluating individual team members.
So you can measure productivity with a reasonable level of accuracy and consistency, but then what? For most organizations it's not actionable so calculating that metric ends up as another form of waste.
You certainly did fail to see the idea being conveyed by a simple, but poignant anecdote. If this is all you have to say, I might argue the real ego being massaged is your own; by yourself.
You can say this about any discipline. The root of the issue is that productivity for productivity's sake is meaningless, and it makes no sense to measure productivity as a general property when outputs vary.
A tire factory has a distinct, singular goal: produce tires. It does this continually. Productivity is meaningful, but only in relation to a target that is typically specified by externalities (e.g. amount of demand)
A software company is usually not in the business of producing consumable commodities so this kind of measurement does not make sense. It can make sense to measure productivity during a period for delivering a particular piece of software within a given time bound, but once it's delivered, productivity becomes meaningless. You always need to understand productivity in relation to some purpose and I don't know how these knuckleheads who think this abstract idea is basically like a concrete measurable essence, like mass, or liquid, got leadership positions.
Seems to me you’re talking about measuring software. Once we pick a measure though, we can calculate output/input (productivity).
Imagine two people tasked with producing the same software, or software satisfying the same requirements or test suite. What would you call the person who produces it faster? More productive?
Measuring software in financial terms or lines of code might not be the right measurement of software in all situations, but surely we can measure time and cost to produce software or software that satisfies equivalent requirements.
You’re responding to the headline, not the article. The article has a plainly stated thesis: “Even if it could be measured, productivity in software does not approximate business value in any meaningful way”
The article contains this sentence, yes, but the rest of the text does not support it.
The article argues that lines-of-code is bad metric, and no one argues about this.
thuanao proposes measuring "number of completed projects", and I think it's a pretty good metric in some situations which don't involve long-term maintenance.
One example I can come up with is data science: two engineers are given identical requirements to ingest the data (once) and calculate certain metrics. If the assignments are identical, we can calculate and compare their productivity, in "projects/month"
Long-term maintenance is a confounding factor of course. People sometimes write software faster with the downside that it wil be harder to maintain.
So remove the confounding factor (at the same time removing most projects we would like to apply this to) and it becomes an easy problem?
But there are many confounding factors. Your data science project results will be different (different metric values outputed) and you don't know which one is correct, or closer to correct. Now how does the "projects/month" number look?
Or maybe one of the engineers uses way more compute-hours for getting the job done. Their projects/month number is better, but is that really a better outcome? Compute is not free.
By the time you remove most confounding factors, the productivity measure will only apply to an insignificant number of projects.
I suppose in the context of this thread, we can look at the productivity of a system that includes money or does not.
A small system, like a tire production line, can measure its productivity in terms of money, which is outside the system. But a large system, like society, includes money, and cannot measure its own productivity using internal things.
By the way, why suddenly the money=bad sentiment is so popular? I thought USSR example loudly showed us what happens when people think that money is evil.
When companies focus on profit at the expense of all other goals, the result is typically disadvantageous to customers, employees, and society at large.
See also: crapification; financialization; parasitic private equity
I can't speak for HN, but I imagine some readers are more interested in the benefits of cool technology beyond putting more money in the pockets of investors and management.
> I would argue that what good software developers do is remove problems. The opposite, in fact, of production
But something is being produced - it is version 2.0 of the software. This is an artifact that is then shipped to users or deployed to a server. Peter’s solution fixed the issue and did not (seemingly) create further maintenance burden, which would have taken attention away from other tasks, i.e. reduced future productivity.
I agree that metrics for programmer productivity are often useless (e.g. using lines of code is a bad idea for obvious reasons), but it seems silly to claim that the entire concept of productivity does not apply to the production of software.
Silly and convenient. No one takes the claim that you can’t measure software productivity seriously and everyone simultaneously agrees that simple scalar metrics often fail to show the big picture in any and all disciplines.
The rest is just usual software guy hubris and lack of awareness of the discipline.
> No one takes the claim that you can’t measure software productivity seriously
At least three people do, and they are mentioned in the article: the author, his colleague, and Martin Fowler.
I suspect that software productivity (like the productivity of scientists, artists, composers, etc.) on any non-trivial project may be measurable on a scale of years, but not necessarily months or calendar quarters. A larger issue is that productivity is often due to external factors as much as it is to individual effort, and overnight success is often the result of extended periods of limited progress or even repeated failure.
Let's say that you have two runners running the same marathon. The first one, Frank, sprints at full speed and eventually tires out and slows down. The second runner, Peter, takes a nap first and then finishes the marathon at the exact same time. Which of these two runners was faster in the race? The answer is: It doesn't matter. And therefore there is no such thing as running speed.
There is such a thing as productivity in programming. If you could measure it it would likely be some combination of peer review and an analysis of the impact implemented features and fixes had. Some companies actually have programmers rate each other. I don't know how well it works and I think it can lead to perverse scenarios. But you can come up with metrics that are positively correlated with productivity.
Thats why bugs and chores have 0 points per default in pivotal tracker. Thats also why there was a push to force us say "as a user i..." at the beginning of story descriptions to make sure this is something creating user value. With these guards in place i don't follow the argument that productivity is not measurable, if a team builds features that solve an expressed user need and lets even say go through a final user acceptance check, this is a productive process and it is also very measurable!
Isn't this whole debate a repeat of the old talking point about what is value? Labour theory of value, that kind of thing?
This is simply never going to end if carried on along the lines of this article, or along the lines of most of the comments here.
There's no way to objectively and reasonably put a value on something.
All we have is a theory of subjective value, which does a bit of handwaving about utility and works out some ways where we can come to a price, regardless of the fact that the values in peoples' heads are subjective.
Thus the distinction between knowledge work and "tangible work" like bricklaying is actually a moot point. Yes, you can measure "productivity" of bricklaying in metres per day, but ultimately you care about value, not amounts of wall.
The arguments about one guy definitely being more productive than another are similar. One person values speed, another values maintenance costs downstream. It is subjective what ought to be more important.
Output is a weak proxy for impact. But it’s the one that makes intuitive sense to people. Doesn’t make it right or useful. I’m sure you all can envision a parable about your subfield of expertise that showcases how a seemingly light touch has a huge positive impact.
It's straight up impossible. Best we can do is observe and attempt to measure proxies.
Take hypothetical situation: VPs A & B debate a business decision. Let's say A wins the argument and their solution leads to a revenue increase of $10M, and let's say we can confidently state that this is the outcome driven primarily by them winning the argument. Is the impact a net growth in business of $10M? In a sense, yes; but in some other sense, perhaps if they went with B's solution, the revenue growth would have been $15M? There's probably no way to know for sure unless you try both approaches, which is often impossible...
Yeah, the comments on the post already said it all. First Isaac says
> You seem to be playing a definition word game. [...] Obviously productivity does approximate business value in a very meaningful way when it is defined in terms of delivered business value.
Then the author responds
> Martin asserts "any true measure of software development productivity must be based on delivered business value". I agree, and I propose there is no such thing. We're better off dropping metaphors altogether and just talking about what programmers do: Solve problems.
So I guess the author wants to replace the term "software productivity" with something like "problem-solving productivity, measured in terms of business value"...? Kind of silly IMO, but to each his own.
>It's a ratio of inputs to outputs.. even in the example the inputs and outputs are measurable.
So more lines of code is better!
Um, we know this doesn't work that way as a good measure.
This is like comparing algorithms that do the same thing to algorithms that do different things. You're not going to get good valid comparisons. Metrics for one thing may not work at all for another.
The output is measured with dollars, not lines of code. So are the inputs.
It's a perfectly cromulent measure so long as we understand the limitations of the measure. For example, trying to measure the productivity of a day or a sprint? That's silly. Measure the output of a team which does not produce an entire product? Won't work because you'd have to figure out how to apportion the productivity.
It's a perfectly good measure except it does not help us at all.
The whole reason for this discussion is situations like Microsoft having 200k employees and making $240B in a year. Which employees, teams, or even departments are more productive? They want to know.
And even if it did not matter, likely the expense of this year influences the income over multiple future years, so you compare the dollars in / dollars out for which periods?
"productivity, in economics, the ratio of what is produced to what is required to produce it. Usually this ratio is in the form of an average, expressing the total output of some category of goods divided by the total input of, say, labour or raw materials."
It complains, but doesn't offer a solution. It simply criticizes and says "all engineers cannot and thusly should not be measured".
The ironic thing is, the blog post is implicitly measuring by not explicitly measuring. The measurement is the bug ticket itself and whatever value attached to it.
But to this end, I generally agree. There are qualitative and quantitative measurements. Quantitative is the value of the ticket commonly ascribed by the team (scrum? agile? whatever). Qualitative should come up in review.
Qualitative is SO HARD. Top down? Team 360? Mixed? But it must be undertaken and refined by the team at each level of the org. Otherwise you will run into the exact situation described by the blog post and you won't know how to judge left from right, good from bad. Maybe the blog post's example isn't that great, too much information is missing to make a solid judgement, but you need to decide who to reward via promotion, annual raises and who to reprimand and who to not change.
But still, all systems are terrible, but you must pick one less it be picked for you.
In the scenario described in the article, Peter and Paul both achieve the same outcome in the same wall clock time. Obviously they are equivalently productive despite different working styles, by construction.
But this doesn't account for the more realistic examples of Prakash, who completely fails to deliver a working solution, or delivers half a solution, and Percy, who gets it done two weeks late. I'm pretty sure you can define a shit_done/time_elapsed productivity metric for those two guys that is worse than that of Peter and Paul.
Maybe I am a cynic but I suspect that some people are upvoting this because the framing makes them feel OK about getting paid six figures to work three hours a day...
Solving problems is productivity. That also means, solving them in a way that the solution doesn’t spawn new problems down the road, which is why we should follow the
best practices. I can’t say whether deleting 100 lines is better than adding 1K lines, because before passing that judgement I would need to see what exactly those 100 and 1K lines are. OP is arguing too black and white IMO, and draws the conclusion before I‘m sold on the premise.
Productivity in economics refers to how many units of output you can generate with a given amount of input, it doesn't take into account quality, usefulness, etc. Complaining "lines-of-code is not a sensical productivity measure" is too kind to the concept of productivity.
Using productivity as a metric leads to the same nonsequitur stuff in many many fields.
It works somewhat for industrial output when you're working with commodities. Or it can work in some more fields as an ancillary measure if you pair it with some other quality, customer satisfaction, outcomes etc measure. But usually you don't want to maximize work while getting good outcomes, you just want the good outcomes.
So if you have a developer who writes no code and closes no tickets, who essentially does nothing except come to meetings, they are just as productive as a developer who writes some code and closes some tickets? Obviously there is a difference there. Negative code is still doing something as well. Back in the 90s we used Personal Software Process from the SEI, and we measured lines added, deleted, and changed. (as well as defects removed and added)
It becomes clear that simple quantity of code and tickets is not enough- but it's also not nothing. Part of what is missing is quality and assessments of task complexity. Part of what is missing is the other parts of the jobs, like design and code reviews.
I don't think it's hopeless, and can at least be used to look into why some people don't seem to make much at all.
You could always look at things in a more abstract way ;)
From one point of view, the users of the software are supposed to enjoy so much of a productivity increase that it's not supposed to matter if the coders are as productive as they could be or not.
Give or take a few hundred percent at least.
I realize that most people who've been to business school still aren't going to develop the needed acumen to handle a situation like this.
Too many times the only training retained is a knee-jerk over-reaction to a fraction of a percent :\
Ever see one of these "leaders" have a cow and it was as stupid as it could have possibly been?
I'm confident there are still some natural leaders that can thrive without worrying about every ounce of nose to the grindstone for their staff.
The difficulty of measuring productivity is particularly felt by senior developers. They save time and effort in ways that are non-obvious, which might be measured by dependencies that were not added, design patterns that were rejected, or processes that they pushed back on. Just like with living things, unchecked growth is unhealthy for an organism but the actions required are difficult to measure. One could start an attempt with a counter-factual narrative, but this does not map cleanly to KPIs.
Asking a software engineer to be more productive is akin to asking a mechanical engineer to be more productive. What does that end up looking like? More useless blueprints? It turns out when you ask software engineers to just crank out code, you just get lots of code.
Good productivity is about being able to do all tasks as planned. That includes the management, as its task is to attribute tasks. And whoever creates tasks. And whoever defines the objectives those tasks should answer to.
Ok, but ironically the article does explain by a comparison of two alternatives just how to measure productivity—-solving problems most efficiently (least amount of maintainable code).
You can very well at least count the problems that were solved (or deleted) you can also probably measure the value those solutions have in revenue or another metric.
It's still true that measuring lines of code, time spent coding, commits or anything else is at best a proxy of productivity. It's also true that without any code changes problems more often than not don't get solved or we at least can't call the activity software development.
I do appreciate the little coda at the end - nice that the author was self-aware enough to realize where the cruft was and cut it (and courteous to the reader!)
Time spent on "incidental complexity" (new features, key fixes, performance) versus time spent on "accidental complexity" (anything and everything else).
Easy metric to understand, easy metric to teach, just remember that it applies to teams and not individuals.
And yet somehow, software developers are doing something that's worth paying them a salary. And it's not always impossible to tell if someone's being held back at a lower salary band than they should be, or was promoted into a salary band that's higher than they're able to keep up with.
Yeah, these complaints are always so duplicitous. Somehow, it’s impossible to measure programmer productivity when it’s time for accountability or professional standards, but not when it’s time to assign annual bonuses.
Wholeheartedly agree. SWEs are not factory workers. They don't punch in at 6am, put on their uniforms, make sure they meet their daily "lines of code" quota before clocking out at 10pm. We should not measure software productivity the same way we measure ball bearings production.
Using lines of code to track productivity is absurd (do people really believe it or is it just a strawman at this point?). I'm reminded of that midwit meme where a junior has very few lines of code written because they don't know the code base well enough, the midwit writes up an whole framework, and the senior engineer has a net negative lines of code contribution.
Theres also a fundamental difference between creating and maintaining code. Something like 10 guys wrote visicalc. Does that mean they were contributing millions of dollars in profit per hours? What about the maintainance to keep it going? Bug fixes? Patches? On call infra guys? What about opportunity cost of putting engineers on deadend projects?
My point is tracking productivity in software dev - maybe all knowledge work for that matter - is complicated. Maybe that's why there's so much "busywork" (emails, slack, tickets, meetings, etc). Everyone wants to look productive but no one knows what it means
"Put another way, productivity has no applicability as a metric in software.
"How much did we create today?" is not a relevant question to ask. Even if it
could be measured, productivity in software does not approximate business value
in any meaningful way. This is because software development is not an activity
that necessarily produces anything.
This is ridiculous, irrelevant, and wrong. Of course software development produces things. It produces software.
Which of these two developers was more "productive" today? The answer is: It doesn't
matter. What matters the that Peter solved the problem, while simultaneously reducing
long term maintenance costs for the team. Frank also solved the problem, but he
increased maintenance costs by producing code, and so (all other things being equal)
his solution is inferior. To call Peter more "productive" is to torture the metaphor
beyond any possible point of utility.
Ohhhhhhhhh. I get it. The author doesn't know what the word productivity means.
Productivity does not mean "increases business value while decreasing maintenance costs [and having no net negative impact in any way]". It doesn't even mean "solving a problem".
Productivity just means "to make something", or more specifically the rate at which something is made. That's all. You can make 10x more of something, and it can be garbage quality, but you did make it, and you did make more of it, so your productivity increased.
If you produce 10x more grain than you did yesterday, you are more productive. The grain might now be full of heavy metals, pesticides and toxins. But you did in fact produce more grain. If you were trying to measure productivity of usable, healthy, high-quality grain, that is a different measurement than just productivity of grain. You may assume everybody knows what you mean when you say "productive", but you'd be wrong.
Strawman argument. It would have been (semi) interesting if the title was "Software productivity can't be measured by lines of code" (but already very eloquently stated in https://folklore.org/Negative_2000_Lines_Of_Code.html ).
And yet … there are some engineers who manage to solve hard problems to get something new working and debug it, however long it takes, and many others who don’t and/or can’t.
I'm not sure I follow. So you failed to measure software productivity in lines of code, therefore it follows that "There's No Such Thing as Software Productivity"? Don't you think that giving up after n=1 attempts at measuring software productivity might be a tad too fast to draw a generalized claim of impossibility? I might argue the real lesson learned is "Lines of Code are Not a Measure of Productivity in an Isolated, Toy Example".
I suspect this sort of thing gets promulgated because it kind of massages our ego, like yes, they can measure other sorts of productivity, but not ours, oh no, we're too complex and intelligent, there's no way to measure the deep sorts of work that we do! Which, yes, OK, we're not exactly bricklayers, but surely, if you had to, you could do better.
> it kind of massages our ego, like yes, they can measure other sorts of productivity, but not ours, oh no, we're too complex and intelligent, there's no way to measure the deep sorts of work that we do! Which, yes, OK, we're not exactly bricklayers, but surely, if you had to, you could do better.
Productivity is not a cut-n-dry stat for any kind of knowledge work.
For an admin, we could say they were more productive if they handled more cases than last month. But this does not account for tricky cases or ambiguous cases, which might have taken 2x time to sort out but on rote numbers it looks they were less productive.
For us, there are no specific goals(except arbitrary imaginary deadlines by clueless management and sprint points) to correctly measure productivity. A junior engineer may spend 4h each day blasting out many lines of codes, a mid/senior may spend more hours in reviews or meetings. A 5 line PR with test coverage is faster to review than a 800+ LOC PR with additional tests and have significant risk of breaking something, so review number is also not a good indicator. I need to coordinate with min. 12 people to get any new credentials or access to specific env owned by another team, which is not countable as productive but unavoidable. How about that garbage meeting where I was just spending time yawning because some new tech lead who believes that a modular monolith system is better if it was rewritten as event based. Productivity measurement is hard.
To take your envy(I sense it on the quoted remark) to another place, think about how you go about measuring productivity of a CEO, PO, PM, Scrum Master, Agilist/Agile-Specialist etc?
Sure, productivity it's not cut-n-dry stat, but it's not completely impossible to measure either. For a team, there are very clear goals: to make product that customers like, and to keep improving it for a long time. The last part is especially hard to measure, so people do all sorts of approximations, some very bad.
For individual, productivity is measured on multiple axis - new long-term features, throw-away and prototypes, code maintenance, fire-fighting, code reviews, high-level design, outside-of-team communications, intra-team communications (for example helping teammates), archeology, specific parts of your project, probably others. In ideal team, everyone is strong on at least one or two axes, so no person is worse than others. In real teams, I've seen people who were weak on every single axis and did not seem to improve with time. I've also seen people who are great at everything, and it's extremely nice when such people exist - but they are very rare.
(Btw, re your specific example: if you weren't productive because you were waiting for others to approve your access then you were not productive. There is nothing complex about this, there were plenty of times when I've said, "I did nothing for project S last week because they still could not give me access")
For a team, there are very clear goals: to make product that customers like, ...
With respect, that goal is far from clear. How do you measure how much they like the product? Have we achieved more "likes" this sprint than last?
Customers like a product because it removes their problems. QED.
> To take your envy(I sense it on the quoted remark) to another place, think about how you go about measuring productivity of a CEO, PO, PM, Scrum Master, Agilist/Agile-Specialist etc?
I have no envy; I'm a software engineer as well. I just don't seem to struggle, as others claim to, in measuring productivity. I find it fairly straightforward to see that some people are more productive than others, at least in my workplace where we can hold most things constant (e.g. meeting count, managerial ability, etc, etc). Yes, there may be external factors, and that's unfortunate. Yes, it's not fair that some people are more productive than others despite putting in half the effort. But I'm not going to stick my head in the sand and pretend it's invisible.
> I just don't seem to struggle, as others claim to, in measuring productivity.
Because you are measuring at a very broad and basic level.
Steve is more productive than Susan.
Great. How much more productive? Can you turn it into a number?
Can you still do it consistently when Steve and Susan are in different teams in different parts of the organisation trying to achieve different goals?
I've done DB upgrades that took 10 minutes and I've done DB upgrades that took 3-4 months. What changed was not my productivity but the nature of the problem. Yet from the outside they were both just DB upgrades.
If Susan had done the DB upgrade in 12 weeks could we confidently claim that Steve could have done it in 11 weeks? Steve hasn't even done a DB upgrade since he joined the company. Perhaps Steve could have done it in 10 minutes?
I don't think anyone can get numbers, but partial ordering is much easier.
If Steve and Susan are in different part of organization, the answer is "cannot compare". If they are doing different job, the answer is the same.
But every once in a while there is a scenarios when you can compare people easily.
There has a weekly rotation to be an support person for other team. During his week, John always answers questions quickly and to the team's satisfaction. Meanwhile James struggles to answer them and cannot troubleshoot product his team is writing. This has been going on for multiple months and hundreds of questions for each, so it's not "bad week" unlucky or fluke. We now know who is better at answering questions about product.
John and James are doing DB migrations, they did many dozens of them. The migrations are assigned randomly. But John is usually finishing his migrations with no problems, while James often caused outages or missing data. A few times James took over two months to migrate, so the task was taken from him and given to John. John had to discard everything James did and migrate everything from scratch. Now there is a migration for a very important client and CEO is fed up with random assignment.. who is he going to choose?
What your scenario doesn't address is that while John finished his migrations on time, James has designed the flagship order processing pipeline something that John could never pull off.
Or maybe while John is technically adept, he's also a huge jerk and belittles people at standup, while James is the quintessential communicator with jr devs, etc.
Real life is messy. I've seen more people get replaced due to attitude or teamfit issues than specifically due to technical incompetence.
I would argue that here you are talking less about productivity and more about basic competence?
"What changed was not my productivity but the nature of the problem"
I think that's the source of the problem: it's impossible to measure the "work" required to solve most software problems. If you tell me to carry a stone up a hill, I can put that in a formula and know exactly how much work I'll have to do. But if you give me a ticket to do a DB upgrade, I can, at best, make an educated guess.
So by the time I close the ticket, how much work have I done, and how do I know whether the time I've spent it proportional to the work I've done?
> Great. How much more productive? Can you turn it into a number?
This is moving goalposts. OP's argument was "There's No Such Thing as Software Productivity", not "You Can't Convert Software Productivity into a Floating Point Number With 3 Decimals of Accuracy."
There’s no real dependable/reproducible single linear measure of software productivity”.
Would that be a fairer assertion?
By that measure there also isn't a real dependable/reproducible scalar that measures athleticism. Nonetheless some people are clearly more athletic than others. Also within a single sport we can easily see that some players are better than others. Here too you could object that people who play offense should not be compared to players in defense. Or that it's not individual players that matters but the team as a whole. And yet, we can still figure out easily who the star players are.
The ability to look productive is a different skill. The ability to create any impression is a skill some have. Creating fires and putting them out like a hero is another skill.
What are your eyes measuring? Are you being fooled?
If you are IC, it's pretty easy to see if your teammates are "looking productive" vs "are really productive", as you know all the code and how it works.
For other teams yeah, it's harder, or maybe even impossible if the teams are isolated.
> If you are IC, it's pretty easy to see if your teammates > are "looking productive" vs "are really productive"
I'm not sure how true this really is. We all like to think that we're good at estimating other people's ability, particularly if we work closely with them, but how can you validate that impression?
Something that happened to me at one point in my career was that I joined a team that was half new hires with little experience, and half long-term employees with a lot of experience. Because the long-term employees were typically extremely busy, I ended up answering a lot of questions from the entry-level devs. So I spent time documenting the system, doing trainings, doing 1-on-1 code reviews, and so on.
During the performance review, feedback from the entry-level devs was that I was highly productive, and feedback from the long-term employees was that I was a slacker.
I've been on all sides of this issue and I guarantee that as an individual contributor you lack the perspective and visibility into all of the factors that make up net productivity for your colleagues. Once you get a little more maturity and experience you'll probably understand this better.
No offense but reading this comment with the utmost attempt at interpreting it in good faith, it kind of sounds like you're just going off of vibes and can't actually quantify productivity any more than anyone else can. Otherwise... How exactly are you quantifying it? If you had to show your work, could you?
I dunno, the opposite side seems equally absurd to me. Do you work at a software company? Do you have coworkers? Are you honestly telling me that, gun to your head, you couldn't say that some are more productive and some are less? That if you had to choose a co-founder for your next startup, that you would have NO idea where to begin, that any of them would be equally as good as the next?
I mean look, some people seem to be extremely productive, and sure, I could say that those people are almost surely more productive than others. It's still genuinely very hard to quantify if those people are actually vastly more productive, or if it just looks like that because they produce more obvious artifacts of their productivity. Hell, what is productivity, is it more productive to fix 1000 bugs or to be the tech lead on a product that reaches 1 million ARR? Does it matter what bugs?
I stand by what I said, whether people like it or not.
Including the part about going off vibes. Please explain how this "you can just kind of tell" mentality is not literally just going off of vibes?
(And I wouldn't choose a cofounder only based on productivity, anyways. I'm not even sure that would be the main criteria.)
How would you choose a cofounder? I would call whatever metric you use to determine "would that person be a good cofounder" productivity. (OK, sure, let's narrow it down and say that you were tasked with choosing a purely technical cofounder; perhaps in this hypothetical you're helping your non-technical friend find one?) Do you still believe you couldn't do better than chance?
Well, a lot of the factors explicitly have nothing to do with general productivity at all, such as alignment. The goal with regards to selecting for productivity would be "someone who seems productive enough" but getting hung up on exactly how productive feels like a mistake, as aside from being unquantifiable, that isn't going to be the only or probably even main factor that decides the fate of the endeavor. Productivity is not the most challenging problem in software development. In my opinion, as far as factors for success for a technology endeavor goes, even overall technical competence winds up factoring in fairly low at the end of the day. What's important is meeting the threshold you need, not getting as high of a score as possible.
I should clarify that I actually agree with you if what you're saying is that the best measure of productivity we have is literally just going off your gut, but my argument is that this gut feeling is terrible. This is in large part because humans are biased, our gut feelings are swayed by things that simply shouldn't factor in. We tend to have more positive opinions of people that we think are like us and we sometimes wind up having negative opinions of someone based on stupid things like disagreeing with them on something that is ultimately irrelevant to whether or not they are productive.
And to me, the ultimate nail in the coffin is really in the question of what really constitutes productivity. Productivity is supposed to measure the efficiency of producing something, which already has pitfalls in and of itself when dealing with things that do have discrete, measurable indicators of progress, but programming doesn't, it's not even always obvious if progress is forward or backward sometimes. What's most important, performance optimizations, disaster planning, features, robustness, minimizing resource utilization? The best answer you can generally say is "It depends," and moreover, everyone will have a different set of competencies they're best at, so a person doesn't really have some single "productivity value" you could summarize them with. Yet, all of those things are pretty important for any serious technology organization, so you would want people with a variety of different affinities.
At that point, it starts to beg the question of whether or not attempting to directly measure software productivity is a worthwhile endeavor. I'd argue not.
I suppose the question is: can you supply a word you would use when deciding on who to hire for a technical role? Does anyone off the street contribute the same amount of that word, or would you be selective?
> Don't you think that giving up after n=1 attempts at measuring software productivity might be a tad too fast to draw a generalized claim of impossibility?
Software developers should look at anyone claiming they can measure software productivity as a snake oil salesman.
We have seen hundreds of attempts over the years and they have all "failed". More accurately they all have large error bars and biases.
Researchers can and should continue looking into how to measure software development productivity. It is likely over the next few decades we will start to understand how to measure it appropriately.
I agree. I know from looking at a story that if Alice picks it up it’ll take her about a day and be decent quality, but if Bob picks it up it’ll take him about a week and be worse quality. And I know that this will be pretty consistent week-in, week-out.
Developer productivity exists and I am totally comfortable describing Alice as being more productive than Bob. The fact that there isn’t a good, generic way to distill this into particular values does not mean that the phenomenon doesn’t exist.
We’ve all worked with Alice and Bobs. Claiming there’s no such thing as developer productivity is denying what we’ve all experienced. We aren’t special snowflakes whose work is beyond the ken of mortal men. Our work has value and sometimes we produce a lot of value in a particular time period and sometimes we product a little. This is productivity.
This is not an isolated toy example, it's reductio ad absurdum (https://en.wikipedia.org/wiki/Reductio_ad_absurdum) - and until today I haven't seen a way to measure software productivity (might also apply to other knowledge work) that is resistant to that...
Ok, but then what is a way to do it?
The text gives an example to the core problem, and to argue differently requires thinking around it.
In practice. I’ve seen many attempts at measuring productivity, but once you dig into them, you see they are just abstraction mechanisms above something that is similar to lines of code.
I have yet to see an idea that sidesteps the core issue described in this post. Also, it applies to many types of work, and software is not unique in any way.
Happy Customers.
All other measures are a proxy for happy customers.
Actually, happy customers is also a proxy (the real measure is profits) but measuring profits directly (in the short term) can lead to decisions that have adverse long term effects. It's too easy to increase profits in the short term by avoiding long-term expenses.
So, if you're in the business of software, the goal is happy customers. (And I use the word Customers carefully here. Not just Users who pay nothing, but Customers who spend money.)
In a business context, it's really the only thing that matters. But, of course, it can be hard to measure (are they Happy?) and relies on multiple disciplines. Production (coding), Marketing, Sales, Support, Documentation, Training- all need to be working well to make it work.
Ultimately if the big picture doesn't lead to Happy Customers (again, I stress, in a businesses context) then no-one is "being productive."
When a measure becomes the objective, it ceases to be a good measure.
Even if you could perfectly measure customer happiness (very hard, as you note) - it's relatively easy to make customers happy by giving them more value than what they pay for. Sure, that may cost your business more money than what it makes with said customers, but hey, who cares, "profit" was not the metric...
(and as you note, if you make "profit" the metric, that has its own set of challenges - e.g. the optimization towards short-term profit in detriment of the long-term sanity, which is what we observe in a lot of corporations).
It’s demand. How much demand is there for your product
As the other commenter pointed out, happy customers means nothing if they aren’t actively paying you
Arguably the business is aiming for Paying Customers, not necessarily Happy ones :)
I recommend measuring job satisfaction instead of developer productivity. It's the "least bad" proxy metric I know of. https://redfin.engineering/measure-job-satisfaction-instead-...
The interesting corollary to this approach seems to be that productivity barriers are largely external.
A potential riske seems to be feedback systems where job satisfaction is determined by high or low pay.
As already mentioned, people measure value return by revenue gain. It is irrelevant to attribute it to some construct like a line of code.
profit generated I think is the high level one, and then you want to dig from there into how much the software development contributed to this.
That's very tricky thing to quantify, especially with "unsung heroes". If my work is in preventing problems, the guy that fixes problems will be seen as the one that contributes more to profit, since impact is directly observed/measured.
This is something that one of the orgs I worked for eventually realized. The people f'ing up, and then fixing their mistakes, were the ones getting promotions/bonuses/raises, because they were the ones interacting with all the execs.
Revenue per employee $ spent.
I think people get hung up on wanting to collapse developer productivity into a single dimension, usually for stack-ranking purposes. This, I think, is always going to punish good engineers and reward bad ones to some degree.
Measuring developer productivity should, in my opinion, have one dimension for speed, one for quality, and one for user impact. LOC can be fine as a measurement for speed, you just don’t want to look at it in isolation. You would want to also measure, for example, escape rate and usage for the features the developer worked on, and be willing to change or refine these if circumstances require it.
You also need to look for different profiles based on the developer’s level of seniority. A senior dev probably shouldn’t be writing as much code as a contributor, but their user impact should be high, and escape rate low. Analyzing differences between teams is important, as well. A team that has a lot of escapes or little user impact probably has issues that need management attention, and may not have anything at all to do with individual developer productivity or ability.
In brief, the numbers are there to help you make better management decisions, not to relieve you of having to make them.
Bricklayers have measures that are usable by people who know little to nothing about the profession who are incapable of doing the work. This is what people want. An objective measure that can be applied by people functionally incapable of actually doing the work.
Wherein the work isn't repeating existing work no such measure should be expected to exist.
Any creative work is going to suffer from the same problem.
If you asked a brick layer to develop a better workflow for his fellows to work more effectively in a particular sitiation it would be the same. Are you going to measure words per minute?
The problem is that corporate types don't see programming as creative work. Type faster!!!! If you push the buttons faster the product gets build faster!
Also imagine every other week as the bricklayer had half a wall built a PM comes out and says, "tear this all down and move it 3 feet to the north", and then the week after, "now move it east 1 foot, and why is it taking you so long?"
Function point analysis works well enough for measuring software productivity in most domains, provided that quality is held constant. Like other software productivity metrics it's only meaningful at the level of a complete product team and worse than useless for evaluating individual team members.
https://doi.ieeecomputersociety.org/10.1109/32.44364
So you can measure productivity with a reasonable level of accuracy and consistency, but then what? For most organizations it's not actionable so calculating that metric ends up as another form of waste.
> Function point analysis works well enough for measuring software productivity in most domains, provided that quality is held constant
Well enough for what? And how do you measure quality?
You certainly did fail to see the idea being conveyed by a simple, but poignant anecdote. If this is all you have to say, I might argue the real ego being massaged is your own; by yourself.
You can say this about any discipline. The root of the issue is that productivity for productivity's sake is meaningless, and it makes no sense to measure productivity as a general property when outputs vary.
A tire factory has a distinct, singular goal: produce tires. It does this continually. Productivity is meaningful, but only in relation to a target that is typically specified by externalities (e.g. amount of demand)
A software company is usually not in the business of producing consumable commodities so this kind of measurement does not make sense. It can make sense to measure productivity during a period for delivering a particular piece of software within a given time bound, but once it's delivered, productivity becomes meaningless. You always need to understand productivity in relation to some purpose and I don't know how these knuckleheads who think this abstract idea is basically like a concrete measurable essence, like mass, or liquid, got leadership positions.
Seems to me you’re talking about measuring software. Once we pick a measure though, we can calculate output/input (productivity).
Imagine two people tasked with producing the same software, or software satisfying the same requirements or test suite. What would you call the person who produces it faster? More productive?
Measuring software in financial terms or lines of code might not be the right measurement of software in all situations, but surely we can measure time and cost to produce software or software that satisfies equivalent requirements.
You’re responding to the headline, not the article. The article has a plainly stated thesis: “Even if it could be measured, productivity in software does not approximate business value in any meaningful way”
The article contains this sentence, yes, but the rest of the text does not support it.
The article argues that lines-of-code is bad metric, and no one argues about this.
thuanao proposes measuring "number of completed projects", and I think it's a pretty good metric in some situations which don't involve long-term maintenance.
One example I can come up with is data science: two engineers are given identical requirements to ingest the data (once) and calculate certain metrics. If the assignments are identical, we can calculate and compare their productivity, in "projects/month"
Long-term maintenance is a confounding factor of course. People sometimes write software faster with the downside that it wil be harder to maintain.
So remove the confounding factor (at the same time removing most projects we would like to apply this to) and it becomes an easy problem?
But there are many confounding factors. Your data science project results will be different (different metric values outputed) and you don't know which one is correct, or closer to correct. Now how does the "projects/month" number look?
Or maybe one of the engineers uses way more compute-hours for getting the job done. Their projects/month number is better, but is that really a better outcome? Compute is not free.
By the time you remove most confounding factors, the productivity measure will only apply to an insignificant number of projects.
Depends in the coming months and years which was more stable, maintainable and extensible.
Isn't the output, the singular goal, always money? (As sad as that may be.)
Money is just a means to obtain goods and services. So no... money isn't really the goal.
Yes, but people lose sight of this.
I suppose in the context of this thread, we can look at the productivity of a system that includes money or does not.
A small system, like a tire production line, can measure its productivity in terms of money, which is outside the system. But a large system, like society, includes money, and cannot measure its own productivity using internal things.
Money is the measure of usefulness.
By the way, why suddenly the money=bad sentiment is so popular? I thought USSR example loudly showed us what happens when people think that money is evil.
"Money is the measure of usefulness?"
Good lord. There's so much wrong with this statement that I doubt I can meaningfully respond to it. I'll try, though.
If we have to reduce measures of usefulness to a single metric, then why not percentage of users who respond favorably?
I daresay that people find air quite useful, and for the most part, air is free.
Suddenly? Who first said money was the root of all evil?
When companies focus on profit at the expense of all other goals, the result is typically disadvantageous to customers, employees, and society at large.
See also: crapification; financialization; parasitic private equity
I can't speak for HN, but I imagine some readers are more interested in the benefits of cool technology beyond putting more money in the pockets of investors and management.
> I would argue that what good software developers do is remove problems. The opposite, in fact, of production
But something is being produced - it is version 2.0 of the software. This is an artifact that is then shipped to users or deployed to a server. Peter’s solution fixed the issue and did not (seemingly) create further maintenance burden, which would have taken attention away from other tasks, i.e. reduced future productivity.
I agree that metrics for programmer productivity are often useless (e.g. using lines of code is a bad idea for obvious reasons), but it seems silly to claim that the entire concept of productivity does not apply to the production of software.
Silly and convenient. No one takes the claim that you can’t measure software productivity seriously and everyone simultaneously agrees that simple scalar metrics often fail to show the big picture in any and all disciplines.
The rest is just usual software guy hubris and lack of awareness of the discipline.
> No one takes the claim that you can’t measure software productivity seriously
At least three people do, and they are mentioned in the article: the author, his colleague, and Martin Fowler.
I suspect that software productivity (like the productivity of scientists, artists, composers, etc.) on any non-trivial project may be measurable on a scale of years, but not necessarily months or calendar quarters. A larger issue is that productivity is often due to external factors as much as it is to individual effort, and overnight success is often the result of extended periods of limited progress or even repeated failure.
Let's say that you have two runners running the same marathon. The first one, Frank, sprints at full speed and eventually tires out and slows down. The second runner, Peter, takes a nap first and then finishes the marathon at the exact same time. Which of these two runners was faster in the race? The answer is: It doesn't matter. And therefore there is no such thing as running speed.
There is such a thing as productivity in programming. If you could measure it it would likely be some combination of peer review and an analysis of the impact implemented features and fixes had. Some companies actually have programmers rate each other. I don't know how well it works and I think it can lead to perverse scenarios. But you can come up with metrics that are positively correlated with productivity.
The answer is neither is faster, because they finished at the same time.
Thats why bugs and chores have 0 points per default in pivotal tracker. Thats also why there was a push to force us say "as a user i..." at the beginning of story descriptions to make sure this is something creating user value. With these guards in place i don't follow the argument that productivity is not measurable, if a team builds features that solve an expressed user need and lets even say go through a final user acceptance check, this is a productive process and it is also very measurable!
Isn't this whole debate a repeat of the old talking point about what is value? Labour theory of value, that kind of thing?
This is simply never going to end if carried on along the lines of this article, or along the lines of most of the comments here.
There's no way to objectively and reasonably put a value on something.
All we have is a theory of subjective value, which does a bit of handwaving about utility and works out some ways where we can come to a price, regardless of the fact that the values in peoples' heads are subjective.
Thus the distinction between knowledge work and "tangible work" like bricklaying is actually a moot point. Yes, you can measure "productivity" of bricklaying in metres per day, but ultimately you care about value, not amounts of wall.
The arguments about one guy definitely being more productive than another are similar. One person values speed, another values maintenance costs downstream. It is subjective what ought to be more important.
Impact is the only useful metric.
It is staggeringly hard to measure.
Output is a weak proxy for impact. But it’s the one that makes intuitive sense to people. Doesn’t make it right or useful. I’m sure you all can envision a parable about your subfield of expertise that showcases how a seemingly light touch has a huge positive impact.
> It is staggeringly hard to measure.
It's straight up impossible. Best we can do is observe and attempt to measure proxies.
Take hypothetical situation: VPs A & B debate a business decision. Let's say A wins the argument and their solution leads to a revenue increase of $10M, and let's say we can confidently state that this is the outcome driven primarily by them winning the argument. Is the impact a net growth in business of $10M? In a sense, yes; but in some other sense, perhaps if they went with B's solution, the revenue growth would have been $15M? There's probably no way to know for sure unless you try both approaches, which is often impossible...
My theory is that impact is heavily context dependent.
If you can solve a problem quickly, at the right time, with buyin from your org, that’s positive impact.
That general case differs wildly when you descend to the particulars.
I think this is just a semantic quibble based on a narrow interpretation of the word "produce".
Because it also means:
"cause (a particular result or situation) to happen or exist."
Yeah, the comments on the post already said it all. First Isaac says
> You seem to be playing a definition word game. [...] Obviously productivity does approximate business value in a very meaningful way when it is defined in terms of delivered business value.
Then the author responds
> Martin asserts "any true measure of software development productivity must be based on delivered business value". I agree, and I propose there is no such thing. We're better off dropping metaphors altogether and just talking about what programmers do: Solve problems.
So I guess the author wants to replace the term "software productivity" with something like "problem-solving productivity, measured in terms of business value"...? Kind of silly IMO, but to each his own.
Clicked for the bait, stayed for the background tile.
Sure there is. It's a ratio of inputs to outputs.. even in the example the inputs and outputs are measurable.
The only thing this article gets at is that engineers may not know how to calculate their own productivity; but it doesn't means it's not calculable.
>It's a ratio of inputs to outputs.. even in the example the inputs and outputs are measurable.
So more lines of code is better!
Um, we know this doesn't work that way as a good measure.
This is like comparing algorithms that do the same thing to algorithms that do different things. You're not going to get good valid comparisons. Metrics for one thing may not work at all for another.
The output is measured with dollars, not lines of code. So are the inputs.
It's a perfectly cromulent measure so long as we understand the limitations of the measure. For example, trying to measure the productivity of a day or a sprint? That's silly. Measure the output of a team which does not produce an entire product? Won't work because you'd have to figure out how to apportion the productivity.
It's a perfectly good measure except it does not help us at all.
The whole reason for this discussion is situations like Microsoft having 200k employees and making $240B in a year. Which employees, teams, or even departments are more productive? They want to know.
And even if it did not matter, likely the expense of this year influences the income over multiple future years, so you compare the dollars in / dollars out for which periods?
That's not the definition.
Eg https://www.britannica.com/money/productivity says:
"productivity, in economics, the ratio of what is produced to what is required to produce it. Usually this ratio is in the form of an average, expressing the total output of some category of goods divided by the total input of, say, labour or raw materials."
> It's a ratio of inputs to outputs.. even in the example the inputs and outputs are measurable.
But reality is never that clear cut. How’s the ratio look when:
- Peter goes to the park and the breakthrough doesn’t come?
- Or it comes 3 weeks later?
- Or he deletes 100 lines of code and introduces a new bug?
I realized why this post rubs me the wrong way.
It complains, but doesn't offer a solution. It simply criticizes and says "all engineers cannot and thusly should not be measured".
The ironic thing is, the blog post is implicitly measuring by not explicitly measuring. The measurement is the bug ticket itself and whatever value attached to it.
But to this end, I generally agree. There are qualitative and quantitative measurements. Quantitative is the value of the ticket commonly ascribed by the team (scrum? agile? whatever). Qualitative should come up in review.
Qualitative is SO HARD. Top down? Team 360? Mixed? But it must be undertaken and refined by the team at each level of the org. Otherwise you will run into the exact situation described by the blog post and you won't know how to judge left from right, good from bad. Maybe the blog post's example isn't that great, too much information is missing to make a solid judgement, but you need to decide who to reward via promotion, annual raises and who to reprimand and who to not change.
But still, all systems are terrible, but you must pick one less it be picked for you.
Why do you need to evaluate people constantly and pit them against each other ? Why not give a raise to everyone and see what happens ?
It's how we sent rockets to the moon, it seemed to work ok.
In the scenario described in the article, Peter and Paul both achieve the same outcome in the same wall clock time. Obviously they are equivalently productive despite different working styles, by construction.
But this doesn't account for the more realistic examples of Prakash, who completely fails to deliver a working solution, or delivers half a solution, and Percy, who gets it done two weeks late. I'm pretty sure you can define a shit_done/time_elapsed productivity metric for those two guys that is worse than that of Peter and Paul.
Maybe I am a cynic but I suspect that some people are upvoting this because the framing makes them feel OK about getting paid six figures to work three hours a day...
Also, there's no such thing as sculpture productivity because all sculptors are doing is removing material.
Solving problems is productivity. That also means, solving them in a way that the solution doesn’t spawn new problems down the road, which is why we should follow the best practices. I can’t say whether deleting 100 lines is better than adding 1K lines, because before passing that judgement I would need to see what exactly those 100 and 1K lines are. OP is arguing too black and white IMO, and draws the conclusion before I‘m sold on the premise.
Productivity in economics refers to how many units of output you can generate with a given amount of input, it doesn't take into account quality, usefulness, etc. Complaining "lines-of-code is not a sensical productivity measure" is too kind to the concept of productivity.
Using productivity as a metric leads to the same nonsequitur stuff in many many fields.
It works somewhat for industrial output when you're working with commodities. Or it can work in some more fields as an ancillary measure if you pair it with some other quality, customer satisfaction, outcomes etc measure. But usually you don't want to maximize work while getting good outcomes, you just want the good outcomes.
So if you have a developer who writes no code and closes no tickets, who essentially does nothing except come to meetings, they are just as productive as a developer who writes some code and closes some tickets? Obviously there is a difference there. Negative code is still doing something as well. Back in the 90s we used Personal Software Process from the SEI, and we measured lines added, deleted, and changed. (as well as defects removed and added)
It becomes clear that simple quantity of code and tickets is not enough- but it's also not nothing. Part of what is missing is quality and assessments of task complexity. Part of what is missing is the other parts of the jobs, like design and code reviews.
I don't think it's hopeless, and can at least be used to look into why some people don't seem to make much at all.
You could always look at things in a more abstract way ;)
From one point of view, the users of the software are supposed to enjoy so much of a productivity increase that it's not supposed to matter if the coders are as productive as they could be or not.
Give or take a few hundred percent at least.
I realize that most people who've been to business school still aren't going to develop the needed acumen to handle a situation like this.
Too many times the only training retained is a knee-jerk over-reaction to a fraction of a percent :\
Ever see one of these "leaders" have a cow and it was as stupid as it could have possibly been?
I'm confident there are still some natural leaders that can thrive without worrying about every ounce of nose to the grindstone for their staff.
Some things you just can't fix.
The difficulty of measuring productivity is particularly felt by senior developers. They save time and effort in ways that are non-obvious, which might be measured by dependencies that were not added, design patterns that were rejected, or processes that they pushed back on. Just like with living things, unchecked growth is unhealthy for an organism but the actions required are difficult to measure. One could start an attempt with a counter-factual narrative, but this does not map cleanly to KPIs.
Asking a software engineer to be more productive is akin to asking a mechanical engineer to be more productive. What does that end up looking like? More useless blueprints? It turns out when you ask software engineers to just crank out code, you just get lots of code.
Good productivity is about being able to do all tasks as planned. That includes the management, as its task is to attribute tasks. And whoever creates tasks. And whoever defines the objectives those tasks should answer to.
Optimizing task completion is different from optimizing task benefit.
the problem is that software does not directly produce value, it is a way to place a bet on what will be valuable
Pretty spot on and that is why devops stuff and automatic tests are so important. Those just let your bet faster.
Ok, but ironically the article does explain by a comparison of two alternatives just how to measure productivity—-solving problems most efficiently (least amount of maintainable code).
You can very well at least count the problems that were solved (or deleted) you can also probably measure the value those solutions have in revenue or another metric.
It's still true that measuring lines of code, time spent coding, commits or anything else is at best a proxy of productivity. It's also true that without any code changes problems more often than not don't get solved or we at least can't call the activity software development.
I do appreciate the little coda at the end - nice that the author was self-aware enough to realize where the cruft was and cut it (and courteous to the reader!)
Time spent on "incidental complexity" (new features, key fixes, performance) versus time spent on "accidental complexity" (anything and everything else).
Easy metric to understand, easy metric to teach, just remember that it applies to teams and not individuals.
See: Out of the Tar Pit.
In an economic sense there is absolutely software productivity, when it results in more outputs from the same economic inputs.
This is fictious btw, or logic as follows is delete all code, deploy nothing, all bugs fixed, and achieve infinite productivity.
And yet somehow, software developers are doing something that's worth paying them a salary. And it's not always impossible to tell if someone's being held back at a lower salary band than they should be, or was promoted into a salary band that's higher than they're able to keep up with.
Yeah, these complaints are always so duplicitous. Somehow, it’s impossible to measure programmer productivity when it’s time for accountability or professional standards, but not when it’s time to assign annual bonuses.
Yep it’s just nonsense. If it’s hard to define or measure I guess it doesn’t exist?
Yay Ben Rady. He has a great podcast with Matt Godbolt (who made Compiler Explorer) called Two's Complement.
Wholeheartedly agree. SWEs are not factory workers. They don't punch in at 6am, put on their uniforms, make sure they meet their daily "lines of code" quota before clocking out at 10pm. We should not measure software productivity the same way we measure ball bearings production.
Using lines of code to track productivity is absurd (do people really believe it or is it just a strawman at this point?). I'm reminded of that midwit meme where a junior has very few lines of code written because they don't know the code base well enough, the midwit writes up an whole framework, and the senior engineer has a net negative lines of code contribution.
Theres also a fundamental difference between creating and maintaining code. Something like 10 guys wrote visicalc. Does that mean they were contributing millions of dollars in profit per hours? What about the maintainance to keep it going? Bug fixes? Patches? On call infra guys? What about opportunity cost of putting engineers on deadend projects?
My point is tracking productivity in software dev - maybe all knowledge work for that matter - is complicated. Maybe that's why there's so much "busywork" (emails, slack, tickets, meetings, etc). Everyone wants to look productive but no one knows what it means
Productivity does not mean "increases business value while decreasing maintenance costs [and having no net negative impact in any way]". It doesn't even mean "solving a problem".
Productivity just means "to make something", or more specifically the rate at which something is made. That's all. You can make 10x more of something, and it can be garbage quality, but you did make it, and you did make more of it, so your productivity increased.
If you produce 10x more grain than you did yesterday, you are more productive. The grain might now be full of heavy metals, pesticides and toxins. But you did in fact produce more grain. If you were trying to measure productivity of usable, healthy, high-quality grain, that is a different measurement than just productivity of grain. You may assume everybody knows what you mean when you say "productive", but you'd be wrong.
Strawman argument. It would have been (semi) interesting if the title was "Software productivity can't be measured by lines of code" (but already very eloquently stated in https://folklore.org/Negative_2000_Lines_Of_Code.html ).
And yet … there are some engineers who manage to solve hard problems to get something new working and debug it, however long it takes, and many others who don’t and/or can’t.
Productivity to me is about choosing the best tradeoff while minimizing the costs of going ahead.
And yet, Jeff Dean is significantly more productive than I am.
[dead]
[dead]
[dead]