Work Is Work(codahale.com)
As a small part of a massive digital transformation project last year, it was clear to everyone involved that no amount of good organizational design could succeed without significant adjustments to organizational behavior and culture.
We made all these logical plans but implementing them was about as effective as dipping them, bloodsoaked, into the piranha pit of legacy culture.
> A better model for staying informed of developments as the organization scales is for groups to publish status updates as part of the regular cadence of their work.
I mean, it sounds great. But politics just eats this stuff alive. And the more I felt empathy for the various factions, the more I realized that real, honest change takes unrelenting and sheer force of management will. Which is itself a kind of culture.
>> honest change takes unrelenting and sheer force of management will
Having also experienced this firsthand, I've come to learn that:
- there are people in the world who have difficulty changing, despite best of intentions
- there are people who are narcissistic, psychopathic, or toxic in other ways
- there are people who are lazy and are just clawing and scratching to keep their paycheque because it's basically free money due to not adding commensurate value back to the organization
- there are people who are in over their heads and find it difficult to improve, no matter how much money you may provide for training budgets, you're basically paying their salary to be full-time students at that point, not employees, and they still fail to graduate from school
Take all of the above, let it rot in a stale culture for a decade, you can easily see why an organization might get into some difficulty. Change is hard and unfortunately sometimes involves layoffs if the organization is to thrive (note I said organization, not company, affects organizations of all types).
Change is hard because people are hard. People are hard because... many reasons. Not everyone can be a driven Steve Jobs to achieve success, damn the torpedoes and other people's feelings, especially if they're just the average joe.
The easy answer is to fire these people. I don't like it, but I don't have a better answer. It sucks. I went through it before, refused to fire anyone for years, believed in people's ability to change. The scar tissue has caused me to completely change my mindset moving forward. If you can't change, you need to be on a performance improvement plan, if you can't achieve it, you're out the door, but hopefully we work at an organization willing to give you some nice severance so that you don't immediately end up on the street.
I was discussing this issue with somebody that lead an engineering transformation effort to drag an organization of around 2000 developers towards modern development practices. They had accreted so much tech debt and bad ways of working that their velocity was too slow to allow them to thrive in a changing market, it took forever to get things into production.
I would have expected that at least the developers would have been on board with this change but the amount of resistance was surprising. They ended up closing 2/6 of their offices, letting the bulk of those employees go and then opening a new office.
Big disruptive top-down initiatives to fix what isn't broken feel like politics on the receiving end too.
This post reeks of mathiness and pseudoscience and analogies taken too literally. Sure there are some interesting ideas, but without evidence, there's not much to it. You want to say that communication costs grow quadratically? You could "show" it with an argument about dyads, but that's not the kind of thing to believe without data. Or that productivity per person is at best constant as an org grows. Or most assertions in this post.
One of the obvious issues is that human communication is not just point to point.
We can all get in a room together or read an email chain at once.
This is a very interesting article, and it contains something that seems true at face value but that I feel compelled to disprove:
> The work capacity of an organization scales, at most, linearly as new members are added.
Let's go back to separating cotton fibers from its seeds, a laborious process. If you had a factory of 100 workers doing this and added another one, then you can expect the work capacity to increase 1%. But what happens if you add a worker that is called Eli Whitney, who makes a machine that automatically separates those elements? You definitely don't get a 1% increase in capacity since the new combination of everyone's capacity is more than 101%.
So the author goes on explaining that capacity is not the same as productivity, that the important task is to change the work itself. But is capacity measured only as the hours worked? An extra CPU processor does not do the same as an extra GPU or TPU processor. I've been playing with Coral.ai and I'm totally mindblown at how my raspberry pi can do ML!
Going back to the company, if your bottleneck in the company is e.g. devops and you hire a devops person, then productivity will increase super-lineally; because with this being the bottleneck it means that other resources were underused. So the new person will not add +1/n of capacity. Sure they might add 1/n of the _working hours_, but the capacity itself (and in all likehood the productivity) will increase more than 1/n.
I don't think what you're saying contradicts the article.
Especially the example of the devops bottleneck -> hire a devops team. That's more or less explicitly covered: "Prioritize the development of force multipliers".
The idea is that adding a new person to a team increases the work capacity of that team linearly to do the work that the team has been doing, in the same way they've been doing it. Finding new, more efficient ways to do the work is how you increase productivity, not work capacity. Hiring an Eli Whitney means that your work capacity is now devoted to doing something else (operating a gin rather than separating by hand), something that has higher productivity per time unit.
This doesn't remove the contradiction, it only moves the contradiction into the article.
It's not as though the article makes a soft claim which should be interpreted generously, it makes a very strong claim - "at most linearly" - despite the fact that there are all sorts of exceptions. That is certainly not "at most linearly", perhaps "at base case linearly".
The author seems to conflate work capacity with work hours, which is what I'm trying to disprove. Work capacity is IMHO the maximum that a company can accomplish given the resources, which in a multi-skill team can be super-linear if a skill is lacking and it's supplied. The work can be the same but if made by a specialist it's more efficient in a non-linear way. I am sure someone was doing devops before but if an expert does it then it'll improve everyone's life by leaps and bounds.
there isn’t a devops “team”. devops isn’t a job role, it’s a way of life.
This might all be pedantic, but in this vein: one draft horse can pull 8,000 pounds, while two horses together can pull 24,000 pounds , even more if the horses have worked together.
Horses do not become exponentially more powerful as you add additional horses. Similarly, if you had me make something by myself I could do some work, but if you had me make something partnering with someone with complementary skills (for me a good frontend designer) then the result would be more than 2x what I would do alone. This also does not scale... you can't add a chef and an ops person and a ballet dancer and a plumber to the team and keep getting better results.
But maybe you add a whole ops team, and now I can do work and almost forget about that task, and the work is more effective and our little team stays little and the big ops team is professional and reliable so that once something is up it is up. Can whole teams complement each other?
So you get all matrixed  and that's good and bad, but you can only have so many of these groups. Maybe you go long and blow everything up into microservices , but it can't go forever.
Can whole business units become complementary? Or are there other ways to break it up? I haven't read the article so I'm not sure if it answers this ;) But from skimming it, it seems Malthusian in its perspective, hell-is-other-people, that scaling up is unfortunate instead of a process of becoming something new.
I just wanted to say thank you for letting me know that a horse can pull that much, and even more for letting me know that two horses can pull 12 tons. Just, wow.
Best TIL I've had in a while :D
I couldn't find a citation for that fact though, the one about horses. I found a lot of people referencing the article you mentioned but no original source.
Yeah, getting clear information is hard.
This bulletin on horses and horsepower from 1926 is pretty great, though it also doesn't answer the question very clearly: https://lib.dr.iastate.edu/cgi/viewcontent.cgi?article=1466&...
There's a bunch of load pull tests there, but they are only done on teams, never on a single horse.
Quoting the article:
> The ceaseless pursuit of force multipliers is the only possible route to superlinear productivity improvements as an organization grows.
The cotton gin qualifies as a force multiplier.
It's weird that the next sentence after your quote says you can't do superlinear.
I believe a more appropriate way to think about this sort of thing is asymptotically.
Sure, your 10th worker might revolutionize the way your company works. Your 9 workers work on a very specific thing and suddenly your new hire makes it more efficient or does away with a bottleneck.
Maybe at 100 people you have a specific team of 10 people that's a bottleneck and the right person comes along and solves their issues.
At 1000 people, your teams are already working on things different enough that a person revolutionizing the way one of them works is unlikely to affect the others, as your work has inevitably diversified. Again, because you have 100 teams, each team has a smaller impact percent-wise on the total output.
At 10000 people it's easy to imagine all these inefficiencies have already been mostly solved (you had after all hired so many competent people who have tried their best to do so).
Not only that, but there's evidence that the denser a city is (the more members it adds), the higher its productivity per capita . I imagine something similar could apply to some orgs. It's why mergers happen, isn't it?
Or imagine two startups, both too small to really invest in tooling. If you combine them, maybe that changes. People can specialize more. The return on building tool X scales with the number of users, so once you have enough people, you can start to make things easier for everyone.
> Keep the work parallel, the groups small, and the resources local.
That how the military gets things done, by having a trained officer corps which can think on a local level and the higher up give general guidelines for how to proceed in different situations.
I don't mean this to sound skeptical, just curious - would you say that the military gets things done better than your average tech company?
I could fill pages with complaints and criticisms of the way companies do business, but sometimes I wonder if it's a matter of being the worst way except for all the others
> would you say that the military gets things done better than your average tech company?
No, because a tech company is all about avoiding work. Tech companies do as little as possible to reach the benchmark and no more.
The military is about "Getting unit X of work done. Period." Whether that work is stupid or irrelevant is not up for debate once it's assigned. The work must get done; it will get done; there is a minimum standard; the cost is now irrelevant."
The military spends a fucking ton to get the results they get, I'd bet they put a million bucks into my training as a junior officer back then (it was a while ago). That's mostly why they get results. It wouldn't work for most businesses.
In wartime it does, usually.
I guess it depends what you mean by that exactly. Military "group" sizes range from squads (~30 people) to corps (~50,000 people). According to the article, the dyad sizes here would be huge; but my hypothesis was that point-to-point communication is often not necessary in military organizations due to the rigid structure.
The military - and many essential services like police, fire, healthcare - differ in that a majority of the work is not "doing" but "preparing". The military does not optimize towards cheaply increasing daily metrics of death and destruction, rather, it aims to use a targeted amount of force to achieve a particular national objective at the exact moment demand for that objective arises. Most units sit idle most of the time, in a runtime system with very low latency and variable throughput. The same analogy is obvious when you think of police(we don't want crimes committed), fire(we don't want lots of fires to fight), healthcare(we don't want tons of patients).
And that leads towards the extensive investments in training, doctrine, equipment, logistics etc. And so the majority of military employees are not "doing work" in the sense that a tech startup is, where every employee is directly engaged in optimizing the system's value chain - that kind of work occurs only in the core command, around the generals and planners and architects and researchers of the next war. The majority of the work is simply maintenance of the system so that the needed response is possible when called for.
And I believe this examined difference in organizational purpose also applies across a lot of businesses. The value chain is always shifting in surprising ways, but generally in the direction of lower maintenance.
Excellent explanation! The difference between these kinds of "work" is a very poignant distinction.
Exactly right, thanks for putting it better than I could.
I mean that each unit in the military has some general orders on how to proceed in their mission, but the commander has a lot of flexibility. That true for the squad the same as for the platoon, they're just different sizes. The work gets done efficiently because there isn't the need to approve each action with the next level up, but often in business they don't trust the low level manager to make those decisions.
The Air Force has the basic doctrine of "Centralized Control, Decentralized Execution". In essence, thousands of aircraft can be controlled by a single commander but only to a certain point. In practice this means that the commander tells his planners what to do and they make a plan for all of the aircraft and how they will coordinate. It is then up to the commanders of those aircraft squadrons to make sure that their aircraft are in the right place at the right time and have the desired effect.
Not necessary, and also actively inhibited via chain-of-command
I feel I myself come from a similar background to the author. We both think heavily from time to time on the theory of work in philosophical, management theory, mechanistic and (computer) scientific frames. We see correlations. However, I always fear I'm falling into some trap of reductive reasoning or worse, finding I'm really an Anti-realist at heart.
The author would do well to look into management theory, initially Harold Leavitt's Diamond in “Applied Organization Change in Industry” now more commonly distilled down to golden triangle of "People, Process and Technology".
I don't know the answer, not can I offer any overarching counter-argument but I feel a deeper understanding of lessons learned in management theory and operating models for the author might help in their next iteration.
One of the basic parts of the author premise is just wrong:
> The work capacity of an organization scales, at most, linearly as new members are added.
Maybe asymptotically, but usually - not at all. Different hires have specialty skills which existing employees/collaborators simply don't, and would have taken extremely long, or forever, to do what the new person can achieve handily. Also, some tasks are simply better geared towards multiple people collaborating; simplest example: assembling large IKEA furniture...
Roughly, if an org is considered as a system, then entropy of it will continue to increase. To keep the order, more and more energy needs to be spent. This energy is in the form of developing internal tooling, more explicit communication which is repeated often, hiring more, establishing processes (which in turn itself adds more tasks on people's plate), etc
I believe there are two sides I'm aware of when it comes to people. Firstly there exists persons that consider work as another extension of life with social merit. Secondly the persons that think of work as work and disregard any social merit because they care more about it outside of work.
In my opinion if we shut out the first classification we're making these persons miserable because they see social merit in work. I also believe work has become overly optimized compared to previous years and where people should still be able to enjoy social merit if they desire it.
> The work capacity of an organization scales, at most, linearly as new members are added.
This is going to sound kind of grim, but a counterexample here is, e.g., a Southern plantation where children of slaves were also born into slavery, yielding a superlinear work capacity. With that said, even though I intuitively agree with the statement, I think other (more contemporary) counterexamples might exist.
> Coherence costs grow quadratically as new members are added.
I see the point here, but I'm not sure how I can reconcile it with the reality of modern-day armed forces: where millions of organizational members function extremely efficiently (possibly due to very rigid organizational structures -- which the article doesn't really touch on). It seems there's a ratio between dyads and "organizational strength" which probably yields "true" coherence.
I don't think your slavery example is really a counterexample... those children are new members. The linear scale is about people to productivity, not 'cost of a new worker' vs productivity.
Also, the military is different; it is not really trying to produce anything new, so isn't the same type of work.
> The linear scale is about people to productivity
According to the article, the scale is about "hiring" to "work capacity" -- the author specifically points out that productivity is another ballgame altogether. (FWIW, I'm not sure what "hiring" means in the context of slavery -- I suppose a Roman-style post-war enslavement could be one equivalent.)
> Also, the military is different; it is not really trying to produce anything new, so isn't the same type of work.
Yeah, this is a good point. The military is not as "creative" as your run-of-the-mill corporation and, as such, probably doesn't need the coherence.
> I'm not sure how I can reconcile it with the reality of modern-day armed forces: where millions of organizational members function extremely efficiently (possibly due to very rigid organizational structures -- which the article doesn't really touch on)
Yes, I think it is due to the organization topology. The quadratic case mentioned ((N^2 - N) / 2) assumes a fully-connected network. Also, it would depend on the weight of each connection. Perhaps the "cost" of having two mathematicians communicate together is higher than the "cost" of having two soldiers communicate together.
> the reality of modern-day armed forces: millions of organizational members function extremely efficiently
How do you define "extremely efficiently" in this case?
I feel like this is what an engineer who doesn't really understand businesses would write. The models he writes about don't even come close to the complex realities of politics, externalities, market factors, accounting, funding, etc... When he started talking about workers like cpu cores, I knew immediately, this is not going to go anywhere.
If you know more, please share some of what you know so the rest of us can learn. It's less helpful just to talk about what someone else doesn't understand.
I will note that I have written the following in jest, due to the title of the post. I have not read your post, before writing the following statements: Arch biet Mach Frei - There was a man once, who believed as such. Although there were a great many more under him, who didn't much like what he had to say. Even though quite many simply went along with it, because it seemed easy, comforting, one might even contend that it was the right thing at the right time, while the others vehemently opposed such thinking and critique.
> I have not read your post
Then don’t comment?