T-Shaped Organizations

post by Vaniver · 2019-12-16T23:48:13.101Z · score: 51 (14 votes) · LW · GW · 7 comments

Back in 2012, the Valve New Employee Handbook made the rounds on the internet, and claimed that Valve was looking for T-shaped individuals, who were broad generalists (the horizontal top of the T) with a depth of expertise in a particular topic (the vertical base of the T). I was reminded by this by thinking about Your Price for Joining [LW · GW]. Eliezer argues that people (and skeptics / individualists in particular) likely overestimate the amount of ‘negotiating pull’ they should have in interacting with groups, and set standards that are too high, especially for incidental factors (like use of a subpar font or money storage method).

According to me, a T-shaped organization should be ‘broadly generalist’, in that it needs to do lots of different things in order to exist at all (like keep money in a bank, and comply with regulations, and have an office, and a website, and so on), and ‘narrowly expert,’ in that it performs exceptionally well on the functions related to succeeding at its primary mission. But with limited attention and resources, whenever the organization is given a choice between doing a lot better on a secondary factor or a little better on a primary factor, it should choose the primary factor. The organization should be half-assing everything; ideally in the way where the derivative of the goal with respect to each input is the same, and failing that hitting each quality threshold with minimal effort, and failing that being a trier at the primary goals and a slacker at all the other goals. The difference between a good bank and an okay bank probably has almost no effect on organizational success, whereas the difference between having a bank at all and having no bank has a huge effect on organizational success.

This means that you should think very differently about organizational performance on primary tasks and non-primary tasks. You almost want to cultivate a sort of Gell-Mann Amnesia, where you see a restaurant’s menu use a font no designer would be caught dead using and forget about it as actually not very informative about the quality of the food. You don’t want to forget it all the way, because underlying generators do cause things to be correlated; just as real budgeting constraints mean that money spent on the menu can’t be spent on the food, real variation in wealth and profitability means that some restaurants can spend more on both than other restaurants can. The same reasoning goes through for human capital in organizations.

Your Price for Joining’s examples implicitly assume that the criticisms levied are on secondary factors, but I think the post’s proposal also goes through for primary factors. If you think the restaurant doesn’t have good food, and you could make the food tastier, either you join as a chef or start a competing restaurant or withhold your support conditional on the food improving to some other standard. But also someone else's ox likely is gored by you joining the restaurant and being a chef, and perhaps what you think is tasty is not what their customers think is tasty. 

7 comments

Comments sorted by top scores.

comment by Vaniver · 2019-12-16T23:55:47.957Z · score: 15 (6 votes) · LW(p) · GW(p)

A related question that I find interesting is, what if the observer and the organization disagree on what the primary factors are?

Continuing the restaurant example, suppose I think that food safety should be one of the restaurant’s primary concerns, and the restaurant thinks it is a secondary concern, where they should slack as much as possible while not being closed down by the health inspector. In a regime where health inspections are sufficient, this probably doesn't matter much, but often information asymmetries will mean that the restaurant has much more ability to affect health than the inspector can check.

This sort of thing--which values you're a trier on, and which you're a slacker on--also seems like it's the sort of thing that gores oxen by default. The restaurant where they have to wash the knives in between uses is slower at getting food out the door than the restaurant that reuses them. [I'm not a professional chef, I don't know what examples are actually relevant on the margin here.]

[edit] It also seems like this is the sort of thing that marketing pretty strongly encourages misrepresentation of. "All children are above-average," in that the restaurant wants to present itself as serving healthy, cheap, tasty food, while also paying its employees well and having good returns for its investors. But several of those variables are in direct tension with each other, and there's not great language for speaking publicly about the tradeoffs you're making. 

comment by Dr_Manhattan · 2019-12-18T13:39:23.196Z · score: 4 (2 votes) · LW(p) · GW(p)
[edit] It also seems like this is the sort of thing that marketing pretty strongly encourages misrepresentation of. "All children are above-average," in that the restaurant wants to present itself as serving healthy, cheap, tasty food, while also paying its employees well and having good returns for its investors. But several of those variables are in direct tension with each other, and there's not great language for speaking publicly about the tradeoffs you're making. 

Couple of reasons spring to mind:

  • Marketing leverages the Halo Effect
  • Not emphasizing being above average on a particular dimension is a chink in the armor against competitors (who can establish beach head by claiming to be superior in that dimension)

comment by mr-hire · 2019-12-19T19:50:09.691Z · score: 4 (2 votes) · LW(p) · GW(p)

One of the things I'm curious about is why you think there should be T-shaped organizations. None if the research I've done into effective organizations points at anything like this, at least not after the first few years when an organization is done finding it's feet.

comment by Vaniver · 2019-12-20T19:00:13.891Z · score: 5 (2 votes) · LW(p) · GW(p)

I think it's a common view that orgs should have a particular focus, vision, or specialization; I think "core competency" is a more standard term for the 'narrowly expert' thing.

I think once an org is large enough that its departments are mini-orgs of their own, the combination of narrow expertise leads to a more well-rounded org, but I think it'll still be the case that, in the competition between departments for resources / executive attention, some departments will be seen as being much more important for the org's overall success.

comment by mr-hire · 2019-12-21T13:42:54.624Z · score: 7 (1 votes) · LW(p) · GW(p)

Its not clear to me that this new language points at a useful distinction(at least not from what you've said so far). Valves t-shaped individuals are actually pointing at the opposite of what you're pointing at - they don't want individuals too specialized because of their unique corporate structure (ie, not having a corporate structure), so they need everyone to have the broad base.

And the long tail on specialization works because they can hire to fill out relevant other competencies. Whereas, it's not always the case that it makes sense for a company with a core competency to outsource functions to other companies, there are Coasian dynamics at play to make that cost more than one would expect.

I would much more, through a process of pushing through constraints, expect a healthy company to look something like vshaped, with having a smooth gradient of competencies based on how central they are to throughput.

comment by G Gordon Worley III (gworley) · 2019-12-26T20:20:52.434Z · score: 5 (2 votes) · LW(p) · GW(p)

Indeed. A modern version of this is the "lean organization", which is a particular methodology for doing the sort of thing you are pointing at here. Alas business terminology is rarely generalized away from implementation methods, so I don't know of a general term to describe what you're pointing at that isn't tied up in implementation details, i.e is purely descriptive of all orgs having a shared property regardless of how it is achieved.