Particularly with the rise of the "EA has a lot of money" meme, I get asked why CEA can't just outsource a lot of its labor. This is my attempt to explain. I think that my view is standard within management circles, but I haven’t found a great source to link to, so I am attempting to write a version in a definition-proposition-proof format that I think might work well for EA readers.
Summary: it’s usually possible to outsource work on things which aren’t core competencies (e.g. payroll). But outsourcing work on core competencies requires substantial managerial overhead, which is the precise thing outsourcing was supposed to relieve. Graphically:
Say that an activity is a core competency if it provides substantial value to the customer and is difficult for competitors to copy. 
- One of Red Bull’s core competencies is marketing. They do not outcompete others because they have a better tasting drink, or a cheaper one, but because it is better marketed.
- One of Walmart’s core competencies is purchasing power. Other companies could build similar stores with similar inventories, but their prices would be higher, because they can’t replicate Walmart’s purchasing power.
Importantly, note that an organization’s core competencies are not necessarily their most visible products. E.g. McDonald’s most core competency is arguably real estate.
Say that an activity is operationally important if it is necessary for the operations of the business. Common examples include: running payroll, processing invoices, and manufacturing the product.
Claim: An activity can be outsourced if and only if it is operationally important but not a core competency.
For example: many companies outsource their payroll processing, as payroll is often operationally important but not a core competency.
By definition, core competencies are hard for other firms to copy. Outsourcing firms are an example of other firms. Therefore it's hard for them to do this work.
I think this proof is true and somewhat useful but it's kind of vacuous. The next section gives some additional evidence that firms do in practice frequently encounter important tasks which cannot be outsourced.
Management costs & team communication costs
Outsourcing doesn’t only cost money, it adds overhead for the team and manager.
In order for an organization to successfully outsource work on its core competencies, those outsourced contractors need to be well integrated with the team: they need to attend all the meetings and retreats, be in the same slack channels, give and receive feedback on their work, etc. This means that successfully outsourcing work requires a lot of the same management capacity which having in-sourced labor requires.
Successful companies tend not to outsource their core competencies
I am not aware of any successful companies which largely outsource labor on their core competencies. All companies outsource secondary priorities (e.g. use outsourced lawyers), and some will occasionally pull in consultants to help with specific initiatives, but I don’t know of any successful company whose labor force on their core competencies primarily consists of contractors.
Indeed, companies often choose to not outsource even their secondary priorities. E.g. I think a large fraction of companies with >100 employees have in-house legal counsel.
Most tech companies (the area I’m most familiar with) go even further than just employing people and use cross functional teams, because even being on a separate team within the same employer is considered too great a distance for good communication and coordination.
Appeal to Authority
I believe that the view I am promoting here is accepted wisdom in the management community. A Google search for “outsourcing decision matrix” shows 4.4 million results (this matrix being a slightly different version of the graph I drew above).
Product development is one of CEA Online Team’s core competencies
If we are able to outsource everything which is not a core competency, this raises the obvious question of what our core competencies are.
In particular: I think there’s a temptation to say something like “CEA’s core competency is in understanding EA’s, not writing code, therefore we can outsource the code writing bits.”
Unfortunately, it’s very hard to separate product understanding from product development. Inspired is a book length treatise on this claim, but I will try to give a short overview here.
Product development is generally not a waterfall process where you first develop requirements, then do a design, etc. but instead an iterative process where you bounce between these different stages unpredictably. One part of your requirements ends up being completely infeasible for engineering reasons so you have to go back to the drawing board, but another part is perfectly fine so it can be released quickly, and a third needs repeated iterations on small aspects of the UI to get right. So you can’t have someone who just implements designs – they need to be giving feedback and generally in deep communication with the “strategist”.
And the strategist can’t just simply give the implementer a specification that the implementer can naïvely follow. Even for the simplest changes there’s a bunch of nuance (what do you do on mobile? What if the user is logged out?) and for complex projects it’s essentially impossible to give a fully detailed specification. So the implementer needs to have a deep understanding of your strategy and product to fill in the gaps in the specification.
Could these issues be addressed even with outsourced labor? They sure can: You can have contractors who don’t need complete specifications because they have attended all your meetings, are in your slack channels and asking regular questions, getting regular feedback about how their implementation is lining up to organizational objectives, providing feedback to the product managers through regularly meeting them and attending user interviews, etc.
But this is all the “management overhead” and EA labor you were hoping to avoid by outsourcing!
Many novice managers, both inside and outside of EA, have an intuition like “I want X and have money, this firm says they will give me X in exchange for money, therefore I can give them money and get X.” In some cases this works well, but in many cases it’s actually quite hard to exchange money for services.
CEA is hiring for a number of positions, including UI, UX, or graphic designers, product managers, engineers, in-house legal counsel, and an office manager.
This work is licensed under a Creative Commons Attribution 4.0 International License.
This is a simplified version of the standard definition
This isn’t quite true: for example, you can’t outsource an activity if there are no external firms to whom you can outsource it. Also I guess technically you could outsource work which is neither a core competency nor operationally important, but you should probably just not be doing that work at all. But this claim captures the most important points
This is such a truism, I sometimes see people using “core competency” to definitionally mean “labor which is not outsourced”. E.g. Red Bull outsources the manufacturing of their drink, and I sometimes hear people assume without further evidence that this means producing a drink is not a core competency of red bull, because if it was then it wouldn’t be outsourced.
Construction in the US is an exception: many construction companies’ labor force primarily consists of independent contractors. However, this is due to labor laws in the US, and in practice construction workers have a relationship to the entity they contract with which is quite similar to how employees relate to their employer.
I saw this title and assumed someone was making a public criticism of CEA.
Then I saw it was written by a present CEA staff member.
And I thought "Wow, creative way to get changes made at your organization." :D
(Someone told me this comment read as hostile to them; FYI I thought it was a funny series of thoughts that I had, no hostility meant at all!)
I didn't think this was hostile at all.
FWIW I didn't interpret it is hostile, though I did change the title to make it more clear that I'm not suggesting CEA change
I also did not interpret it as hostile fwiw (though I'm not Ben/CEA)
Outsourcing a function is usually not binary. For example, Red Bull's brand was actually originally developed by an outside agency https://kastner.agency/work/red-bull-brand and they still use a mix of internal and external marketing teams today. Often internal teams at a company function serve as a bridge between the company and contractors.
With that said, I wonder if the people asking about outsourcing are thinking of it in the literal "employe vs contractor" sense that you covered. When I have heard these debates I believe that people meant "If we have money now, why not hire non-EAs for these positions?".
This is a great counterexample/clarification, thanks!
I might do a future post about how important it is to hire value aligned people; I agree that this is a slightly different question.
On the whole I think you are correct. However, things like restaurant franchising seem to push against this specific point a little. If McDonalds, Coca-Cola etc. had never thought of using franchises, they probably would have thought of running restaurants and making soft drinks as being core competencies.
Yeah, you can definitely get into something like the old joke about two economists who find a $20 bill on the ground: your activities must be core competencies, because if they weren't you would already have outsourced them.
Thanks for this! Most of what you wrote here matches my experience and what I've seen grantees experience. It often feels weird and frustrating (and counter to econ 101 intuitions) to be like "idk, you just can't exchange money for good and services the obvious way, sorry, no, you can't just pay more money to get out of having to manage that person and have them still do their work well" and I appreciate this explanation of why.
This post, written by somebody at Wave, argues that you should outsource even less than this post argues for in many cases. I found it interesting so wanted to link it here.
Wait Dan Luu is at Wave now? What an unexpected crossover! 😍😍😍
This is a really interesting post, and I had not seen it before. Thanks for sharing!
Very interesting, I like this idea!
There are only two things that concern me:
- It seems that the cost of outsourced services is sometimes unduly inflated. Take, for example, legal services. Besides drafting contracts, the lawyer also informs the client about the risks associated with a particular activity and how to mitigate them. Unfortunately, this area is prone to lawyer abuse in the form of artificially generating problems which translates into higher costs. For this reason, I guess small organisations tend to leave this kind of work to administrative or management personnel, and larger organisations tend to hire in-house lawyers.
- Second thing is that generalists and junior staff sometimes tend to switch within an organisation into other positions, so for EA orgs, it could be an opportunity for spotting talent who has a good fit for working in, as you said, core competencies, which probably would not be possible otherwise.
I get that outsourcing doesn't work for core competencies but why does outsourcing not work for operationally unimportant activities? Basically I'm confused by the bottom-left quadrant.
> Also I guess technically you could outsource work which is neither a core competency nor operationally important, but you should probably just not be doing that work at all.
Yeah, this is what footnote 2 was trying to explain.
I was considering marking that quadrant some third color to indicate that you should just not do those things, but thought that would overcomplicate the message of "only outsource the bottom right quadrant". Peter's comment that this is confusing is helpful feedback that maybe that was the wrong call – thanks!
What's an example of something that is a core competency yet operationally unimportant (the top-left grid)? I'm starting to think the entire operational importance axis isn't needed.
An example is something like "having a CEO who is considered a prestigious thought leader in their field." The day-to-day operations of the business aren't really impacted by this, but it's also not something you can really outsource.
(That being said, maybe it would have been a lot simpler and almost as correct to just leave this axis off, like you suggest.)
Ah ok that makes sense
Ben_West thanks for this text. I have some examples of outsourcing work in our organisations. And even when sometimes they are for free, for me as a Project Manager it is an additional cost, mostly time which is precious in NGO. Communication and trying to explain what is matter for us is sometimes very complicated, especially if someone wants to contact by mail. From last cooperation we did many improvements, like for example adding people to our slack and explaining the way we work in IT (even about code to not have broken windows after cooperation like that). But I still think that outsourcing should be rethought twice to not have more work than less. So thank you for your view for this topic.
I have trouble understanding this notion in practice, especially (as you've noted) from the outside it is often hard to tell what core competencies of companies actually are. For example, what is the core competency of Uber/Lyft if not the network effects of having lots of drivers and customers in one place? Other plausible contenders like "brand" or "UI" or "scalable software" don't seem quite right to me.
I would have said "network effect" (though I don't know much about either of them). This seems analogous to the claim that one of Walmart's core competencies is purchasing power.
Hmm doesn't network effect entail that people in the network (here, drivers and arguably customers) are your core competency? Or am I missing something?
Yeah I think that's right. "Having drivers in every major city in the world" is a core competency.
I also quite appreciate the organization/management posts, as someone else commented above.
I'll also add that cultivating and retaining institutional knowledge with a regular team is already quite a task in itself, and sustaining sufficient ontological and practical coherence with contractors can lead to a lot of inheritance type issues. This kind of overhead cost can often be overlooked or trivialized, in my experience.
However, I can imagine a scenario where there could be the deliberate training of certain contractors to fulfill a particular ecosystem's common needs. For example, most EA organizations would benefit from legal work (amongst other things like accounting) supported by specialists that are at least sympathetic to the philosophy of the community. In which case the overhead cost of one organization taking on outsourcing might reduce the cost of another organization in the same community taking on the same contractor. This could make sense for the contractor too, because they can develop their own niche and get more clients. To coordinate this though would require exceptional sharing of learning/intel across the community, which warrants more writing than this comment can serve.
Accounting for "overhead", or higher order costs/gains that a lower order action may create, is really interesting to me. I hope there'll be more discussions in the future.
Upvoted! I find your posts on management very beneficial for our organisation. Thank you!
Thanks! I'm glad to hear they are helpful.
Perhaps a dumb question, but do companies tend to have an accurate picture of what their core competencies are?
I'm wondering if trying to outsource X and seeing that go negatively/positively might help update towards X being/not being a core competency. Or is there too much noise to update significantly in practice?
Just following the example from the post: did CEA ever try to outsource some product-related tasks? Are there any circumstances in which it would it make sense for you to do so even if it's a core competency?
This is an interesting question! The world is certainly full of stories of outsourcing gone awry, which is some evidence that companies don't always have an accurate picture. But I don't know rigorous statistics on this.
Yep, for example we have a contract software engineer who works on effectivealtruism.org. I think this works because we can have a relatively "complete contract": we can hand them a design specification and say "turn this into code please". EA.org is a relatively straightforward static website, which makes it easier for the design specification to be complete.
The things that are harder to outsource are when the design is harder to define. As the design becomes more complex, a complete specification often ends up just becoming code ("if the user is on mobile and logged in and has at least 1000 karma do this, if they are on mobile and logged in and have 50-100 karma do that,… Oh actually we have to consider this race condition which you basically need to be a programmer to have thought of...").
I don't think this is a dumb question at all.
Thanks for writing this Ben.
I find these kind of post with structured line of your reasoning very impactful and I would recommend people here to share it with other people in management roles that may skip this post.
I encountered a lot of examples of organizations doing optimally for themselves when not internalizing this concept. This is especially tricky when outsourcing can give you benefit in the short-term, but is negative in the long-term. I often found this to be case in groups that outsource some legal counseling, people operations, marketing or fundraising.
Two more points.
1. On trading money for work
There is similar point made by Holden Karnofsky in 2013 for GiveWell - We can’t (simply) buy capacity.
It's more about just spending money to hire people, but not only and lists things that can be outsourced:
People who found this post useful may found the link helpful.
2. On core competencies
I think this temptation to declare something not one's core competency comes from some kind of bias that groups have. It makes people not want to do things that are not natural for organization's ethos. It reminds me a bit of points made by Paul Graham in Do things that don't scale that engineers want to code not to run their sales which he thinks is a wrong approach:
So IT companies want to code, advocacy groups want to campaign and effective altruism would probably want to reason how to do the most good. In my group we he had this when we started as investigative group, but fortunately realized quickly it wasn't working as intended.
I generally think it's dangerous to go path of outsourcing / buying capacity if an organization wants to be successful, so I think it would be good to have a heuristic that you should not outsource anything by default and then understand properly the cases where it's fine to outsource. Yet, I think the reverse is the default when I talk to some groups (although usually not from the EA space).
Yes! This. Thank you for writing.
I often get asked why LessWrong doesn't hire contractors in the meantime while we're hiring, and this is the answer. In particular the fact that getting contractors to do good work would require all of the onboarding that getting a team member to do good work would require.
Just curious, has CEA ever hired/talked to a management consultant about all of this? I think it would be more persuasive to some people if you said “we hired/contracted a management consultant to discuss the situation and they largely concurred, saying that we can’t contract out our core competencies even with a funding overhang” rather than the “proof” (which TBH I thought was a bit silly) and the appeal to common practices in business (although this would be a good supplemental explanation).
I think that appeals to successful business/management models make some sense at least to give some contextual perspective. However, CEA is not a “business” with “competitors” in the sense that 1) CEA’s primary mission involves producing large amounts of positive externalities, and 2) CEA is funded by grant makers and is thus less dependent on “customers’” (grant/service recipients’) funding.
Additionally, the situation seems to be a major funding overhang relative to talent, and one of CEA’s primary purposes is to improve the talent pipeline (e.g., by supporting community building).
Ultimately, in a competitive market, shoveling money at a “core competency” to attempt to scale up may indeed be unsustainable, especially if your primary value is money and if you have a competitor that will continue to profit and may use those funds to outcompete you later. A firm will not choose to indefinitely produce 200 widgets at a profit of $0 per widget vs. indefinitely produce 100 widgets at a profit of $6 per widget if it cares about profits instead of net widget production/sales. But in this context it seems that even if outsourcing some core competencies produces inefficiencies on a per-dollar basis (e.g., it requires hiring more managerial support such as secretaries even if their work is not as “efficient” as the average secretary), it might still make sense.
Of course, it might be the case that no (tolerable) amount of throwing money at the problem produces efficiency gains, I’m just trying to say that the appeal to standard market practices might not be justified if the reasoning is “they need to do this to compete/profit.”
It seemed to me like the key variable you’re trying to highlight is management costs or something similar (while “core competency” is just an indirect/less-explicit way of referring to that), but perhaps I’m misunderstanding? Is there a reason you can’t just make the Y-axis “Demand on management” (or some other cost/resource)?
After all, it seems quite plausible that a company’s “core competency” could be something like “ability to effectively outsource work and manage contractors,” in which case the diagram doesn’t seem to make as much sense: much of the work is through outsourcing.
Hmmm, no, I think the ability to outsource well is not itself easily outsourceable. E.g. if you have some method of identifying whether an outsourced factory will produce high-quality products, I guess you could train an outsourced team to do that identification, but that doesn't seem remarkably easier than hiring staff and training them on your identification methods.
I think there may be some confusion over the semantics of “core competency”—I wasn’t trying to say you could outsource the outsourcing, I was just saying “a company’s biggest strength can be that it is effective at outsourcing”—but I feel like that confusion further reinforces my main point, in the first paragraph: it seems to me like “management complexity/demand” would be a better Y-axis label than “core competency-ness”?
I think you are saying something like: "outsourcing is a managerial task, therefore bottlenecks on outsourcing are by definition bottlenecked on management."
I think this is true, but I don't think it's the most helpful way of phrasing it. E.g. many biology labs can't outsource their research (or even have it be replicated by labs which are almost identical) because their work relies on a bunch of tiny things like "you should incubate the cells at 30°C except if you notice some of them starting to turn a little yellowish increase the heat to 32°C but then also you maybe need to add this nutrient bath…"
You could argue that documenting these procedures is a managerial task, and therefore the outsourcing is bottlenecked on management – again, I think this is true, but it seems more insightful to describe these biological procedures as a core competency of the lab. (To me, at least, YMMV.)
I suppose "management complexity/demand" might indeed be a bit too narrow, but either way it just feels like you're basically trying to define "core competency-ness" as "difficulty of outsourcing this task [whether for management demand or other reasons]," in which case I think it would make more sense to just replace "core competency-ness" with "difficulty of outsourcing this task."
My worry is that trying to define "core competency-ness" that way feels a bit unintuitive, and could end up leading to accidental equivocation/motte-and-baileys if someone who isn't familiar with management theory/jargon interprets "core competency" as important functions X, Y, and Z, but you only mean it to refer to X and Y, reasoning that "Z is some really core part of our operation that we are competent at, but it can be outsourced, therefore it's not a core competency."
TLDR: Outsource tasks that do not require critical thinking or lead to growth within EA to EA group-vetted contractors in emerging economies and encourage others to do so too. Require some understanding of EA ideas (e. g. intro program) for non-leadership tasks with opportunities for growth in EA-related work independence. Only highly engaged community members should develop strategies and lead thoughts and narratives.
1) Would you consider outsourcing to specialists with strong discipline-specific skills who are vetted by EA groups in emerging economies (Pakistan, India, ...)? This can reduce costs, increase efficiencies, and even enable people to run a part-time project alongside a full-time job, if the wage differential is 5x.
The vetting may be key to prevent reputational loss risk related to the nature of EA work, for example if a graphic designer wants their suffering neighbors to be assisted first, regardless of the cost-effectiveness, or knows that a family celebration has to have meat (so reducing meat consumption is against families) (these are illustrative claims).
2) Why would you not employ EA community members (rather than non-EA freelancers), if there are those interested in such work? There may be persons in high purchasing power nations with low nominal wage alternatives looking for employment. Also, depending on the definition of an EA community member, this can be relatively facile (or affordable) to achieve (for example, if you pay someone (upon application to check if they are there for impact) to participate for 20 hours in the Intro to EA fellowship, also to increase their productivity).
Would you seek to employ persons looking for growth within EA, including within their specialization, or rather those who would have limited ability or willingness to further upskill? If this depends on the role, how would you ensure that persons who seek to advance EA are supported in such, those seeking to increase their employment compensation and security can do that while prioritizing EA work but being independent on it, and those seeking or able to just keep an entry-level position (e. g. data entry, online search) can do that?
3) How will you assess position core-competencyness requirements? I judge that this relates to the person's internalization of EA objectives (e. g. actually altruistic and seeking effectiveness) - but, perhaps the central meta-question is: what if someone can exhibit this internalization at work but all they care about is to have a good job to pursue their goals (such as happy family, buying a nice car, ...), what is their position within or around the community?
Relatedly, is there a list of some tasks that you would outsource to 'external' labeled contractors, less engaged community members, and highly engaged members? I would estimate:
External: field officer work of EA-related charities that does not require problem solving or critical thinking (such as distributing bednets, enumerating whether bednet is hanging, ...), data entry and management or personal assistance for EA non-decisionmakers and non-influencers (making spreadsheet look better, re-posting content, ...)
Less engaged: field officer work that would benefit from critical thinking on higher impact (e. g. enumerating whether bednet is hanging and seeing why people do not use them, what their priorities are and how they can be better assisted with equal resources), assistance to EA decisionmakers or influencers (e. g. making spreadsheets look better while learning about further possibilities to do good within the organization or its cause area, drafting content based on discussions, ...)
Highly engaged: strategic planning, thought, and narrative leadership (e. g. what should we do about regions that experience malaria and other issues that is most cost-effective but is also well-received locally, what kind of AI safety image should be designed to limit reputational loss risk of companies that profit from AI advancement for a poster, ...)
In general I do have a preference for hiring people involved in EA. Among other things, these people are easier to vet (as you remark).
When I hire contractors who are not EA's (including those in developing countries), it's usually because we need a specific skill set and can't easily find EA's with that skill set.
The breakdown you list at the bottom of your comment seems approximately similar to how I think about how easy work is to outsource.
Ok, that makes sense. Then, just encourage people to be flexibly hirable?
This can be minor but I was also making this comment for an event feedback: is there some staff that should be considered external to the community, such as 'dead-end' professions who are not interested in self-development, such as cloakroom staff? As in, you would not make a community member to specialize in carrying coats (or some other brainless, especially manual, work, unless it involves skilled emotional work)? (This is not to say that e. g. cloakroom person cannot be active in EA, if they are e. g. working part-time and reading about EA-related topics.) But, this is an argument for a preference for hiring non-EA involved people (or working with companies that contract staff). I am actually realizing that if it is possible, even persons who carry coats may be able to entertain one on 'hmm I know this brand, they are sourcing environmentally and socially sustainable materials, but it looks funny' etc.
Regarding software projects specifically: Why not outsource fixing bugs rather than product development?
Good question; it seems pretty similar to feature development.
I think the reason the second one is hard to outsource is general problem.
(E.g. how you would verify that the outsourced firm investigated it correctly? Verifying that they actually fixed this hard to debug issue is often almost as complicated as just fixing it yourself. So you basically have to trust that the firm did it correctly, and getting an engineer who you trust to fix these problems is basically what it means to hire and onboard an engineer.)
Great post — definitely aligns with my world model and experience!
One small thing to add: When adding someone to an org, it's really expensive to transfer context to them. In CEA's case, I imagine this cost would be even higher if the person were unfamiliar with EA .
Btw, one possible exception to this rule might be a really good product/ui designer, because they can be enough of an empath to quickly pick up the critical considerations in conversation. (But people at this level are really rare.)