Cultural over/under-fitting and transfer learning. Or why the “Netflix Culture” won’t work in your company.
A couple of weeks back, I gave a talk in the SFELC (San Francisco Engineering Leadership Conference). As I was preparing the slides for the talk I reflected on something pretty interesting: I have been managing technical teams for over 25 years. I have also been giving public talks for about the same time. However, this was the first time I was giving a talk about managing teams. I was also surprised that, despite a packed house and many people congratulating me after the talk, my tweets about the talk and the conference got very little engagement. I guess technical management doesn’t get as much traction as AI/Machine Learning. I get it. In any case, one of the most important points I made during the talk directly related to a some machine learning concepts like transfer learning, or under/over-fitting. So, maybe by sprinkling some of the “AI hype” onto management cultureI will be able to get people’s attention on some very important issues for not only any company, but also every founder, manager, and potentially any employee .
Overfitting and underfitting
In very simple terms, underfitting happens when we try to explain a complex real-world phenomenon with a model that is too simple. As an example, this often happens when we “rush” to simplistic conclusions to explain something after just observing one of the causes without realizing that there are many more. On the other hand, overfitting happens when we do the opposite: we use too complex a model to explain a phenomenon. This complex model might explain the current observation, but does not generalize well to similar situations. While over/under-fitting is a very important issue in machine learning, it is also present in every day life in many ways and forms. We humans underfit and overfit all the time.
Examples of over/under-fitting in company culture
So now that we have the “theory” under our belt, let’s see some examples of underfitting and overfitting in company culture. I will refer to some popular management books for that purpose.
First, let’s start with underfitting. The way to uncover underfitting is to look for the following pattern: someone claims (or implies) that the success of a given organization is explained by one (or a very limited set of) aspect. A good example of this is the book “Measure What Matters” by John Doerr. As much as I like OKRs, if you read that book you might come to the conclusion that all the companies listed in the case studies (e.g. Google) have been successful because of OKRs. While OKRs might indeed have played a minor role, it should be clear that there are many other variables that have had an impact on why Google and others in the book have been successful. You could even argue that OKRs may have not caused them to be successful, but rather the other way around. Other successful cultural and organizational traits caused the companies to adopt OKRs, so these are reflecting success rather than causing it. In other words, and for the geeks reading this post, while there might be some correlation, there is no causation.
There are as many or probably even more examples of overfitting. The pattern we are looking for here is someone claiming that a complex model that was successful in one (or a very limited set of) company, will generalize to others. Of course, you can already see how this pattern repeats in almost every other management book: famous business person A explains what they did at successful company X and will claim you should do the same. A prime example of this is Ray Dalio trying to convince you that the “extreme” culture he implemented at Bridgewater is going to work for you. There are actually many such books or posts all around. Laszlo Bock will try to convince you that the Google culture works by just giving you one example of success: Google itself. Patty McCord does the same thing in her book Powerful when trying to get her point across that the Netflix Culture should generalize well and apply to almost any tech company.
Let me dive into on of these examples a bit more, because I think they are very dangerous, both for companies trying to define who they are or for older companies trying to re-invent themselves. I will use the example of the Netflix Culture because it’s, out of all these, the one I know best.
I should start by saying that I loved the Netflix Culture. I really thrived in that environment, and when I was there I thought it was almost perfect. Even now I think it is maybe 90% good. Still, I can also say that this culture will not work anywhere else, even if you apply everything in Patty’s book step by step. For example, are you willing to give up on hiring junior candidates and/or interns just because they are likely to do poorly in such culture? Are you willing to bias your company towards some personality types that will thrive in that culture while penalizing others? Are you ok with giving up on having remote offices? And, so on. The Netflix Culture only worked at Netflix because of hundreds of small details that end up connecting directly to the business model, mission, and leadership team. I will say it once again: it will not work anywhere else.
A particularly annoying manifestation of overfitting is when people in a company use the following rationale: “Everyone who works at my company is happy and successful, that proves that my culture is right”. In Principles, Ray Dalio says that new employees usually have a hard time at Bridgewater during the first few months, but then end up loving it there. Patty McCord says in Powerful that “all engineers” despise process, and that is why all are happy at a culture like Netflix’. That is, of course, BS, and it is just an example of survival bias. Everyone who “survived” at your company did so because they adapted to the culture and were in some way predetermined to do so. This proves nothing except that you managed to rule out everyone else.
In machine learning, we talk about transfer learning when a model that has been originally trained for one problem can also be applied to a different problem, maybe with some “minor tweaks”. Similarly, while in the previous section I made it sound like it is hard to apply learnings from one company to another, there are certainly many lessons that have repeated enough to be transferable. As a matter of fact, even most of what is written in the books I criticized above can be transferred if you are careful enough to adapt them to your situation and context.
After having been at very different companies, all the way from large 100 year old multinationals to my own startup with only 3 people on day one, I feel like I am in a relatively good position to highlight things that are easily transferrable and you might want to apply anywhere, and things that are much harder to adapt. Of course, I might be over/under-fitting myself, so take everything with a grain of salt.
So, what are the traits that over and over I have seen contribute to a culture’s success no matter the context, and the company?
- People above all (aka You are exactly as good as your team)
This might be the most important invariant: It doesn’t matter how good you are as a manager, you will fail with a bad team, and even the worst manager will seem good when leading an awesome team. Of course, it is unlikely that a great manager ends up leading a bad team or for a bad manager to assemble a great team.
So, how do you build a great team that will make you (and everyone else) successful? There is one, and only one, answer: hiring. As a manager/leader, recruiting should be your number one priority. This might seem an overstatement, but this is indeed an important thing that separates successful teams from unsuccessful ones, and successful companies from unsuccessful ones. And, if you still think this is extreme, you might want to know that at Netflix managers who were struggling to hire someone important to their team were encouraged to stop doing everything else and focus on that hire 100% until they closed someone. And, when I say everything, I mean everything. They would stop even attending “important” meetings.
Maybe one of the saddest moments of my hiring manager career was when a manager that I was interviewing told me that if they could, they would fire 40% of their team. That didn’t say much of the manager, and even less of the company as a whole. Interestingly, the company does not exist anymore.
There are of course other important things to keep your team happy and performing for a long time after they have been hired. It is beyond the scope of this post to go into all the different management techniques that can help. However, if I had to highlight just one that seems to valid no matter the context and the situation, that would be the importance of regular 1:1s with the team.
2. Culture eats strategy for breakfast
It is extremely important for a company’s success to have a strong/clear culture. This might be even more important than having a strong/clear strategy, because in fact a strong culture will likely lead to a strong strategy, but not the other way around. Think about any of the successful companies right now. Google, Apple, Netflix, Salesforce, Amazon… they all have a strong clear culture that you could describe. Note that they don’t all have the same culture, and indeed some of them are very different, but they all have strong and clear traits that define who they are.
My take on this is that, as Google discovered, one of the most important keys to success in any organization is psychological safety. While there is much more to it, an important aspect of such safety is having clarity and understanding of how people will respond and act in the organization in response to anything you do or say. This is only possible in an environment where you have a strong culture and norms that dictate what everyone should expect of each other.
A word of caution here. Strong culture as I have used above might be misunderstood as one that strongly limits freedom or strongly selects individuals who are “fit” for it. Besides the many well-known benefits of diversity in any organization, you should build a culture to scale and therefore to include as many (different) individuals as possible. Of course, there will always be some “kinds” of people that might be more or less attracted to who you are. That will depend not only on your culture, but also your mission, product, and stage. That being said, you should do your best to minimize “culture cliques” unless you want to face the issues that many early successful companies face as soon as they start scaling up.
As an example of this, when we started our company Curai, I talked about setting up our cultural values asap. However, my cofounder pointed me to a post where they advised not to write down your company culture until you had at least 20 employees. After thinking about we agreed they were right. If you set up the culture with only 2 people in the room, you run the risk of that culture mirroring your own values. Having input from at least 20 peoples will make your culture much stronger.
3. Clarity & ownership are important
As mentioned in the previous paragraph, one of the most important keys to success is clarity. No matter how big or small a company is, it is important to have clarity in mission, strategy, and ownership. Of course, such clarity is not easy, but it is on leaders and managers to push for as much clarity as possible. It is really hard for anyone to work in a confusing environment where goals are not clear and the direction is changing constantly. You might think that this is precisely the definition of early startup, and you are probably right. However, even in such environment you can seek and provide clarity.
A couple of techniques that I like are the concept of DRIship, and OKRs.
DRI, or Directly Responsible Individual, is a concept used to explicitly describe that any activity in an organization has to have a clear owner. Confusion over ownership is one of the worst things for a team. Have you ever been involved in a situation where someone thought someone else was in charge of making the decision they should have made? Of course. We all have. I won’t have time in this post to convince you why ownership matters so much, but I would recommend the book Extreme Ownership, which makes a really compelling case of why that matters in all kinds of situations.
OKRs (Objectives and Key Results) are also a good way to provide clarity on goals and ways to measure progress in an organization. While I did refer to this book as an example of underfitting in the first section, I do think that John Doerr’s Measure What Matters is well worth a read.
4. Process matters
Make no mistake: every company has some flavor of process. Of course some companies might have way too much process while others like Netflix strive to only define the MVP (minimum viable process). In any case, process matters. Process brings clarity on what to expect and how things are run. As you will read in the Netflix slide deck, there is such thing as “good process”.
Even when deciding not to define a process you are implicitly defining one and you are inviting others (teams, individuals) to implement their own. So, you should be intentional about what process needs to be defined to bring clarity and what process can be left out for people to figure out on their own. That decision should, of course, be directly connected to the company culture.
What needs adaptation
There are many other things that cannot be easily transferred from one company to another. Here is my top three selection.
This might be a bit counter-intuitive. I mentioned above that having a strong culture is a “must-have” for any successful company. What I am saying here is that the details of that particular culture will not travel well from one company to another. In other words, as I mentioned in the title of this post, the Netflix Culture will not work in your company. Another way to see this is by looking at the many successful companies I listed before (Google, Apple, Netflix, Salesforce…) and realizing that their cultures are not at all similar. Some times they are very, very different.
So, why is that? I have already mentioned a few things in this post that partly explain this. A company culture needs to be aligned to the company mission, stage, business model, focus area, strategy, and people, among many others. For example, as much as I loved Netflix Culture, this culture is based on the premise that anyone you hire has to be “experienced”. I knew this was not going to be very successful when going into Quora and realizing that there were many bright recent grads, and my boss and CEO was twenty-something.
This anecdote is true for any other culture. You want to adapt Bridgewater’s culture outside of finance? Good luck with that. Please let me know if anyone is even mildly successful at that attempt. Do you want to copy Facebook’s culture in your company? Good luck if your company is anything but a social media startup. As a matter of fact, Facebook is even struggling with scaling their culture now that they are not a startup anymore!
Similarly to culture, while process always matters, the details of the process are hard to directly transfer from one company to another. Yes, you will be able to use some concrete techniques like OKRs, but beyond that, most things will need to be adapted to your particular context. As a matter of fact, process is a representation of your culture plus your strategy, so they need to go hand in hand.
Finally, and perhaps more importantly, people are very different from one company to another. While, as I mentioned above, you should strive to make your culture as inclusive as possible, your company will attract different people, and for different reasons, than any other company.
This is key, since everything else, including culture and process, will need to align to the people you have in the company.
Is that all?
Of course not, there are many other things that matter in a company culture. For example, I am a fan of the three C’s: Challenge, Care, and Communicate. This is an extension from the two-dimensional challenge and care framework presented in Radical Candor. I do think those three things in combination are really key. However, I am also aware that there are many examples of successful companies that do not observe one or more of them. For example, Apple does not care too much about the “communicate”, while Netflix does not care too much about the “caring” aspect either.
Some reading recommendations
I know in this post I have been generally critical of management book. However, there are many great books on these topics out there. Even the ones I criticized above I would recommend if you take with a grain of salt. However, if I had to pick a few to recommend, these are probably my favorite five. Note that any of them in isolation is also likely to over/under-fit. However, in combination, I think they do give a great way to start addressing culture: