Let’s play three quick coordination games:

  1. You and I each have to guess what the other person will answer to the question, “Name a British rock band that got very famous in the 1960s.” What do you guess?
  2. You’re on Family Feud and the survey question is, “Name a resort where kids want to go on vacation.” What’s the #1 answer?
  3. Imagine we’re both in Paris and absolutely have to meet up tomorrow, but we lose contact before deciding on a time or location. Where and when would you show up tomorrow to maximize our chances of finding one another?

Think of your answers, and keep them in the back of your mind for a paragraph while I talk about teamwork.

An intuitive model of teamwork is “we click as a team → so we work well together”. You see this when people can finish each other’s sentences, or when a pair of basketball players pull off an alley-oop, or in a design team that collaborates well. This has intuition going for it, but little explanatory power. Really, clicking is a synonym for working well together, and so can’t be the cause of it. The real question is, what does it mean to click?

Ok, now pause that while we go back to the coordination games. What were your answers for the three questions up top?

My answers: The Beatles, Disney World, and the Eiffel Tower at noon. There’s a good chance those were yours. If they were, we just clicked a bit. But what if you’d said that set of answers, and I’d said, “The Who, Legoland, and the Mona Lisa at 9 a.m.”? You’d think something like, “Those are horrible answers, why would you guess those? And let’s not meet up in Paris, you’re probably not going to fit in if those are your guesses.”

But they’re only horrible answers because we don’t know each other, which means the only viable strategy is to default to the universally most popular answers. Those answers are obviously not the universally most popular. Anybody who uses them either (a) doesn’t realize that those answers are non-obvious, or (b) is playing according to some other strategy they think is better. In either case, they’re ipso facto not going to be good teammates for Beatles-Disney-Eiffel12 players.

Now, though, suppose that we’re a married couple who’s obsessed with The Who, just took our kids to Legoland, and are actually in Paris right now with an existing plan to go see the Mona Lisa tomorrow at 9 a.m. Those answers are suddenly the best guess. Doubly so, actually: they’re our shared context, and because the context is unique to us, if we both play Who-Lego-Mona9 we’ll click especially well.

The band OK Go invented a game based on this to pass the time when they’re on tour. It’s called Say the Same Thing. You play with a friend. Here’s how it works:

The formal name for these things is “focal points”, or Schelling points after Thomas Schelling, the economist/game theorist who pioneered the concept. In The Strategy of Conflict, Schelling describes it like this (I’ve added the bold emphasis):

People can often concert their intentions or expectations with others if each knows that the other is trying to do the same. Most situations — perhaps every situation for people who are practiced at this kind of game — provide some clue for coordinating behavior, some focal point for each person’s expectation of what the other expects him to expect to be expected to do.

You keep recursing on expectations until you reach a stable outcome (that’s the goal of Say the Same Thing, and probably means you have a good answer), or until you exceed the maximum call stack size and have to figure out some other approach (e.g. “screw it, I’ll go to the Arc de Triomphe and wait there all day”).

In pseudocode, it would be something like:

function getFocalPoint(priorPoint) {
  // What you'd expect the other person to guess if
  // they know you're going to guess priorPoint.
  const newPoint = getPartnerGuessFromMyGuess(priorPoint);

  if (newPoint === priorPoint) {
    return newPoint;
  } else {
    return getFocalPoint(newPoint);

Of course nobody thinks that way literally, but it’s a model of what’s going on when you guess at a Schelling point. So with respect to teamwork (remember teamwork? It’s an essay about teamwork.), the question becomes how to guess Schelling points well as you build stuff.

Wait — what does this have to do with building stuff?

Well, what percentage of the time is someone looking over your shoulder at work? Even if they were, would they be able to spot a difference in real time when you make an A- choice vs. an A+ choice? What does an A+ choice even look like at the moment it’s made? Your decisions have to interact with your teammates’ decisions. Which means the “best” decision is partially a function of what your teammates are doing — or did a month ago, or will do next year. So your decisions have to align with decisions that you won’t learn about until after they’ve already happened.

That kind of alignment isn’t sufficient to get the team to the right destination. But it is necessary if the team’s going to get where they want to go. Whether “the right destination” and “where the team wants to go” are the same place is a separate question, but those high-level strategic choices are also the result of a team process, and the health of that process itself depends on your meta-choices being aligned.

So you can only thrive if your work augments your teammates’ work. And their work has to augment yours. And because it’s impossible even in theory* to communicate with anything like the fidelity you’d need for a literal step-by-step plan, you are in an unspoken game of Say the Same Thing with your coworkers, all day every day. If you play the game really well, congratulations, your work is going to align and you’ll probably create something cool. If not, it’s going to be painful.

(*Why’s it impossible? For a variety of reasons, starting with (a) you aren’t watching each other minute-to-minute, (b) even if you were, you can’t see in enough detail to see the thousands of incremental decisions people make as they work, (c) even if you could, there’s no form of communication that can adequately relay the context behind those decisions, (d) even if there were, you’d need to account for how those decisions recursively interact with the thousands of other decisions happening elsewhere on the team.)

You might have just thought, “But we do make step-by-step plans.” Sure, but the steps aren’t infinitely granular. It’s a question of how many games of Say the Same Thing happen between each step. Quick math: if each step is two days long, there are six people on the team, and each person makes just 100 microdecisions a day, then each two-day interval is hiding 6000 games.

The result: working together well means that when you all go off and do your thing, you can have faith that everyone’s making their thousands of decisions in a way that’s compatible with how you’re making yours.

This applies to some obvious things. Coordination between two engineers working on the server and client halves of an API. Between devops and database folks. Between design and front-end engineering. Marketing and product. Sales and customer service. Between the people within and across all those teams. Etc.

It also applies in less obvious ways. The unspoken nuances of how you run meetings. The way you hire. The way you negotiate or have difficult conversations. How you make small talk. How you make big talk. When people get to work, what they they wear, and how little or how much people care about those things.

There’s often no reason that any particular answer to a coordination question is objectively best. We can do code reviews in Way A or Way B. Whatever. If the team uses A and someone prefers B, maybe that someone is right! But if they start using B unannounced, and if A and B aren’t compatible … they’ll get the same response they would if, using the Paris example, you showed up at the Eiffel Tower and they proudly showed up at their favorite small boulangerie. A strong feeling of “I’m sorry, we’re just not clicking.”

So if being a great team means being great at finding Schelling points, how can we build that skill? High-quality day-to-day processes are awesome — kickoffs, retros, check-ins, well-run meetings (and not too many of them), etc. But they’re tactical, and (significant meta-content aside) focused on making the right decisions. Finding Schelling points is focused on executing the decisions you’ve made. And that means aligning on high-level goals and values.

“Aligning the team on high-levels goals and values” is such an important process that there’s actually already a short way to say it: leadership. This is the most important thing the boss does. They’re not the only one doing it (anyone on the team can help), but they are the one responsible for seeing that it gets done. Excellence in this is the defining trait of good leadership. Incompetence in it is the fastest way to sap a team.

That’s because the team’s goals and values are the framework everyone will use as their guide when they go off to make their thousands of decisions.

In other words: it’s the map to the Schelling points! If the map is bad, you’re never going to find them. If it’s good, you’ll have a shot. If it’s great, it’ll waze you right to them.

Note that this doesn’t mean everyone thinking identically. To the contrary, useful disagreement requires you to have an accurate model of what the other person’s thinking, what they expect you to be thinking, what they expect you to expect them to be thinking, etc. Schelling points again! Recalling Say the Same Thing, the trick isn’t to think the same way, the trick is to recurse all the way to the bottom of how the other person is thinking. Which is the ideal way to disagree constructively.

Take all this together, and it’s startlingly broad: your competence, as a team, at finding Schelling points is the limiting reagent for your ability to do high-quality work, hire well, disagree well, make small talk, negotiate, collaborate, plan a roadmap, and generally to feel like you click.

There’s a word for this broad category of “how well are we working together?” tests. It might have popped into your head already: culture. I think we can bring these ideas together, and formalize the definition a bit: your Schelling points are your culture. Having a lot of shared ones is the definition of an effective culture. Not having them is by definition a broken culture. Being great at finding Schelling points together isn’t enough by itself (you might be effective and going 200 mph in the wrong direction — against team blind spots, industry trends, competition, etc.). But it certainly is required. So it’s a good place to start.