What is a manager for?

Soccernomics found that wages explain most of where an English club finishes over the long run. Google abolished its engineering managers in 2002 and brought them back within months. Both fields went looking for proof that managers don't matter, and both found an effect that is real, smaller than the myth and in odd places.

The most quoted attempt to show that football managers don't matter is a regression. In Soccernomics, Simon Kuper and Stefan Szymanski took the spending of forty English clubs between 1978 and 1997 and found that transfer outlay explained 16 percent of the variation in league position while wages explained 92 percent; rerun for the Premier League and Championship over 1998–2007, wages "still explained 89 percent." The number usually travels as "about 90%", stripped of its window and of the fact that it is a long-run figure across dozens of clubs and more than one division, which is a very different claim from "the wage bill decides this season" and leaves room for a manager. The cleanest attempt to measure that room, a Bundesliga paper by Hentschel, Muehlheusser and Sliwka, used the league's high manager turnover to separate manager from club and found that one at the 75th percentile of ability earns about 0.25 points a game more than one at the 25th, around 18% of the average points per game, with estimates that predict future results. Ranked by how far their teams finished above their wage rank over the five seasons to 2008, English football's best managers were Tony Pulis, Steve McClaren and Sam Allardyce, a list no leadership book has been built around.

Software ran the same experiment with fewer spreadsheets. In 2002 Google tried "a completely flat organization, eliminating engineering managers", and it lasted a few months, until "too many people went directly to Page with questions about expense reports" and interpersonal conflicts. In 2009 its people analytics group returned to a question that, by HBR's account, "had come up again and again since the company's founding: Do managers matter?" Neal Patel, co-lead of Project Oxygen, put the method plainly: "we actually ended up trying to prove the opposite case—that managers don't matter. Luckily, we failed." Project Oxygen drew on over 10,000 data points from performance reviews, feedback surveys and interviews, and found that teams with highly effective managers were more likely to get better results, be happier and turn over less.

What it found managers were for is the part that gets less attention. The eight behaviours in the original list, as the NYT reported them in 2011, open with "Be a good coach" and put "Express interest in team members' success and personal well-being" third, while technical expertise, the thing an engineering organisation would have bet on, "ranked dead last among Google's big eight." So both fields set out to show the manager was overhead, and both came back with an effect that is real, a good deal smaller than the back-page version of the job, and located somewhere other than tactics or technical judgement.

In my experience that location is the individual. Line management exists to serve the company, but you can only do it well by serving the employee, because the person you manage has to trust that you are there to help them do their best work, and that trust only comes from actually getting to know them. I've also found the reverse holds: at companies where the structure doesn't let you serve the individual, you start to feel quite quickly that you aren't adding anything, whatever the org chart says your job is.

Footballers describe the coaches they rate in the same terms, and Mourinho's Chelsea players are the clearest public case. Frank Lampard said in 2019 that Mourinho "was the first manager that day-to-day made me feel really close to him", and that "we worked on the mental side without working on it; it was just this confidence that he gave me which forms a bond." Didier Drogba told the BBC's Football Focus in 2018 that "when he decided to sign me he changed my life", and has described wanting to leave during a hard first season until Mourinho told him "No, you have to stay, I will help you, you must stay"; Drogba says the challenge of that conversation is "why I became the player that I was." John Terry, back under him in 2013 after being phased out by the previous regime, said Mourinho "showed faith in me, and to get that arm around you is something everyone needs, even if you are one of the more experienced players in the team". None of the three talks about the 4-3-3; they talk about being known, and about being pushed toward something they wanted for themselves.

The same manager is also the survivorship warning. In December 2015, seven months after that second Chelsea spell had run away with the title, he was sacked with the club a point above the relegation zone, the technical director citing "palpable discord between manager and players" and Mourinho saying his work had been "betrayed." Whatever he gave Lampard and Drogba, it could sour inside a season.

The analogy breaks at the scoreboard. Kuper and Szymanski could run their regression because both sides of it are public: a league table everyone agrees on, and wage bills taken from clubs' published accounts. No engineering organisation has either. Oxygen's outcome measures were Google's own reviews and surveys, the club marking its own homework, and its effect shows up as satisfaction and retention rather than points. The other break is in what the player wants: when Lampard or Drogba reached a personal goal, it arrived as a Chelsea trophy, whereas an engineer's goals can point somewhere the company has no use for, which makes "serve the individual" a harder sell to the people paying for it.

If the effect is real but narrow, the question is where it shows up, and this series takes five places, one post each. The first is the split football makes and software keeps merging, between the captain who leads on the pitch with, in the words of Law 3, "no special status or privileges", and the manager on the touchline; that is the gap between tech lead and engineering manager, and the home of what Tanya Reilly calls glue work. The second is the decision taken with the clock running, the substitution at 60 minutes or the incident at 2am, where Google's SRE book says its process "is based on the Incident Command System", which separates running the response from making changes to the system, much as the touchline separates the manager from the ball. The third is measurement, where football analytics is still trying to value the run that pulls a defender away and engineering is having an earlier version of the same argument. The fourth is load, which Klopp named when he left Liverpool saying he was "running out of energy". The fifth is succession, because what a manager leaves behind, as Moyes found out at United, tells you whether the team was good because of a system or because of a person.

That last one is where the two attempts to disprove managers come back round: football can watch a club after its manager leaves and see what remains. An engineering manager usually can't, and if the effect really lives in individual trust, a lot of it walks out of the door with them. How much of what a team does well would still be there the week after its manager left?

No comments yet