If you don't understand a club's workflow, you can't sell them your metrics

You gotta spell out in detail how clubs can use your metric/model

Before I begin, just a reminder that I’m expanding the weekly output of the Soccer Analytics newsletter to 2-3 posts a week, with one free one. The paywall is going up on November 1st, so consider this month your free preview. And if you want to sign up ahead of time, please:

Otherwise, enjoy! And if you like what you read here, please share it! I nuked my twitter account, and so I’m relying on you to help me get my work out into the world.

Understanding club workflow

You’ll find it in pretty much any analytics post that presents some sort of new metric or model. It usually comes toward the end of the post.

“I believe this [metric, model whatever] has good applications for both recruitment and performance analysis.”

Sometimes, there may a little more in the way of added information here, but that’s pretty much it.

In fairness, it’s not the analyst’s job to try to make clear, in explicit terms, the specifics of how their metric might realistically be applied.

But, in my experience, unless you spell this out, clubs are not going to use it no matter how powerful it is unless they also hire the analyst who developed it (which maybe isn’t such a bad thing). And if you come up with a metric that won’t realistically fit in with most clubs’ workflows, congratulations—you’ve given birth to a beautiful loser.

Too often, I think, in the analytics sphere, the how of the model or metric is left up to teams to figure out on their own. Maybe it’s because every club has their own unique workflow, and analysts don’t want to presume how teams actually operate. A lot of the time, it’s because analysts lack concrete knowledge of how teams operate in practice. They’ll need to peek behind the curtain before they can recommend how to apply, say, a predictive possession-based metric to something like recruitment.

But this information is potentially critical in selling a good metric. For one, before you can recommend it to a club, you might need to know what kind of data a club has access to or is willing to pay for. You’d also want to know who’s in charge of what, and how amenable they would be to incorporating abstract metrics into their decision-making process, and whether your metric will really help. And, in terms of things like recruitment, you’d want to know the relative size of the player pool they’re working with, which may be limited by things like who the manager personally knows, or which agent the director of football generally relies on to get these deals done.

These may seem like distractions from the main event of your really cool metric, but in reality, they are vitally important for that metric’s future success. They mean the difference between a beautiful and accurate and predictive KPI that no club has a practical use case for, and a decent but less-than-stellar metric that clubs will use because it doesn’t require a lot of data to generate and won’t be applied to a giant pool of prospective players a team might not ever realistically look at anyway.

What is ‘workflow’ exactly?

I’m sure there’s some stuffy corporate definition online somewhere, but a workflow is essentially how a club moves from the raw fact of its existence to fielding a team of eleven players from god knows how many clubs and countries and academies who know more or less what their job is.

It’s the nitty-gritty, soup-to-nuts details of a recruitment strategy, of training, physio, tactical preparation (how a manager gets ideas from his brain onto the pitch).

And it’s something analysts need to know intimately in order to make their case.

Obviously, workflow will differ from club to club. Some DoFs might come up with a list of potential transfer targets and present them to a manager, or they might collaborate with a manager to come up with a list that they have to present to the chairman (I honestly don’t know, it’s all a vague soup to me).

But football is still largely an old boys’ network, and most will at some point adopt something that resembles ‘industry best practices.’ So there will be a lot of similarities from club to club.

The problem is, save for the insight of a few former club insiders, it’s incredible how little we know about club workflows! Yet this information is absolutely critical to selling analytics to prospective teams.

Let me try to offer you a vivid example.

Let’s say you’ve come up with a brilliant way to figure out the value a player of any position adds to a given possession based on, say, the likelihood their actions will contribute to a goal. One of Opta’s analysts just did this the other day, by the way.

One of your aims is to aid clubs in more effective performance analysis. So first, how soon after the most recent match can they access this information? In what form would it appear? Who will have access to it? The first-team coach? The director of football (perhaps they want to know who to sell on)? The performance analyst?

If it’s the in-house PA, how will they incorporate it into their own model or database, and who are they specifically talking to on the team? If it’s the first-team coach, how might they translate the metric’s findings into an actionable change? Particularly if they think in terms of traditional tactics?

Don’t advertise the stock, advertise the return

What I’m trying to say is it’s not enough for analysts to sell good metrics. You have to make a vivid use case of who on the team might use them in what kind of capacity, and to what specific advantage, as part of the club’s weekly workflow.

Because when you sell the metric alone, you’re selling to people who already value metrics. And that’s not always going to get the most from these valuable pieces of information. Because what can invariably happen is that they go into a quiver of tools a PA will share with a particular decision-maker in the team, and the PA’s advice will either be accepted or ignored based on whether who they’re talking to already agrees with them or not.

Far better to present a good metric but then couch it into a specific, tangible use case. So, in the case of Nils McKay’s cool new stat, you might present an example that reveals how a well-known player might, in fact, be massively overvalued, and the savings selling that player before their contract comes up.

This is a bad example because, like the vast majority of people, I don’t know the workflow of most clubs. But there are a lot of analysts that do, analysts who still publish public stuff, analysts who might not only be able to develop great predictive metrics, but who can also spell out a convincing example of how they might realistically be used by your average club.

And they should do it in each and every post!

Some stuff

• This is a vital post on data discrepancies and how they can fuck up things like xG.

• Grace on some statistical oddballs emerging this season.