Product, metrics, and who to serve 📊

Thursday, April 12, 2018 :: Tagged under: management culture. ⏰ 6 minutes.

Hey! Thanks for reading! Just a reminder that I wrote this some years ago, and may have much more complicated feelings about this topic than I did when I wrote it. Happy to elaborate, feel free to reach out to me! 😄

The song for this post is Stress, by Justice.

I wrote a few thoughts on Twitter the other day and I think I'll elaborate on them.


Paperclips at all costs

There's a risk when you pick a specific metric that is anything but the core of your org's mission, you'll end up taking away from your org's ability to succeed in that mission. If your mission is "to remove Weebles" and you measure that by some other value "Pobbles", you'll end up with a strategy based on Pobbles instead of Weebles. It sounds obvious, but this doesn't prevent it from happening, especially in larger orgs with lots of nested teams.

As an example, let's consider MMR in video games:

When you queue up for a match in StarCraft, the algorithm does its best to match you up to someone of a similar skill. But "skill" is fuzzy; you could judge a player by any number of dimensions:

All of these matter, but for determining your matchmaking rating, the only thing the game cares about is did you win or did you lose. The rest of the metrics are useful to me as a player, but if any of the above was weighed into rating me in the matchmaking system, weird things would happen. For example, suppose they weighed players who killed more units somewhere into the rating:

Suddently, your MMR system becomes some weird hybrid that isn't totally removed from skill, but it isn't entirely about someone's ability to win a game, either.

If your organization is large enough, it'll need teams with specialized roles, and it stands to reason that they can't all have the exact same goals. However, if you go too many levels removed from the top-level metrics, your company as a whole may end up with lumpy results. You also risk teams with competing or zero-sum metrics having an unhealthy, undercutting relationship.

Rat Effect

The Wikipedia description is pretty good:

[...] in Hanoi, Vietnam, under French colonial rule [...] the colonial regime created a bounty program that paid a reward for each rat killed. To obtain the bounty, people would provide the severed rat tail. Colonial officials, however, began noticing rats in Hanoi with no tails. The Vietnamese rat catchers would capture rats, lop off their tails, and then release them back into the sewers so that they could procreate and produce more rats, thereby increasing the rat catchers' revenue.

If you do everything with metrics, your organization risks becoming a paperclip maximizer: if you attack anything but True North, you'll go somewhere other than True North.

Kind of Blue

Do you feel like Google was a leader in the Design space in 2009? Do you know many amazingly-designed Google products from then, or meet any folks who did design for them who went on to make a splash?

Most didn't because it was so data-driven, and designers weren't allowed to design:

[...] With every new design decision, critics cry foul. Without conviction, doubt creeps in. Instincts fail. "Is this the right move?" When a company is filled with engineers, it turns to engineering to solve problems. Reduce each decision to a simple logic problem. Remove all subjectivity and just look at the data. Data in your favor? Ok, launch it. Data shows negative effects? Back to the drawing board. And that data eventually becomes a crutch for every decision, paralyzing the company and preventing it from making any daring design decisions.


Yes, it’s true that a team at Google couldn’t decide between two blues, so they’re testing 41 shades between each blue to see which one performs better. I had a recent debate over whether a border should be 3, 4 or 5 pixels wide, and was asked to prove my case. I can’t operate in an environment like that. I’ve grown tired of debating such minuscule design decisions. There are more exciting design problems in this world to tackle.

Doing things in the absence of data or process is foolish, but you also can't use data to substitute vision or leadership. Use data as a tool to improve your product, not to abdicate responsibility.

Why does this happen?

Ass covering

Product Managers and Engineering Managers, like most people in a job they like, don't want to get fired. One way not to get fired is to reduce risk in product launches, and to have an unimpeachable process to determine your product offering.

Deferring to data is a good way to achieve both of these goals: nobody will fault you if can point to data, whether or not the feature succeeded. Like security theater, it matters most that you were seen to implement a sound process. A solid negotiation strategy is to appeal to a higher authority: what higher authority is there than data?

Visible impact, narratives around intelligence

PMs and EMs, like most people in a job they like, enjoy feeling impactful, important, and necessary. One way to do that is to insist on process and documents, and to make them as fancy as possible. Much like an engineer applying SingletonAbstractProxyFactory patterns for creating an object with two variants, if you ship process, it's tempting to ship a featureful, complex process when your team and its goals may be served by something simpler.

I mentioned a related point in my bit about meetings (emphasis not in original):

[...] there are people who thrive in dysfunctional meeting cultures like fungal spores in moisture, and will (often subconsciously) perpetuate and defend it, since it keeps the spotlight on where they are strong rather than where they are clueless. Commanding a high salary while always being in Important Meetings is one of the most accepted and desirable narratives in professional life, so to many people, this is success. It shouldn't surprise that there's an unwillingness to question whether all these meeetings are the expensive illusion of success. [...]

For many people, producing slick documents with Professional Language that spin tales about industries and consumer habits, full of percent signs and bar charts is what intelligence and risk reduction looks like. Many sell their value to organizations along these lines: without these charts, docs, or appeals to data, your org will make Bad Decisions.

There will be times when intuitions are incorrect and data will save the day. However, there also exist good decisions that don't require weeks or months of validation, product specs, and experimentation.

Just be careful

I ended my Twitter thread with "this is all wildly reductive," and this expanded blog post is too. Most of the processes I'm alluding to are themselves excellent responses to capricious, poor leadership, where an org will build things based on how the boss felt about the article they read over breakfast. Hiring competent and charming people who use data effectively is still a great strategy, all things considered.

That said, be wary if your org swings the needle too far the other way. It'll be a slow death, and you'll feel smart marching to it every step of the way.

Thanks for the read! Disagreed? Violent agreement!? Feel free to join my mailing list, drop me a line at , or leave a comment below! I'd love to hear from you 😄