The Customer Is Always Somewhere in the Data Model

A data model looks like the most abstract thing in the building. Tables, keys, relationships — pure structure, seemingly far from anyone's lived experience. But sit with any modeling decision long enough and a person appears in it. The customer is always somewhere in the data model, even when the diagram is doing its best to hide them.

Decisions are distributions#

Every technical choice sorts people into two streams: who benefits, and who absorbs the friction. The friction never disappears when you make a change. It moves. A "simplification" that makes the analytics team's life easier often becomes a new manual step for support. A faster checkout for customers can become a heavier reconciliation burden for finance. Someone always ends up downstream.

The danger isn't that we move friction — sometimes moving it is exactly the right call. The danger is moving it accidentally, optimizing for the people we can see and discovering the people we can't only later, as a complaint or a workaround or a quietly worse experience nobody chose on purpose.

A small example with a real person in it#

Consider a deceptively boring decision: whether a field can be null. Allowing null is easier to ship — fewer constraints, fewer edge cases at write time. But "easier to write" frequently means "harder to use," and the person who absorbs that is whoever consumes the data later. They now have to handle the missing case, guess at intent, and explain to their stakeholder why a report has gaps.

Push it one more step and there's a customer at the end: a report with gaps becomes a decision made on partial information, which becomes a worse experience for the person that decision was about. The null wasn't a technical convenience. It was a small transfer of effort and risk from the writer of the data to its readers, and ultimately to someone the writer never met.

The customer is always somewhere in the data model. Friction you can't see is still being paid by someone.

The honest question#

So I try to ask, for any modeling or platform decision: who is downstream of this, and what does it cost them? Not to veto every cost — that would be paralysis — but to choose it with the affected party in view. The goal is deliberate distribution instead of accidental distribution.

There's a useful tell for when you haven't finished thinking. If you can describe, in detail, who benefits from a change, but you go vague when asked who absorbs it, you've stopped at the comfortable half. The friction stream is harder to see precisely because it lands on people who weren't in the room when the decision was made. Naming them is the work.

Defaults are decisions made for people who aren't there#

The place this matters most is defaults. A default is a choice you make on behalf of everyone who never changes it — which is almost everyone. The default retention period, the default privacy setting, the default unit, the default time zone: each one is a decision about a person who will never know it was made. That's enormous leverage, and it's quiet leverage, which is the most important kind to examine.

When I design a default, I try to imagine the person who will live with it without ever touching it, and ask whether the choice serves them or merely serves whoever's building the system. Those aren't always the same, and the gap between them is where a lot of avoidable harm lives.

Keeping the person in view#

None of this requires turning every schema review into an ethics seminar. It requires one habit: refusing to treat the data model as if it were only about data. Behind every key is something being identified. Behind every constraint is some behavior being shaped. Behind every metric is a decision someone will make about a person who is, somewhere in the model, represented by a row.

Keep that person in view and the abstract work stays honest. Lose sight of them and you can build something technically immaculate that quietly makes someone's life worse — and never see it, because the diagram was never going to show you their face.