A unified schema means every platform in an API answers in the same shape. Ask for a profile on one network or another and you get the same field names in the same places, so your code has one parser instead of ten.
That is a real saving. It is also the easiest place in a data product to tell a small lie at scale.
What it genuinely solves
Platforms disagree about everything. One calls it followers_count, another followerCount, another puts it three levels deep in a hydration blob. One returns an ISO timestamp, another a Unix epoch, another a relative string. Normalising that once, centrally, and testing it is work that should not be repeated in every client.
The first way it lies: inventing a field
If a platform does not expose something, a unified schema has three options for that key:
| Option | What the consumer reads |
|---|---|
| Omit the key | "This platform is different" — but now the shape is not unified |
Return null |
"We do not have this" |
Return 0 or false |
"This account has none" — a claim the platform never made |
The third is the lie, and it is popular because it keeps dashboards from showing gaps. A false in a verification field says the account is unverified. A null says nobody knows. On one federated network there is no identity verification at all — only link verification — so any boolean in that field is asserting something the platform does not measure.
The second way it lies: flattening different things
A star is not a like. A repository is not a post. A boost is not something the account wrote. When a schema maps all of those into metrics and posts_count because the shape demands a number there, the shape has started producing fiction.
The honest handling is to leave the field null and say why, which is what every one of our route pages does in a section headed what this endpoint does not return — for example the GitHub profile route, where post count stays null because a repository is not a post.
How to evaluate one
Ask a vendor which fields come back null for which platforms, and why. A vendor that has measured its sources can answer per field. A vendor that cannot will send you a coverage table with ticks in every cell, which is the answer you should worry about.