Brand drift
The accumulated distance between the brand a business decided on and the brand it is currently executing. It builds through small, individually defensible choices across colour, type, voice and positioning rather than through any single bad decision, which is why it behaves as a decay process with no alarm and no obvious moment to catch.
Brand drift is not Brand inconsistency.
Inconsistency is a snapshot: two artefacts sitting side by side that disagree, and you can fix it with an edit. Drift is the trend that produced them, measured over time against a stated definition, and editing one post does nothing to it because the cause is the missing force that was supposed to hold the definition in place.
In practice
Nobody rebrands by accident, and that is exactly what makes drift expensive. There is no event to point at, no artefact to blame and no error message. There is a body of work that looked coherent in March and reads like three different companies by September, assembled entirely out of choices that were each reasonable at the time. The useful mental model is not that a mistake was made. It is that no force was maintaining coherence, and coherence is not a brand's default state.
Drift accumulates in several dimensions at once, which is the second reason it hides. The colours are slightly off in three places, the writing has become a little more corporate, the audience description now names who you sell to rather than who you started for, and the logo has been used at four sizes with four different margins. Any one of those is a shrug. Together they are the reason a feed stops reading as one thing. The common sources are an offer that changed while its description did not, an audience that quietly shifted, voice eroding under volume, visual decisions re-made in the moment because the kit was in another tab, and new tooling re-asking questions you had already answered.
Nothing trips an alarm because brand coherence, unlike almost everything else in a business, has no detector. Revenue has a dashboard, uptime has a monitor, code has a test suite. Brand has historically had a person who happens to care and happens to be looking, which is not a detector but a hope, and it fails in the way hopes fail: it holds while attention is available and stops exactly when you get busy, which is when drift accelerates fastest.
Making drift measurable takes three things most setups lack: one authoritative description rather than a folder and a shared understanding; structure, because prose cannot be scored but fields can; and dependency awareness, so the system knows which outputs came from which fields and can flag what is now potentially stale. Score each field on its own and resist rolling up into a single number. "Audience: underspecified" tells you what to write next. A single brand-health percentage is a mood. This is what Repic's Quality Engine does, and it is worth building even without a tool.
The fix is not a rebrand. A rebrand replaces the definition; drift is the gap between whatever definition you hold and what you ship, so a rebrand simply resets the clock and drift begins again from the new starting point. The durable answer is a definition that is inspectable, a habit of reading recent output against it, and a system that tells you the size of a change when you make one deliberately.