Then ninety percent of the actual business happens somewhere else: the hundredth onboarding email, the routine invoice, the support ticket about the same thing as last week. Nobody rehearses those. They are not important enough individually to warrant the attention.
Which is exactly why quality decays there first, and why you usually find out months later.
Table of Contents
Why routine work degrades and nobody notices
A big moment has an audience. Multiple people look at the deck, the contract gets read twice, someone checks the numbers. The stakes are visible, so the attention arrives on its own.
Routine work has one person and no audience. They have done it ninety-nine times, they know what they are doing, and the time pressure is real. So the version that gets produced is the fastest acceptable one, every time, and acceptable slowly shifts.
Nothing breaks. That is the problem. Each instance is fine, so there is no failure to investigate, and the drift only becomes visible in aggregate, usually when somebody outside looks at a sample.
The most thoroughly measured example of this
American healthcare documentation is the best-audited version of this failure, because a federal agency samples it and publishes the results.
A hospital stay generates documentation from a dozen people over several days. High stakes, many hands, detailed record. A routine clinic appointment generates one note, typed by one clinician in whatever time is left after the conversation. Nothing dramatic happened, so nothing detailed gets written. “Doing well.” “Stable.” “Continue current medication.”
Each note is accurate. In aggregate they produce a file that cannot support the conclusions in it.
The US Office of Inspector General audits exactly that. Across diagnoses it considers high risk, it reports roughly 70 percent were not supported by the associated medical records, with some categories above 90 percent. In one audit of 97 records carrying an acute stroke diagnosis, none were supported as acute strokes by the documentation. In 68 of those cases the patient had genuinely had a stroke, years earlier, and the record still described it as current.
Nobody fabricated anything. Somewhere in a series of unremarkable appointments, a condition got carried forward without anyone noting it had become history.
Three things this should change about how you run things
- Audit the routine, not the exceptional. Your QA probably checks the big deliverables, because those feel like where the risk is. Pull twenty of your most ordinary outputs from last month instead, the standard invoices or the standard onboarding sequence, and read them properly. That is where drift lives.
- Watch for summary language. “Handled”, “sorted”, “client happy” in a CRM is the business equivalent of “stable”. It records that something happened and nothing about what. Six months later it supports no decision at all.
- Fix it at the point of work, not at review. A field that prompts for the specific thing people habitually skip costs ten seconds in the moment. Catching the omission in a review three weeks later means reconstructing it from memory, if the person is even still available.
The general principle
Quality systems get designed around the moments that feel risky. The actual exposure accumulates in the volume work that feels safe, one acceptable shortcut at a time, until a sample shows you what the aggregate looks like.
Healthcare has started addressing its version directly with outpatient cdi software, which prompts clinicians for the missing specificity during routine appointments rather than sending a reviewer to reconstruct it afterwards. The target is deliberate: not the dramatic cases, the ordinary ones that make up the bulk of the record.
The equivalent in your business is less sophisticated and the logic is identical. Find the work nobody checks because it never goes wrong, and check it.
