De-identify a fact pattern in 60 seconds without losing the question

1 · Learn the move · Template pattern

The question survives de-identification; the client does not need to. 'Taxpayer A', 'Entity B', '$X' and 'a city in the Mountain West' carry every tax issue the real facts carry, and none of the identity. The template is four moves. Strip identifiers: names, numbers, addresses, dates of birth. Generalise narrowing facts: the exact figure becomes a bracket or a placeholder, the city becomes the region, the industry becomes the kind of business only if the kind matters to the question. Keep the facts the answer turns on: the use, the dates relative to each other, the elections made, the entity type. Then read the result as a stranger and ask whether it points at one person. One honesty line, every time: removing names reduces risk, but it may not de-identify an unusual fact pattern, and whether a de-identified question falls outside s. 7216 is unresolved [M]. A client whose situation is one of a kind is still identifiable with every name gone. The model can run the template; it cannot tell you the result is safe.

De-identify this fact pattern for a research question. Replace every name with Taxpayer A, B or Entity A, B; every Social Security number, EIN and account number with [REDACTED]; every exact dollar figure with $X, $Y or a rounded bracket; every city with its region; every exact date with a relative date (year 1, three months later). Keep every fact the tax question turns on. Then show me a two-column diff: what you removed, what you replaced it with. End with the facts that still narrow the set, if any, and the line that this is not a safe harbor.