What makes data “decision-grade”

“Decision-grade” gets used loosely, so it is worth being concrete about it. Data is decision-grade when someone can act on it without re-checking it first. That sounds modest, but most data fails the test. An analyst pulls a number, then spends twenty minutes confirming it against a second source before they will put it in front of a decision-maker. The number is usually correct. The problem is that nobody can tell that it is correct without redoing the work, so they redo it, every time.

Three properties separate data you can act on from data you have to verify. None of them is exotic. They are just rarely all present at once.

You can see where it came from

Provenance is the cheapest thing to add and the first thing missing. For any figure on a report, you should be able to answer: which feed, pulled when, transformed by which step. When a value looks wrong, the question is always wrong compared to what, and as of when. If the answer takes an afternoon to assemble, the data is not decision-grade. It is a starting point for an investigation.

It reconciles, and you know when it doesn't

Most real datasets are assembled from more than one source, and the sources disagree. Decision-grade data does not pretend otherwise. It states a reconciliation rule and reports the exceptions instead of hiding them. The failure mode is the silent merge: a job that quietly keeps whichever row it read last and presents a clean-looking result. The output looks finished. It is also unaccountable.

The unknowns are handled on purpose

Real feeds contain records that do not parse, do not match, or do not make sense. Two answers are common and both are wrong: drop them, which quietly understates everything downstream, or force them in, which corrupts the view people rely on. The right answer is to set them aside: quarantine the bad record, keep the main table clean, and leave a trail you can reprocess once the cause is understood.

The payoff is not a prettier dashboard. It is the time nobody spends re-checking, multiplied across every number and every day. Decision-grade data is quiet. It does its job and does not ask to be verified, which is exactly the point.