Read evidence like an operator

Most marketing conversations fail before they start, because they begin with a conclusion ("our SEO is bad," "social isn't working") and then go shopping for numbers that agree. An operator inverts this. She starts with the evidence map: what was observed, what is Unknown, and what would change her mind. The question is never "are we winning?" It's "what do we know, how do we know it, and what's the cheapest way to know more?"

This discipline has a compounding effect that most teams miss. When every claim carries its source and its confidence, disagreements get cheaper: instead of arguing about opinions, you argue about the next verification. "Our GBP listing is thin" becomes "direct scan only, medium confidence; connect GBP and re-check." The team stops performing certainty and starts producing knowledge. Over a year, the team that records Unknowns honestly tends to outperform the team that grades itself, because it keeps finding the real constraints instead of the flattering ones.

The third habit is the hardest: let the evidence change the plan without changing the past. When new evidence arrives, update the recommendation and say what changed. Don't quietly rewrite what you believed last month. Clients and teams forgive being wrong; they don't forgive discovering the story changed without them.

The Diagnostic enforces the vocabulary: every area is Observed or Unknown, observed areas carry a confidence (high, medium, or low), and every run states its evidence basis (a grounded read, direct scan only, or unavailable). Every task names its metric, source, and window under "How we'll know it worked" before the work starts, and the "Log a task outcome" panel keeps completed tasks in view until their results are recorded. Decision Memory (Command Center → Decisions) keeps the record of what was decided and why, so a changed recommendation can point to the evidence that changed it. And there is no composite score anywhere to hide behind.