health-readlisted

# Health Read A health score is a hypothesis. Almost nowhere is it treated as one. Most scores were designed once, in a room, with weights set by intuition. They were never back-tested against the accounts that actually churned, and they are never revisited when one of them does. That is not a scoring accident. It is a method nobody validated, producing a number everyone acts on. The failure this exists to prevent: **"the score was green" offered as an explanation for a churn, when the score being green is the thing that needs explaining.** --- ## What this needs **Shared context.** If an `account-context` document exists, read it first. What healthy usage looks like in this business is the single most important input here, and without it a low number and a normal number are indistinguishable. Where it is absent, carry on and name the assumption. **Minimum: the score and what goes into it,** even roughly. If the user does not know the inputs, that is the first finding. **Better with** the input values for this account, the trend over the last two or three periods, and how the account compares to its own history. **Best with** the list of accounts that churned in the last year and what their scores were three months before they left. That single list settles most of what follows. --- ## Step 1: Establish what the score actually measures Before reading the number, get the inputs on the table. For each one: - **What it is** - **Its weight**, if anyone knows - **Who