Composite customer satisfaction
Every customer carries a single happiness reading -- from Delighted at the top down through Happy, Content and Unsettled to At risk. It folds three things into one number: are you keeping your uptime promise, do they trust you from how past asks have gone, and how do they feel lately. A name you can read at a glance, and a relationship you can lose.
A customer high on the scale refers friends and forgives the odd hiccup. One sliding toward At risk is a churn warning you can still act on -- catch a breach, say yes to a fair request, give them a clean stretch -- before they walk. Cramming on more tenants than you can keep happy shows up here first, customer by customer, long before it shows up in your reputation.
Detailed explanation
satisfaction() folds health (are they inside their availability target), relationship trust (the running record of honored vs refused asks), and lived sentiment (recent event mood) into a 0..1 score, mapped to a plain band (Delighted / Happy / Content / Unsettled / At risk). It is surfaced as the "Standing" column for customers and as the avg_satisfaction loop-stat across the active base.
Satisfaction moves before churn does, so the band is the actionable signal: an At-risk tenant is the one a clean hour, an accepted request, or a resolved incident can still save. It is per-customer where reputation is your public aggregate -- satisfaction is what you manage; reputation is what the market sees.