BuiltMind Pulse
Active per day
Active users per hour, last 48 hours
Prague timeRecently active
Client health
Clients with at least one enabled product and one active seat. Products are what the client has enabled (developers table), not necessarily paid. Sorted by risk: a renewal coming up combined with no use, falling use or an unused product. Some companies have one client per person (e.g. "YIT PL - <name>").
Exports, last 30 days
Excel and file downloads made through the API (price list, price update, project and market report exports), per client. Exports the browser builds by itself are not seen by the backend.
Login friction
Logins per user per active day, 30 days
Clients that log in most often
Failed logins, last 7 days
An unknown email has no user: Cognito does not say which email was typed. Days before 19 Oct have counts only, so the list fills in over the first week.
Features, last 30 days
New features
A feature is new for 90 days after it was introduced (catalog.json), or until a route nobody has named yet gets a reviewed name. Then it moves into the ratings above like any other feature.
Dynamic Pricing funnel
Clients with Dynamic Pricing enabled and an active seat, by the furthest step reached. Steps from the DB (ever) and price list use from the logs (30 days). Project dates before mid-2026 can be off: older projects got created_at when the column was added.
Stuck clients
Recommendation uptake
Catalog and coverage
Lambdas without request logging
Counted from invocations only: no client, and no route when the Lambda serves several.
Routes waiting for a reviewed name
Run python -m scripts.discover or the update-catalog skill.
Removal candidates: no calls in 30 days
Only routes whose calls can be seen individually. "No frontend call" means nothing in the app calls them either.
Needs attention
Pricing API, last 7 days
Requests per hour
Latency per hour
5XX responses per hour
server errors4XX responses per hour
mostly auth and validationPricing endpoints, last 7 days
Latency is the backing Lambda's duration: it leaves out cold starts and API Gateway overhead. "Errors" counts crashes only; a handled 500 response is not counted.