Aggregate metrics hide exactly the customer who is about to churn. Adding a tenant label to everything is the obvious fix and it is also how you get a cardinality explosion that takes down the monitoring before it takes down the platform.
Updated Aug 2026 · Grounded in real Forward Deployed Engineer interview loops and written to a senior-engineer editorial bar.
Aggregate metrics hide exactly the customer who is about to churn. Adding a tenant label to everything is the obvious fix and it is also how you get a cardinality explosion that takes down the monitoring before it takes down the platform.
The senior signal is recognising that per-tenant visibility and metric cardinality are in direct tension, and pricing that tension rather than picking a side. The reserve probe is what happens when tenant count grows tenfold: a design that multiplies every series by tenant works at forty and dies at four hundred, and a candidate who has run this has an opinion about which metrics get the label. Strong answers alert on any-tenant-breaching rather than on the aggregate, and separate cheap always-on counters from expensive per-request detail. Watch for anyone who labels histograms per tenant without noticing that each bucket is its own series.
No comments yet — be the first to share your approach.
