A small studio can't staff a platform team, but it still has to know — fast — when something breaks in production. The answer isn't a sprawling observability suite; it's a tight set of signals wired straight into the tools the team already lives in.
Three signals that matter
The studio watches three things on every product: crash-free session rate (is the app falling over?), error rate on critical paths (auth, payments, programme writes), and a few business heartbeats (sign-ups, workouts logged). Everything else is noise until one of those moves.
The stack
- Firebase Crashlytics — crash-free rate and stack traces, mobile + web
- Cloud Logging + Error Reporting — server-side errors grouped and de-duplicated
- Cloud Monitoring alerts — fire only on the three signals above, nothing else
- A single Slack channel — every alert lands here; if it's not actionable, the alert gets deleted
Alert discipline
The fastest way to make a small team ignore alerts is to send too many. The studio runs a hard rule: every alert must be actionable and rare. An alert that fires daily and gets dismissed daily is worse than no alert — it trains the team to ignore the channel. Noisy alerts get tuned or deleted within a day.
The weekly review
Once a week the team looks at the trends, not the alerts: is the crash-free rate drifting down, is a particular error climbing, is a funnel step leaking. Catching a slow degradation early is worth more than reacting to a single spike — and it's the part most small teams skip.
The fastest way to make a team ignore alerts
Send too many. An alert that fires daily and gets dismissed daily trains everyone to ignore the channel — so when the real one fires, nobody looks. The hard rule we hold: every alert must be both actionable and rare. A noisy alert gets tuned or deleted within a day, because a channel full of noise is worse than no monitoring at all. Discipline about what not to alert on is the whole skill.
Three signals, then stop
A small team cannot watch a hundred dashboards, so we watch three things: crash-free session rate, error rate on the critical paths (auth, payments, data writes), and a couple of business heartbeats like sign-ups and core actions. Everything else is noise until one of those moves. The weekly job is to look at the trends, not the alerts — catching a slow degradation early beats reacting to a spike, and it is the part most teams skip.