Who This Helps
This is for you, the Junior Analyst, when your boss asks why a key number dropped 15% overnight. It’s part of the Data Reliability Leadership course, which helps you build trust by fixing problems calmly.
Mini Case
Your daily active user report shows a sudden 20% dip. The team is guessing: ‘Is it the new app version?’ ‘Maybe the tracking broke?’ Last time this happened, it took 3 days to find the real issue—a backend data pipeline failure. Let’s fix that.
Do This Now (5 Steps)
- Pause the guessing. Call a 30-minute huddle with just the core people needed: you, the data engineer, and the product owner.
- State the contract. Open your metric definition. Say: ‘Our contract says DAU is counted from this source table. Let’s verify that first.’
- Check the source. Look at the raw data pipeline for the last 24 hours. Is it running? Are rows missing? This takes 5 minutes.
- Check the calculation. Run your core analysis query on the raw data from yesterday and today. Do the numbers match your report’s logic?
- Declare the finding. In the last 5 minutes, agree on the one most likely root cause. Is it a data break, a logic change, or a real user shift?
Avoid These Traps
- Don’t invite 10 people to the call. Chaos slows you down.
- Don’t start by blaming the new feature. Check the data first.
- Don’t skip defining the metric at the start. You’ll all talk about different things.
- Don’t let the meeting run over 30 minutes. Timebox it to force focus.
- Don’t forget to note what you checked and what you ruled out.
- Don’t move to ‘fix it’ mode before you agree on the ‘why’.
- Don’t use jargon. Say ‘the pipeline stopped at 2 AM’ not ‘the ingest job failed’.
- Don’t forget to tell your stakeholder you’ve found the cause and what’s next. A quick Slack update works wonders.
Your Win by Friday
Run one practice triage on a stable metric this week. Grab a teammate and walk through the 5 steps in 20 minutes. You’ll have a calm, repeatable playbook for the next real alarm. You’ll ship analysis with clear recommendations because you know exactly where the problem started. And you’ll look like the person who keeps a cool head when the numbers get wobbly—pretty neat, right?