Who This Helps
Hey Junior Analyst. You just saw a key metric drop 15% week-over-week. Your team is asking why. This is for you. It’s a core skill from the Product Metrics Basics course. You’ll move from panic to pinpointing the root cause.
Mini Case
Priya’s activation rate fell from 42% to 36% in a week. The overall dashboard showed the drop but gave no clues. She felt stuck. By building one Segment Funnel Snapshot (a mission from the course), she isolated the issue to new users from social media ads. Their completion of a key step had plummeted from 70% to 48%. Boom. Root cause found.
Do This Now (5 Steps)
- Name the Dropped KPI. Write down the exact metric and the size of the drop (e.g., "Week 1 Activation down 6 percentage points").
- Pick Your Suspect Segment. Choose one user group you suspect is driving the change. Think: new vs. returning, sign-up source, device type.
- Map Their Funnel. List the 3-5 key steps that user segment must take to reach your KPI.
- Pull the Numbers. Get the conversion rate for each step for this segment, comparing the good period to the bad period.
- Spot the Leak. Find the single step with the biggest percentage point decline. That’s your likely culprit. Now you have a specific problem to investigate.
Avoid These Traps
- Don’t try to analyze all segments at once. You’ll drown in data. One segment, one funnel.
- Don’t skip defining the funnel steps clearly. Vague steps lead to vague answers.
- Don’t just look at the final conversion rate. The leak is almost always in the middle of the journey.
- Avoid jumping to solutions before confirming the leak. Diagnose first, prescribe second.
- Don’t use different time windows for your comparison. Keep it consistent (e.g., always Monday-Sunday).
Your Win by Friday
By Friday, you can walk into a meeting and say: "The 15% drop in our North Star metric is being driven by a 22-point fall in step three for users on mobile web. Here’s the data." You’ll have transformed a scary blob of a problem into a clear, actionable target. That’s how you ship clean analysis. Go find that leak!