Diagnose a KPI Drop: Junior Analyst's 5-Step Fix

Who This Helps

This is for junior analysts who get a Slack message that says "revenue dropped 12% this week" and need to respond with a clear, useful answer. If you're in the Finance Basics for Operators course, you already know cash and profit tell different stories. This is the same thinking, applied to any KPI.

Mini Case

Viktor, a junior analyst at a SaaS company, saw weekly active users drop 12% in 7 days. His boss wanted one root cause and one recommendation by end of day. Viktor used the approach below and found the issue was a broken onboarding email link. He shipped a fix in 3 hours.

Do This Now (5 Steps)

1. Define the KPI exactly. Don't assume. Write down the formula. For example, "weekly active users = unique users who completed at least one session in the last 7 days."

1. Check the data source. Is the drop real? Maybe a tracking pixel broke. Look at raw counts, not just percentages.

1. Segment the drop. Break the KPI by channel, region, or user type. Viktor found the drop was 90% in new users from email campaigns.

1. Find the change point. What changed in the same time window? Viktor checked deployment logs and saw an email template update 8 days ago.

1. Form one recommendation. Don't list 5 options. Pick the one with highest impact. Viktor's recommendation: "Revert the email template and test the onboarding link."

Avoid These Traps

• Don't blame the data first. 80% of KPI drops are real. Check the data, but assume it's correct until proven otherwise.
• Don't over-analyze. You don't need a 10-page report. One page with the root cause and one recommendation is enough.
• Don't forget the business context. A 12% drop in free users matters less than a 5% drop in paying customers.
• Don't skip the fix. Your job isn't done when you find the cause. Ship the recommendation.

Your Win by Friday

By Friday, you'll have a repeatable process for any KPI drop. You'll be the person who answers "what happened?" with a clear, data-backed answer. And you'll do it in under 2 hours, not 2 days. That's the kind of analyst every team wants.