The Truth Behind Absolute Averages

What's something that Product Managers use often but is actually quite misleading?

Averages.

Imagine if a T-shirt store decided the sizes they wanted to offer by taking an average of 10 people that visited them on a specific day.

They'd probably end up with a T-shirt size that fits very few of their audience members perfectly. Bummer.

In our recruitment products, we often try to quantify job posting behavior.

Taking a sweeping average is pointless. One on hand, we have Fortune 500s that post thousands of jobs while on the other, we have startups that seldom hire. An average would be massively skewed.

Some PMs love to quote average time on site but that suffers from the same problem. You end up conflating behavior of various populations: power users vs light users vs. idle etc.

Lessons for me:
1. Never use absolute averages. Go granular & analyze specific cohorts with similar properties.

2. If you want to negate the effect of outliers, consider the median of a data set.

3. The "average" is not equal to a "typical" customer.

4. Inspect raw data over time per customer & create a scheme to highlight anomalies/trends.