MEPX
Chapter 4 of 10All chapters
  1. 1What is one row
  2. 2Rates and denominators
  3. 3Beyond the average
  4. 4Compared with what
  5. 5Aggregation traps
  6. 6Noise and significance
  7. 7Metrics and proxies
  8. 8Missing data
  9. 9Reading forecasts
  10. 10A reading routine

Chapter 4 of 10

Compared with what

The missing half of most claims.

A baseline

A number alone cannot be good or bad. Previous period, a control group, a competitor or a target gives it meaning, and choosing the comparison shapes the conclusion.

  • Year on year comparisons handle seasonality better than month on month.
  • Compare like with like: changed definitions break time series silently.

Before and after

A change after an intervention is not evidence the intervention caused it. Without a comparison group, seasonality and trend are equally good explanations.