Data Analysis Best Practices

Chapter 1 · Practice 4 of 4

Learn how the data was collected before you analyse it

From Frame the question first

Why it matters

Before touching the main question, find out how each table is produced. Google's guide recommends confirming the collection setup, unusual time ranges such as holidays or launches, and which populations were included.

Questions to answer first:

  • What creates a row: a page view, a server event, a nightly snapshot?
  • Are bots, test accounts or internal users included or filtered out?
  • Did tracking change during your time range, such as a new app version or a renamed event?
  • Which time zone are the timestamps in?

A ten-minute conversation with the team that owns the data often saves days of analysing an artefact of logging.

Resources for this chapter

Go deeper