Define the data contract
Record the measured property, time zone, period, filters and column definitions. Sessions, users and events are different units. Maintain separate raw-data, calculation and presentation sheets. A new export should not erase the assumptions attached to the previous one.
Normalise without hiding errors
Check date formats, decimal separators and parameterised URLs. Distinguish missing observations from observed zeros. Do not silently replace invalid records with blank cells. Group totals should reconcile with the overall total, and any exclusions should be recorded.
Share conclusions with limitations
Compare equivalent scopes and identify tagging changes. Direct traffic does not necessarily mean someone typed the address. Differences between systems can reflect definitions, consent and measurement methods. Attach filters and dates to visualisations so another analyst can reproduce them.
- Definitions
- Raw export
- Checks
- Interpretation
Worked example
A monthly report can contain date, channel, sessions, defined conversions and source filename. When a tagging change removes a conversion observation, label the period non-comparable rather than concluding that customer demand fell.
Check before moving on
These checkboxes are not saved.
Sources & method
Sources reviewed on 1 October 2026. Worked calculations use explicitly hypothetical inputs.
Topic provenance: Gros plan sur Excellent Analytics, le plug-in Analytics pour Excel (2010-02). Search Engine Feng Shui. Byline retained in the publication feed: fhouste.
This guide is newly written in 2026. It does not reproduce the earlier text and is not attributed to its historical author.