Delta Audit Tool

Features

• Rolling window slider (3-14 days, default 7)
• Detects the single largest shift in the rolling average of daily clicks
• Weekly aggregation with weeks ending Monday, partial weeks excluded
• Plotly interactive line charts with the change date marked
• Colour-coded delta indicators for clicks/impressions
• Custom column names via sidebar text inputs (exact, case sensitive)

How to use

1. 1 Export GSC data as CSV with date, clicks and impressions columns
2. 2 Type the exact column names in the sidebar (case sensitive; defaults are lowercase)
3. 3 Set rolling window size (3-14 days)
4. 4 Review the detected significant change date and week
5. 5 Explore interactive Plotly charts
6. 6 Analyse the week-before vs week-after comparison table
7. 7 Download the comparison report CSV

Frequently asked questions

Why does the tool say my columns are missing when they are clearly there?

Column matching is exact and case sensitive, and the defaults are lowercase 'date', 'clicks' and 'impressions'. A Search Console UI export (Dates.csv) capitalises its headers as Date, Clicks and Impressions, so type the actual names into the sidebar fields. If your export has multiple rows per date (for example date plus query), rows are summed per date automatically, so any dated GSC export works.

How is the 'significant change' date actually found?

The tool computes a rolling average of daily clicks (window 3 to 14 days, default 7), takes the day-to-day difference of that average, and flags the single date with the largest absolute change. Detection uses clicks only; impressions appear only in the week-over-week comparison table.

Can it detect more than one algorithm update or traffic event?

No, it surfaces exactly one date: the maximum deviation in the rolling average. If your date range spans several updates, only the biggest shift is flagged. Re-run the tool on narrower date ranges to isolate other events.

Why do I get 'Not enough data to compare the weeks'?

The comparison table needs a full week of data both one week before and one week after the significant week. If the detected change sits near the start or end of your export, one of those weeks does not exist and the comparison (and CSV download) is skipped. Export a wider date range around the suspected event.