Export routes
Dashboards — export the view you are looking at, filtered to the period and segment you need. Through the assistant — ask Claude or ChatGPT for the data and have it formatted however you want. This is usually the fastest route for anything non-standard:Export last month’s organic performance by landing page as a table.
Give me a summary of every fix shipped this month, grouped by type.
Pull ROAS by campaign for the last 90 days so I can paste it into a sheet.Agent Activity — the log of every action taken, with timestamps. This is what you want for showing a client what was actually done.
For client reporting
Agencies typically need three things per client, and all three come from the workspace:- What was done — Agent Activity, timestamped
- What changed — performance against the recorded baseline
- What is next — prioritised findings not yet actioned
Practices worth keeping
Export against a fixed baseline. A report showing only current numbers is not a report. Record the starting point at onboarding and compare to it every time. State what has not moved. Reports that only show improvements stop being believed the first time something goes wrong. Including flat and negative movement is what makes the good numbers credible. Keep periods consistent. Comparing a 31-day month to a 28-day month, or a period containing a sale to one that does not, produces conclusions that are artefacts of the calendar. Note the latency. If the report covers a month that included content publication, say plainly that organic results are not expected yet — 2–6 weeks for ranking movement, longer for AI citations. See SEO reporting.Data caveats to carry into exports
- Search Console lags 2–3 days. The most recent days are always incomplete.
- Platform numbers disagree. Google Ads, Meta, GA4, and Shopify use different attribution. Do not present a reconciled figure that none of them would produce.
- AI citation counts are small and volatile. Present them as trends over a quarter, not as week-on-week percentages.