The cockpit combines sources, not populations

The selected cockpit period applies to activity where a dated event series exists. AEO and Commerce cards can instead show their own latest measurement snapshots. Source-specific labels are essential: selecting seven days does not reconstruct a fresh seven-day AI measurement when the latest stored analysis covers another window. Read the period under each card.

Search Console is an authorised property report

Search Console data comes from the connected property and returned dimensions, not a general crawl of Google's search results. Clicks, impressions, click-through rate, and average position describe that property and report scope. CTR is clicks divided by impressions, not the mean of independently rounded daily percentages. Query-level and page-level views can have different completeness and grouping, so compare like scopes.

Reporting delay and the effective end date

Google's reported data can lag the current day. Preset cockpit search periods use the last reported day as their effective end where supported. The activity chart can still include today and mark it provisional. A requested period, an effective source period, and a last import time are therefore different labels. A fresh connector sync cannot manufacture a day the upstream source has not reported.

Unknown is a gap, not a flat zero

A missing daily observation remains missing unless coverage establishes that zero events were measured. A chart should not interpolate a convincing performance trend through unmeasured days. A line also needs at least two comparable observations; one saved score is a point, not evidence of a rising line. Tooltips and legends should state which series is measured and which period is incomplete.

Chat, lead, and advisor activity

Chats, leads, and advisor events are independent measures. A lead count is not an attributed purchase conversion rate, and a product-click event is not an order. Lead coverage is constrained by available event history and the applicable period; the current day can be provisional. Timezone handling matters near day boundaries, particularly when a dashboard uses Europe/Berlin dates and another data source uses UTC weeks.

When a delta is justified

Use complete, equal-length, non-overlapping periods with compatible definitions and coverage. A rate difference is expressed in percentage points unless explicitly labelled relative change. Snapshot-to-snapshot comparisons additionally need compatible question cohorts, model/method, business objective, and formula. 'No comparable previous value' is more useful than a number manufactured from an incompatible baseline.

Troubleshooting a mismatch

First identify the source and exact metric. Then compare requested versus effective dates, property/workspace, grouping dimensions, import freshness, eligible records, and missing-day coverage. Check whether the card is an activity total or last snapshot. Report an issue when the same scope and metric still disagree; do not hide a mismatch by changing the label to an unrelated concept.

Compare statistics without mixing time periodsAEO rates, retailer wins, and unclear answers explainedWeekly account synchronisation and background jobs