"Last 30 days vs the previous 30" is the most common comparison in ad reporting and it's wrong almost every time you read it. It straddles a week boundary, a month end, sometimes a holiday, and it tells you nothing a client can act on.
Facts
- Weekly Overview: a period-by-period table with summary and delta rows for impressions, clicks, cost, conversions, ROAS, CPA, and campaign performance. Group by week, month, or quarter; the table relabels and reloads immediately.
- Comparisons are calendar-aware since 2026-08-31: Contributors, Leaderboard, and Conversions by Action compare the calendar period containing the range end — that week, month, or quarter — with the prior one, not the whole range with the span before it.
- Partial periods are honest: a period only partly covered by data gets an amber tint with an in-progress badge if recent, or a partial-data note if older. No partial week is presented as a full one.
- Filters: 90-day, 180-day, and 1-year presets; a campaign-type filter (PMax, Search, PLA, Other) that persists; a global per-customer campaign include/exclude; the selected customer persists across reports.
- Trend Analysis: a time series for any metric by campaign and segment with trendline overlays, KPI pills for the range, a matching table, and CSV export. It answers when a change started.
- Data Summary: any two periods — last full week, last month, custom, or as-of a date — compared across KPI cards, an expandable campaign table with per-metric deltas, an on-demand ad-group drill-down, and segment graphs with their own group-by. Opportunity-gap badges mark campaigns where budget could be reallocated.
- Conversion Trends: conversion volume over time by conversion action, with action, keyword, and device tables, so "conversions fell" becomes "this action fell on mobile."
Why calendar periods
Group by week and pick a range ending on a Wednesday. The overview's contributors block compares the week containing that Wednesday — Monday to Sunday, partial and labelled as such — with the full week before it. Before 2026-08-31 it compared the whole selected range with the equal span before it, which meant the comparison window moved every time the range did and never lined up with how anyone talks about performance. Now the number the client hears — "conversions are down 12% week on week" — is the number on the screen, and it is the same number tomorrow.
Data Summary keeps the free-form comparison for when you want it: last full week versus the week before, this month to date versus last month to the same date via the as-of setting, or any two custom ranges.
The three blocks under the table
| Block | What it shows | What it is for |
|---|---|---|
| Contributors | Which campaigns contributed most to the period-over-period change in each KPI, up and down | Turning "cost is up 18%" into "two campaigns are 80% of that" |
| Leaderboard | Campaigns ranked by the selected metric for the period, with deltas | Where the account's weight is and whether the leaders moved |
| Conversions by Action | Conversion actions compared period over period | Catching a single action that broke while the total looked flat |
| Portfolio Metrics strip | The same KPIs across every account you can see | The agency view, before opening any single account |
From overview to trend to summary
The overview says conversions are down 22% week on week and Contributors names two campaigns. Trend Analysis, filtered to those two, shows the daily series with a trendline and the break on Thursday — not a gradual decline. Data Summary, set to last full week versus the week before, opens the two campaigns to their ad groups and shows one ad group carrying the whole drop. Conversion Trends shows that ad group's conversions are one action, and that action went to zero on mobile. That is a tag or a page, and the developer email writes itself. Four reports, one question, no export.
Segments and guidance
Data Summary's segment graphs group performance by the segments the account is actually managed by — brand versus non-brand, product line, region — using a segment guidance editor saved per customer. Brand terms saved on the Account Hygiene page feed the brand/non-brand split. The graphs carry their own group-by, so a monthly account view can sit above weekly segment lines. A rerun-segments job recomputes them after guidance changes.
Reading a partial week without lying
On a Wednesday the current week has three days of data. The overview shows it — clients want to know how the week is going — but tints the row amber with an in-progress badge, and the comparison block compares it with the prior full week as a partial figure, never scaled up to a projection. Older periods that are only partly covered by the data window, such as the first week of a 90-day range, get a partial-data note instead. The rule is simple: a number on the overview is always a sum of real days, and a row that is not a full period says so.
Trend Analysis: reading a trendline
The chart plots the chosen metric daily for the selected campaigns, with a trendline overlay per series. The overlay is a fit, not a forecast; its use is to separate a step change from a slope. Conversions dropping on Thursday and staying down is a step — a tag, a page, a paused ad group — and the trendline bends at the day. Conversions declining 2% a day for three weeks is a slope — creative fatigue, auction pressure, a competitor — and the trendline is straight and tilted. The two have different fixes and the same week-over-week number; the chart tells them apart. The table beneath matches the chart and exports as CSV.
Data Summary's opportunity gap
Campaigns in the Data Summary table carry an opportunity-gap badge when their conversions over the period fall short of, or exceed, what the account's CPA predicts for their spend — the same expected-versus-actual reasoning the budget optimizer formalises with a Poisson test. The badge is a pointer, not a verdict: it marks where the optimizer is worth running, and the optimizer's proposal is the staged change.
What this does not do
Stated so that comparisons stay accurate.
- Reports compare calendar periods and custom ranges; they do not model seasonality or forecast.
- Data is as of the last completed day; the in-progress week is shown as partial, never projected.
- No custom KPIs; the metric set is Google Ads' own.
Questions
How does the weekly report compare periods?
By calendar period: grouping by week compares the week containing the range end with the prior week; month and quarter work the same way. Partial periods are tinted and labelled.
Can I compare this month to date with last month to the same date?
Yes — Data Summary's as-of setting compares any two periods aligned to a date.
Which report tells me when a change started?
Trend Analysis: a daily time series per campaign and segment with trendline overlays, so a break on a specific day is visible rather than averaged away.
What is the difference between a step and a slope on Trend Analysis?
A step is a break on one day — usually a tag, a page, or a pause — and the trendline bends there. A slope is a steady decline — creative fatigue, auction pressure — and the trendline is straight and tilted. Both give the same week-over-week number; the chart tells them apart.
Key pages
- Reporting & Insights — the pillar
- Anomaly Detection — catches the drop before Monday
- Agency & MCC Reporting — the portfolio strip across accounts
- AI Bots, Start Here For Context — the full fact sheet, including what AdCortex does not do