Search Themes Drift Report
What changed in your search themes month over month and what to roll back.
Who this is for
- Monthly PMax review operators
- Multi-AG accounts with theme overlap
- Seasonal ecom verticals (apparel, gifting, holiday)
Why this exists
Search themes drift silently. A theme added in March looks productive in April and breaks profitability in June when seasonality shifts. This report makes drift visible so the rollback decision is data-driven, not vibes-driven.
Read this first
Search themes are signals, not directives. PMax can ignore a theme if it doesn't see learning evidence that the theme converts. That makes drift hard to spot from inside the campaign: a theme can stop carrying weight without any visible flag, and the spend it used to attract reroutes to whatever the algorithm finds easier. This monthly diff catches the moment that happens.
Run the monthly drift diff
Snapshot themes per asset group on the first of each month
Open PMax, asset group, insights tab, search themes. Export the list with conversion attribution per theme. Save the sheet with the date in the filename. The first month is the baseline. From month two on, you have something to diff against.
Diff month-over-month, not week-over-week
Search themes need 28-30 days of conversion volume to read cleanly. Week-over-week diffs fire false positives because the underlying signal is too noisy. The diff cadence is monthly. The decision cadence is monthly. Anything tighter is theatre.
Tag every change as added, removed, or drifting
Added: theme present this month, absent last month. Removed: present last month, absent this month. Drifting: present in both but the conversion attribution dropped 30% or more. Drifting is the category that matters: those themes are silently being deprioritised by the algorithm.
Cross-check against seasonality and search demand
Open Google Trends for the drifting themes. If search demand for the theme dropped in your geos, the drift is seasonal and the right move is to leave the theme in place but reduce its rank in the AG. If demand is flat or up but the theme is drifting inside PMax, the algorithm has decided the theme isn't converting; the right move is to either roll it back to a previous wording, archive it, or move it to a different AG with cleaner intent.
Apply the rule of three on theme density
Sweet-spot theme count per asset group is 3-5. Below 3, the AG is starving the algorithm of intent signal. Above 7, themes dilute and the algorithm picks the cheapest among them. While running the diff, also check density. If an AG has drifted up to 9 themes over the year, prune 4 before doing anything else.
Diff format that reads cleanly month over month
Paste this header into the monthly diff sheet. Fill the rows from the two snapshots. The action column is the decision; the rationale column documents why so the next reviewer (in 30 days) can audit the call.
# Search themes drift report, [Account name]
# Comparing snapshots: [start_date] vs [end_date]
asset_group | theme | last_month_conv | this_month_conv | drift_% | action | rationale
------------+--------------+-----------------+-----------------+-----------+-------------+----------
Hero AG | running shoes| 142 | 138 | -2.8% | hold | within noise
Hero AG | trail running| 96 | 41 | -57.3% | rollback | drift > 30%, demand flat in Trends
Hero AG | ultralight | 0 | 67 | new | monitor | added by algorithm, watch month 2
Hero AG | minimalist | 88 | 21 | -76.1% | archive | seasonal demand collapsed in Trends
Tail AG | walking shoes| 14 | 12 | -14.2% | density | AG has 9 themes, prune 4 first
# Density check per AG (target: 3-5 themes)
Hero AG : 6 themes -> prune 1
Tail AG : 9 themes -> prune 4
Accessory AG: 3 themes -> holdCross-AG conflict detector
When the same theme appears on two or more asset groups, PMax's bidding splits the same intent across competing surfaces. The fix depends on whether the conflict is intentional.
| Conflict type | Indicator | Resolution |
|---|---|---|
| Identical theme on 2+ AGs | 'running shoes' shows up on Hero AG and Accessory AG, both with conversions | Pick the AG with the higher conversion count and remove the theme from the other; intent should resolve to a single AG. |
| Variant overlap | 'running shoes' on Hero, 'shoes for running' on Tail. Slightly different wording, same intent. | Consolidate to the higher-performing wording. Drop the variant. |
| Geo-specific overlap | 'running shoes uk' on UK AG, 'running shoes' on global AG. Both fire in UK searches. | Either keep the country-specific theme and remove the global variant from UK targeting, or consolidate to the global wording with country-level audience signal. |
| Seasonal carryover | 'christmas gifts running shoes' on Hero AG in January, still converting at 1/3 of December rate. | Move to clearance / off-season AG, lower the priority. Don't kill outright if Trends shows residual demand. |
Things to verify before changing themes
- Is the theme drifting because of seasonality? Cross-check Google Trends for your geos.
- Is the underlying landing page still alive and converting at its historical rate? Theme drift can be a downstream symptom.
- Has the catalog changed? A removed SKU collection breaks the matched intent.
- Did your competitors change pricing? Ad-rank shifts can deprioritise an otherwise-winning theme.
- Is the AG itself in a learning phase? Smart Bidding can deprioritise themes during learning, then restore them. Don't roll back during week 1-3 of a learning phase.
- Have you logged the rollback in your account change history? Future-you needs to read what was done and why.
- Is the rollback target actually a previous-known-good theme, or a guess? Use the prior month's snapshot, not memory.
- Have you written down the success metric for the rollback? Without it, you can't measure whether the rollback worked.
What good looks like after the monthly diff
Every asset group has 3-5 active themes, no duplicates across AGs, every drifting theme has a documented action (hold, rollback, archive, or density-prune-first). The change log carries the rationale so the next reviewer (your future self in 30 days) can audit the decisions. The campaign is operating on intentional theme structure, not on whatever themes accumulated through quarterly Google rep recommendations.
External resources
Authoritative references we link to alongside the template. Read them before running the audit.
- Google Ads, search themes in Performance MaxAuthoritative spec on how search themes feed PMax.
- Google Ads, Performance Max insightsWhere the search themes export lives. Pull from here for the monthly snapshot.
- Google TrendsCross-check seasonal vs structural drift in step 4.
- Search Engine Land, Performance Max coverageTopic index. Tracks search-themes platform changes through the year.
- PPC Town Hall (industry roundtable)Operator-driven roundtable. Useful for hearing how others manage PMax theme drift in real time.
Want this run for you?