Skip to content
whitead

Fix: Consolidate fragmented Performance Max asset groups

finding google ads updated 2026.05.28 9 min read
// contents · 10 sections
  1. Why this matters
  2. How to verify the issue
  3. How to fix it
  4. When to split instead of consolidate
  5. How to confirm the fix worked
  6. Diagnostic checklist — run all five 14-21 days post-consolidation
  7. Edge cases — when not to apply this fix
  8. Industry benchmarks
  9. Related rules + concepts
  10. Sources

A Performance Max campaign with eight or twelve thinly-spread asset groups is almost always burning learning signal it can't afford to lose. Each asset group pools its own conversions to inform the matching model, so once you drop below roughly thirty conversions per asset group per month, Google's algorithm is making decisions on noise — and the fix is usually to merge, not to add more creative.

Why this matters

Performance Max bids and matches at the asset-group level, not the campaign level. That means every asset group runs its own internal learning loop on the conversions it accrues — and when one asset group bleeds spend without ever clearing the ~30 conv/30d threshold Smart Bidding needs to stabilise, it never exits cold-start. Worse, the conversions it does collect are siloed: they don't pool back up to help the rest of the campaign learn faster.

Google's own Performance Max best-practices guidance is explicit: "For the best performance, we recommend consolidating your Performance Max campaigns where possible" [1]. Practitioner consensus from agencies running PMax at scale extends that principle one level down — Optmyzr's PPC Town Hall guidance is to "spread spend across fewer, more meaningful segments rather than fragmenting across hundreds of micro-segments," explicitly calling out that consolidation "concentrates budget volume so algorithms receive sufficient data to optimize" [2]. The SKAG-era instinct to split by every audience or keyword variation actively hurts PMax because the algorithm is designed to learn cross-signal — fragmenting the input starves it.

The Whitead rule fires MEDIUM on any ENABLED PMax campaign where at least one asset group either (a) recorded fewer than 30 conversions in the trailing 30-day window, or (b) shares its dominant theme or audience-signal composition with another asset group in the same campaign. Neither condition alone is catastrophic — but both signal that the campaign is spending creative-management effort and learning budget on units of analysis the algorithm cannot meaningfully separate.

How to verify the issue

  1. Open Campaigns → select the Performance Max campaign → Asset groups. Read the row count. Anything beyond five ENABLED asset groups deserves the rest of this check; campaigns at three or fewer are almost certainly fine on this axis.
  2. For each ENABLED asset group, switch the table to show Conversions with the date range set to Last 30 days. Flag every asset group sitting below 30 conversions — those are consolidation candidates by the volume gate.
  3. For each asset group also open View details → Search themes + Audience signal. Note the dominant theme bucket (e.g. "running shoes" vs "trail running shoes") and the audience-signal stack. If two asset groups share the same theme bucket AND overlapping first-party signals (same Customer Match list, same Website visitors segment), they're the same unit to the algorithm — consolidation candidates by the duplication gate.
  4. Cross-reference the campaign's total conversion volume. A PMax doing 500 conv/30d split across ten asset groups means the average asset group sees 50 — borderline. A PMax doing 80 conv/30d across six asset groups (average 13) is structurally starved and consolidation is mandatory.

Steps 2 and 3 together identify which specific asset groups to merge. If step 1 returns three to five ENABLED asset groups and step 2 finds none under 30 conv/30d, the rule's primary trigger is absent — leave the structure alone.

How to fix it

Plan a single 45-minute session per campaign. Resist the urge to re-shoot creative — most of the work is in the asset-group inventory decisions, not the assets themselves.

  1. List your asset groups with their 30-day conversions, dominant theme, and signal stack. A simple spreadsheet beats memory here. Sort descending by conversions — the bottom of the list is where the merge candidates sit.
  2. Group the merge candidates by theme + signal overlap. Two asset groups that both target "running shoes" with the same Customer Match list should merge into one. Two asset groups with truly distinct themes (e.g. "running shoes" vs "hiking boots") but each running 15 conv/30d may merge by signal into a broader "outdoor footwear" asset group, or — if margin profiles diverge — escalate to the question of whether they belong in the same campaign at all (see the over-segmentation finding).
  3. Decide the survivor asset group. Keep the one with the strongest creative inventory (most ENABLED images, ENABLED video assets, healthy Ad Strength), the freshest audience signals, and the highest historical conversion rate. Pause the donor asset groups rather than deleting them — pausing preserves the historical learning record and lets you reverse the change if performance degrades.
  4. Migrate the best assets from donor → survivor. Pull any unique images, headlines, or videos from the paused asset groups that aren't already in the survivor. Don't import duplicates and don't import weak performers — Ad Strength on the survivor is what matters going forward.
  5. Refresh audience signals on the survivor. With more conversion volume now flowing into one asset group, the signal stack carries more weight. Verify a Customer Match list + Website visitors segment + Custom Segment are all attached (see Fix: PMax audience signals weak or missing).
  6. Do nothing else for 14 days. Smart Bidding needs a full learning window to re-stabilise on the consolidated structure. Resist the temptation to tweak budget, targets, or assets during this window — that's a separate failure mode covered in Fix: Bidding change made during learning phase.

When to split instead of consolidate

Splitting an asset group only makes sense when one of two specific conditions holds:

  • Distinct margin tier. A product line with materially different unit economics (e.g. 60% margin SaaS Pro tier vs 25% margin Starter tier) belongs in a separate asset group if you intend to set divergent tCPA / tROAS expectations — and at scale, usually in a separate campaign entirely.
  • Distinct creative requirements. A campaign that needs vertical 9:16 video for one product line and product-feed-only Shopping inventory for another is genuinely producing two different creative briefs. That's a creative-management reason, not a learning-signal reason — and it only justifies the split if both halves clear the 30 conv/30d gate on their own.

"Different keyword phrases customers use" and "different audience segments" are signals, not separate units. Encode them in Search themes and Audience signal inside one asset group — that's exactly what those surfaces exist for.

How to confirm the fix worked

Diagnostic checklist — run all five 14-21 days post-consolidation

  • Campaign now has 3-5 ENABLED asset groups (or fewer, if the campaign legitimately covers a tight theme).
  • Every ENABLED asset group cleared ≥30 conversions in the past 30 days or is on credible pace to clear it by day 30.
  • Each asset group has a distinct dominant theme OR a distinct signal composition — no two ENABLED asset groups duplicate each other on both axes.
  • Cost-per-conversion at the campaign level held within ±15% of the pre-change baseline, with conversion volume stable or up.
  • The survivor asset group's Ad Strength is "Good" or "Excellent" — consolidating creative onto one group should improve, not regress, asset density.

If all five pass, re-run the audit — the pmax_asset_group_consolidation rule moves from failed → passed.

Edge cases — when not to apply this fix

  • Feed-only PMax (Shopping inventory only, no asset groups). Filtered out upstream by the asset-group presence check — this finding cannot fire on a pure feed-driven PMax.
  • Brand vs non-brand PMax in the same account. If you've correctly separated branded PMax (with brand exclusions on the non-brand side), each PMax may legitimately run only one or two asset groups — that's the campaign doing the separation work, not the asset groups, and is healthy.
  • Brand-new campaign (<14 days live). Conversion-volume signals aren't reliable yet. Wait until the campaign has at least one full learning window before applying the 30 conv/30d gate.
  • Highly seasonal account (e.g. tax software, holiday gifting). The 30 conv/30d threshold should be evaluated against in-season volume, not annual average. Off-peak, every asset group will look underpowered — that's a budget question, not a consolidation question.

Industry benchmarks

  • 3-5 asset groups per PMax campaign is the practitioner consensus for healthy signal pooling. Google does not publish a specific number, but the platform cap of 100 asset groups per campaign is a ceiling, not a target [1].
  • 30 conversions per asset group per 30 days mirrors the Smart Bidding tCPA learning-phase threshold — see Smart Bidding learning phase for the underlying volume math.
  • Google's own consolidation framing applies one level up too: "we recommend consolidating your Performance Max campaigns where possible" — the same logic propagates down to asset groups within a campaign [1].

Sources

[1] Google Ads Help — About Performance Max campaigns. https://support.google.com/google-ads/answer/10724817 (accessed 2026-05-27)
[2] Optmyzr PPC Town Hall — Performance Max Optimization: Structure, Signals & Intent Done Right. https://www.optmyzr.com/ppctownhall/pmax-campaign-optimization/ (accessed 2026-05-27)

// was this useful?
// anonymous · no personal data stored
← Back to Wiki