Performance Max (PMAX) campaigns are best used as a learning tool, not a permanent campaign structure. The core principle: run PMAX for approximately 90 days to gather broad performance data across Google's inventory, then migrate to more targeted Search and/or Shopping campaigns using what was learned.
This approach avoids the long-term inefficiency of PMAX — which "spends a bunch of money trying things" — while still benefiting from its cross-channel discovery capability early in a product's advertising lifecycle.
"You use PMAX to figure out what works, and then you convert it to other types of campaigns."
— Mark Hope, 2026-03-25 weekly call
Maintain separate PMAX (or Search/Shopping) campaigns for your highest-volume products. In the Doodle Farms account, these are:
- Black Beans
- Popcorn (Doodla and Old World)
- Yellow Cornmeal
Each has distinct CPA targets and budget needs. Bundling them obscures performance signals.
Any products not covered by a dedicated campaign should be captured in a low-budget catch-all Shopping campaign. This ensures full catalog coverage without diluting budget from priority products.
"You don't want to let stuff slip through. It's kind of like the old gold banner."
— Mark Hope, 2026-03-25 weekly call
Low optimization scores on PMAX campaigns are frequently caused by missing assets. Required assets for a healthy PMAX campaign:
| Asset Type | Notes |
|---|---|
| Videos | Must include correct aspect ratios (landscape 16:9 and vertical 9:16). Build in Canva or use Google's in-platform video builder as a starting point. |
| Images | Lifestyle images preferred; ensure correct dimensions for each placement type |
| Headlines & Descriptions | Multiple variations required |
| Logo | Must meet dimension requirements (2:1 ratio recommended for horizontal logo slot) |
| Site Links | Set at account level to cover all campaigns efficiently |
| Call-outs & Snippets | Free real estate — always add these; missing them wastes ad space |
See also: [2]
| Campaign Type | Recommended Bid Strategy |
|---|---|
| Brand Search | Target Impression Share (95% absolute top) — never leave brand searches on the table |
| PMAX (learning phase) | Target CPA set to realistic observed CPA, not aspirational |
| PMAX (mature) | Migrate to Search/Shopping with Maximize Conversions |
| Catch-All Shopping | Maximize Clicks with a low budget cap |
| Old World Search (new) | Start with Maximize Clicks; switch to Maximize Conversions after sufficient conversion data |
Key finding from 2026-03-25 audit: The Doodle Farms Black Beans PMAX had a target CPA of $17 but the actual observed CPA was ~$30. Underbidding relative to reality causes the algorithm to under-serve the campaign. The target was raised to $30 to align with actual performance.
Google throttles impressions for campaigns with an Optimization Score below 80%. During the 2026-03-25 audit, the Doodle Farms account was at 60% — a significant risk to delivery.
Common fixes that raise score without increasing budget:
- Add missing videos to PMAX asset groups
- Add site links, call-outs, and structured snippets (set at account level for efficiency)
- Fix broken conversion tracking (uncounted conversions suppress smart bidding)
- Align target CPA with observed CPA
Target: above 80% as a floor; above 90% as a goal. Getting from 90% to 100% typically requires budget increases, which may not be warranted.
PMAX and smart bidding strategies are only as good as the conversion data they receive. A broken tracking setup will cause the algorithm to optimize toward nothing.
Example from 2026-03-25 audit (Doodle Farms Display Remarketing):
- Campaign was spending ~$300/month with 0 reported conversions
- Audit revealed 574 conversions were occurring but not being counted due to a tracking misconfiguration
- The campaign was rebuilt to target cart abandoners with all-product messaging (replacing a beans-only focus)
Always verify:
- Conversion actions are set as Primary (not Secondary) where appropriate
- Tags are firing correctly on thank-you/confirmation pages
- No duplicate or conflicting conversion actions
See also: [3] for a parallel case of broken conversion tracking causing zero reported conversions on a $33/day spend.