Performance Max reports 9x ROAS but new customer share is falling
An online pet-supply retailer spends $120k/month on Performance Max and reports 9.0 ROAS. New customer share is 22%, down from 35% last year. The account includes brand keywords in the PMax asset group because brand search is profitable, and returning customers drive 60% of revenue. The team uses 'Maximize conversion value' with a target ROAS of 700%. First-party data shows 12-month LTV of new customers is $180 vs $95 for repeat. Leadership wants to cut PMax because new customer acquisition is weak, but PMax is the largest revenue driver. GA4 shows branded search converts at 12x ROAS and non-brand at 4x. The team has no holdout and relies on platform-reported conversions. Budget is fixed for the quarter. The CFO asks for a recommendation.
1. Which metrics should the team use to judge PMax performance beyond blended ROAS?
Read the worked answer
Blended ROAS of 9.0 mixes brand, returning customers, and new customers, so it cannot answer whether PMax is adding demand. Use new customer acquisition cost, new-customer revenue, and new-customer ROAS. Separate brand from non-brand: branded search at 12x and non-brand at 4x means the blended 9x hides weak non-brand efficiency. Compute marginal ROAS by budget segment; if brand demand is capped, extra spend may go to non-brand at a lower return. Compare 12-month LTV: new customers are worth $180 vs $95 for repeat, so a new customer is worth about 1.9x a repeat customer. A lower immediate ROAS on new customers can therefore be acceptable. Use GA4 path analysis and Google Ads new customer acquisition reporting, but do not ignore repeat revenue because it may subsidize acquisition. The core question is incremental new-customer value, not platform ROAS.
2. Recommend changes to campaign structure and bidding to improve new customer acquisition without losing profitable repeat revenue.
Read the worked answer
Use Performance Max with a new customer acquisition goal and set a separate value or target for new customers. Exclude existing customers from a new-customer-only asset group where eligible, using customer match exclusions carefully. Move brand search into a separate Search campaign with exact-match brand terms and its own budget so PMax cannot claim brand demand. Feed first-party LTV segments as audience signals for high-LTV new customers. Set a tROAS that reflects blended LTV but monitor new-customer CPA separately. Test one change at a time: first move brand out, then add the new customer acquisition goal, then adjust tROAS. Do not cut PMax entirely; instead cap brand and measure incrementality. Keep a repeat-customer path for retention, but do not let it dominate the acquisition signal.
3. How would you measure whether PMax is incrementally acquiring new customers?
Read the worked answer
Run a geo holdout or a PMax experiment with a control region, matched on pre-period revenue and new-customer share. Measure incremental new customers and revenue, not platform-attributed conversions. Use a 4-6 week test and compare test vs control with difference-in-differences. If the holdout shows incremental new-customer ROAS above the LTV-based threshold, keep and scale; if not, restructure around brand separation and new-customer-only goals. Also monitor brand search volume and direct traffic as guardrails so the test does not mistake demand capture for demand creation. Document confidence intervals and avoid calling a small lift a win.