I've audited a few hundred ecom accounts in the last three years. The single most common pattern in accounts stalled between $30K and $100K monthly spend is the same: branded search is baked into Performance Max, and the dashboard ROAS reads way higher than blended (Shopify) ROAS because the platform is taking credit for conversions it didn't drive.
of dashboard ROAS often comes from branded queries that would have converted anyway
Pattern observed across 100+ Ad-Lab audits
The audit takes 30 minutes. The fix takes 14 days to land. The unlock is honest cold-funnel scaling because Smart Bidding finally sees the real conversion pattern instead of the brand-search overlap.
Why this happens
Performance Max sees a buyer search your brand name and click your ad. It records a conversion. The buyer was already going to convert; they searched the brand name. PMax gets credit for the cold-funnel work it didn't do because Google's attribution model can't distinguish between branded and non-branded queries inside PMax the way Search campaigns can.
The result: dashboard ROAS shows 8x. Blended ROAS shows 3.5x. The gap is brand bleed. The bidder learns the wrong lesson and over-allocates to whatever asset group looks profitable in the inflated ROAS, which usually isn't the asset group actually driving cold-funnel demand.
What this costs you
Brands stuck in this pattern routinely show 25-40% of "PMax-driven" revenue is actually branded search overlap. Strip it out and you can see exactly which asset groups produce real cold-funnel demand. Without stripping it out, you're scaling the wrong groups and starving the right ones.
The 30-minute audit
Step 1: pull the PMax search-terms report
Open your Performance Max campaign in Google Ads. Insights tab. Search categories report. Export the last 30 days. You'll get a list of search categories along with conversion value attributed to each.
Step 2: filter to brand-name queries
Add a filter for any term containing your brand name (and common misspellings). Sum the conversion value for those rows. That number is your brand bleed.
Step 3: calculate the bleed percentage
Divide brand-bleed conversion value by total PMax conversion value. If it's above 15%, you have a problem. Above 25%, the dashboard ROAS is meaningfully lying. Above 40%, the bidder is making decisions on noise.
Step 4: cross-check with Shopify
Pull the same 30 days of PMax-attributed revenue from Google Ads. Compare to Shopify's reported paid-traffic revenue. The gap divided by Google's reported number is your second confirmation of the bleed.
Step 5: write the rebuild plan
If you have meaningful bleed, draft the 14-day rebuild: dedicated branded Search campaign, brand terms in the PMax negative-keyword list, dashboard-ROAS expectation reset for stakeholders. The fix is straightforward; the buy-in is the part that takes work.
What the fix looks like
Two campaigns instead of one. A dedicated Search campaign with exact-match branded keywords on a tight POAS target (typically 8-15x ROAS depending on AOV; the calculator below converts that into a target ROAS for your specific margin). The existing PMax with brand terms added to the account-level negative-keyword list.
Before and after the rebuild
Before
Single PMax campaign. Dashboard ROAS 8x. Blended ROAS 3.5x. Decisions made on the inflated dashboard number.
- Brand search baked into PMax
- Dashboard ROAS overstates by 30-50%
- Asset-group performance comparisons are noise
- Smart Bidding learns from biased signal
After (14 days post-rebuild)
Two campaigns. Dashboard ROAS drops to 4.2x because the inflated number is gone. Blended holds steady at 3.5x. Decisions now defensible.
- Branded Search campaign on tight POAS target
- PMax handles cold funnel + non-branded
- Asset group performance reads honestly
- Smart Bidding learns from clean cold-funnel signal
What the dashboard does in week 1 post-fix
Dashboard ROAS will drop 25-40% in the first 7 days. Blended (Shopify) ROAS will hold steady. Stakeholders that don't know the playbook panic at the dashboard drop. Have the explanation ready: dashboard isn't getting worse, it's getting honest. Cold-funnel scale is now possible.
The numbers that defend the change
When the CFO asks why dashboard ROAS dropped, the chart you bring is dashboard ROAS, blended ROAS, and MER on one axis over the last 60 days. The dashboard line drops; the other two lines hold. That's the visual proof attribution got honest, not that performance got worse.
Run the audit + fix with these
When this isn't the fix
Brand bleed isn't always the issue. If your blended ROAS is already inside 15% of dashboard ROAS, your PMax structure is mostly clean and the next leak is somewhere else (catch-all search themes, weak audience signals, no Search backstop). Run the audit anyway because confirming clean structure is itself a win, but don't expect a structural unlock from this fix specifically.