The situation
We took over the account in June 2021, in the middle of the pandemic eCommerce surge. That half-year ran standard Shopping on about $13K of spend at a 2.97% conversion rate — a return no accessories brand has seen since, and not a baseline anyone should plan against. When Performance Max launched in 2022 we moved the account onto it, and the real question started: could spend grow every year without the return going the other way?
The measurement had to be honest first. The primary conversion is a GA4 purchase import — real transactions, not a modeled proxy — under data-driven attribution that shares credit with organic touchpoints. That makes every ROAS figure below more conservative than the same number measured last-click.
What we did
- Made revenue the only conversion that counts. The account also tracks add-to-cart, begin-checkout, and view-item actions, and together they report $4.2M in “conversion value.” None of that is revenue. Only the GA4 purchase import sits in the account goals, so bidding optimizes toward orders and the numbers we report are the ones the client sees in their own analytics.
- Scaled spend in steps, not jumps. Annual Performance Max spend went $17K → $74K → $83K → $99K. Each step held or improved ROAS before the next one. 2022 to 2023 was the one flat year, 3.67× both times, and it’s in the table.
- Cut reach to find better buyers. In 2026 impressions fell 31.8% year over year while revenue rose 13.5%. Tightening the audience is the mechanism: average order value climbed from $209 to $244 and conversion rate from 1.11% to 1.29% since launch. The campaign isn’t buying cheaper clicks; it’s finding people who buy more.
- Reported one source per claim. Google Ads and the client’s GA4 disagree by about 8% for the same window, as they should — different attribution models, different windows. We never mix them in one sentence, and every figure carries its pull date.
The results
Google Ads figures are Performance Max only, Jan 1, 2022–Sep 4, 2026, pulled Sep 5, 2026; 2026 is a partial year and is never annualized. Channel share is from the client’s GA4, Jan 1–Sep 3. Total site revenue was flat year over year (+1.1%): this is a story about efficiency and share of revenue, not about growing the company.
Why it worked
Efficiency usually degrades as a campaign scales: the cheap conversions come first and every added dollar buys a worse one. This account went the other way for five years because bidding only ever optimized toward real orders, spend moved up in steps the return had already earned, and reach was cut whenever it bought impressions rather than buyers. What we carry to every eCommerce account: a return that improves at 5.7× the spend is what happens when the conversion you optimize for is the one the client banks.