Usually Performance Max is optimizing toward the wrong signal: it counts every sale equally instead of profit, leans on cheap brand and remarketing traffic, and reports inflated conversions. Fix this by feeding it profit data, excluding brand searches, and judging it on real margin, not platform-reported ROAS.
- Performance Max optimizes to the conversion value you feed it — if every order reports the same revenue, it can't tell a high-margin sale from a money-losing one.
- PMax frequently absorbs branded searches and remarketing audiences, taking credit for sales you'd have won anyway and inflating reported ROAS.
- Platform-reported ROAS counts view-through and modeled conversions; real profit only shows up when you compare ad spend to margin in your own books.
- Sending net profit (revenue minus COGS, shipping, fees) as the conversion value instead of revenue retrains PMax toward genuinely profitable orders.
- Google Ads management in Canada typically runs $1,500–$5,000/mo flat or 10–20% of spend, separate from the ad budget most SMBs set at $2,000–$10,000/mo.
It's optimizing toward the wrong number
The most common reason: Performance Max is doing exactly what you told it to do, and you told it the wrong thing. PMax is a value-maximizing machine. It chases whatever you set as the conversion value — and by default that's revenue, not profit.
If a $400 order on a 12% margin and a $400 order on a 60% margin both report "$400" to Google, the algorithm sees them as equally good. It will happily pour budget into your lowest-margin, highest-volume products because they convert easily and hit your target ROAS on paper. Your dashboard looks healthy. Your bank account doesn't.
The same trap catches discount-driven sales. If your conversion value is the pre-discount price, or you don't subtract returns, refunds, shipping, and payment fees, PMax is optimizing toward a fantasy number that never reaches your books.
The fix is to change the signal. Send margin-aware conversion values instead of raw revenue — either net profit per order, or revenue weighted by product-level margin tiers. In practice this means calculating profit at checkout (revenue minus cost of goods, shipping, and fees) and passing that figure as the conversion value through your tag or server-side tracking. Once PMax sees that a $400 low-margin sale is really worth $48 and a $400 high-margin sale is worth $240, it shifts spend toward the products that actually make you money. Pair this with a target ROAS recalculated against profit, not revenue, and the campaign starts pulling in the same direction as your business.
It's taking credit for sales you'd win anyway
Second big reason: Performance Max is eating your branded and remarketing traffic and reporting it as new, ad-driven profit. It usually isn't.
Left unchecked, PMax bids on people searching your exact brand name and re-targets visitors who already added to cart. Those people were going to buy. When PMax intercepts them, it claims a high ROAS that is mostly cannibalized — you paid for a sale you'd have gotten for free through organic, direct, or email. Strip out brand and remarketing, and the "profitable" campaign often turns out to be acquiring almost no genuinely new customers at a profit.
This is why a campaign can show a strong reported ROAS while your overall blended profit flatlines or drops. The platform number goes up; the business number doesn't.
To separate real acquisition from cannibalization, add brand keywords as account-level negatives (or use the brand exclusion list) so PMax can't claim branded searches. Watch your new-customer acquisition metrics and Google's "new customers" reporting, and use the higher-value bidding option for first-time buyers if repeat purchase value justifies it. Then look at blended performance — total ad spend against total profit across all channels — rather than trusting any single campaign's self-reported ROAS. If PMax spend rises but blended new-customer profit stays flat, you've found your leak. The goal isn't to kill PMax; it's to force it to earn incremental sales instead of taking credit for demand you already created.
Reported ROAS and real profit are different numbers
Third reason, and the one that fools the most people: the ROAS in your Google Ads dashboard is not your real ROAS, and definitely not your profit. PMax counts conversions your business never feels.
Google's reported value includes view-through conversions (someone saw a display impression, bought later), modeled conversions (statistical estimates when cookies are missing), and a generous attribution window. None of that is wrong, exactly — but it consistently reports more conversion value than lands in your store. A 5x reported ROAS can be a 2.5x real ROAS once you reconcile against actual orders, and a real ROAS below your break-even point means every sale loses money even while the dashboard celebrates.
Broken or duplicate conversion tracking makes this worse. Double-firing purchase tags, counting the same order twice, or importing revenue with tax and shipping baked in all inflate the number PMax optimizes against — so it scales spend on phantom returns.
To find the truth, reconcile monthly: pull total PMax-attributed revenue from Google and compare it to actual orders in Shopify or your platform for the same period. Calculate your break-even ROAS (1 ÷ profit margin) and judge campaigns against that, not against an arbitrary target. Audit the tag with Tag Assistant to confirm purchases fire once with the correct value. When reported and real numbers diverge sharply, trust your own books. The platform optimizes to what it can measure; profitability lives in the gap between that and your P&L, and closing that gap is where most of the wasted spend gets recovered.
Structure, feed, and budget starve good products
Fourth reason: the campaign structure and product feed quietly route budget to the wrong products. PMax is only as smart as the data and constraints you give it.
A single catch-all PMax campaign with one asset group treats your hero product and your clearance junk as one pool. Without listing groups or separate campaigns, it can't prioritize the margins and conversion rates that matter — so it drifts toward whatever is cheapest to convert. A weak product feed makes this worse: missing or vague titles, no GTINs, thin descriptions, and poor images all suppress your best products in Shopping inventory while the algorithm spends elsewhere.
Budget and ramp also get misread as failure. PMax needs roughly 30 conversions in the learning period to stabilize; judging a campaign in week one, or changing targets every few days, keeps it perpetually relearning and bleeding budget. Equally, an unrealistic target ROAS can choke a campaign into spending on a tiny, easy slice of demand that doesn't scale.
The practical fixes: segment by margin or product category into separate campaigns or asset groups so you can set distinct ROAS targets — aggressive on high-margin lines, protective on thin ones. Clean the feed (accurate titles, GTINs, real benefit-led descriptions, strong primary images) so your profitable products actually surface. Set a target that reflects break-even plus a profit cushion, then leave it alone long enough to learn. And exclude poor-performing or out-of-stock SKUs rather than letting spend leak into them. Structure is leverage: it's how you tell a black-box algorithm which sales are worth chasing.
Related questions
Profit, wherever you can. If you send raw revenue, PMax can't distinguish a high-margin sale from a money-losing one and will chase volume on your cheapest-to-convert, lowest-margin products. Passing net profit (or margin-weighted values) per order retrains it toward genuinely profitable sales. At minimum, subtract returns, shipping, and payment fees from the value you report.
Two usual culprits: PMax is cannibalizing branded and remarketing traffic you'd have converted for free, and the reported ROAS includes view-through and modeled conversions that never reach your books. Exclude brand searches, look at blended profit across all channels, and reconcile Google's attributed revenue against actual orders in your store before trusting any campaign-level number.
Give it the learning period to stabilize — roughly 30 conversions, often two to four weeks depending on volume — and avoid changing targets or budgets every few days, which forces it to relearn. That said, if conversion tracking is broken or you're feeding it revenue instead of profit, fix those first; no amount of patience overcomes a campaign optimizing toward the wrong signal.
Often yes for e-commerce. A single catch-all campaign treats high-margin heroes and thin-margin clearance as one pool and drifts toward whatever converts cheapest. Segmenting by margin or category — into separate campaigns or asset groups with distinct ROAS targets — lets you push hard on profitable lines and protect thin ones, instead of letting the algorithm average everything together.
Want a second opinion on your situation?
Get a free, no-obligation proposal. We’ll look at your site and your market and tell you honestly what we’d do — and what we wouldn’t.
Get Free Proposal →