POAS vs ROAS: Why Profit-Based Bidding Changes eCommerce Advertising

Why ROAS lies — same ROAS, different profit

I've seen this play out dozens of times: a client looks at the dashboard, sees a 4× ROAS across all three of their product categories, and concludes they're equally healthy. In reality, one category is generating solid profit, another is barely breathing, and a third — at an identical 4× ROAS — is quietly losing money on every sale. ROAS can't tell these three scenarios apart because it knows nothing about the cost of goods sold.

The problem isn't that ROAS is a "bad" metric — it does exactly what it's built to do, measuring the ratio of revenue to spend. The problem is that people use ROAS as a proxy for profitability, when profitability depends on a third number ROAS never accounts for: cost of goods sold (COGS). Once that number varies by product — and in practice it almost always does, even in seemingly simple catalogs — ROAS stops being a reliable guide for budget decisions.

This shows up most clearly on eCommerce accounts with a broad assortment. A multi-category store selling both high-margin accessories and low-margin "anchor" products (branded electronics at a 5–10% margin, stocked purely to drive traffic) will almost always turn up a finding like the table below — somewhere in the portfolio sits a product with a good ROAS that is quietly eating profit out of the rest of the account. Without a POAS view, that product keeps getting funded because it looks like a "winner" on every standard report.

Here's an illustrative example with three products that share an identical 4× ROAS but wildly different margins — the numbers are deliberately round and simplified for clarity, not pulled from any real client account:

This is why a "great ROAS" isn't always good news. Product C would look like the best campaign in the account on paper — same ROAS, maybe even higher sales volume because of its lower price point — when it's actually the only one actively draining profit from the rest of the account.

Break-even math: ROAS and POAS

This is where most people trip up, so let's go step by step. There are two formulas and they get mixed up easily:

Break-even ROAS depends on margin — the lower the margin, the higher the ROAS you need just to cover spend. Here's a table that shows it directly:

POAS break-even is always 1× — which is why POAS is mentally easier to work with. You don't have to memorize a table for every margin; you just check whether the number is above or below 1. If POAS is 1.5×, for every € spent on ads you got back €1 of spend plus €0.50 of pure profit. If POAS is 0.8×, you're losing 20 cents on every € spent, no matter how "good" the ROAS dashboard looks.

The same logic works in reverse — you can derive the ROAS you'd need for a target profit from a target POAS. Say you want a POAS of 1.5× at a 30% margin: required ROAS = 1.5 ÷ 0.30 = 5.0×. That's a useful trick when talking to a client who still thinks in ROAS terms — translate their POAS target back into a familiar ROAS target, but one anchored to their actual catalog margin, not some industry average from a report.

How to implement POAS in practice

POAS isn't a native metric in the Google Ads interface — Google doesn't know your margin and has no reason to assume one. That means implementation requires you to inject a profit signal where Google would otherwise only see revenue. This isn't a trivial technical detail — each of the approaches below carries a different level of risk, development effort, and precision, so the choice depends on how quickly you need visibility versus how deeply you want the algorithm to actually change its behavior. Here's the order I work through on accounts, from simplest to most sophisticated:

The order I recommend: start with the custom column (zero risk, fast), confirm the numbers match your bookkeeping for a month or two, and only then change conversion value so the algorithm actually starts optimizing toward profit. Skipping that verification step and pushing profit straight into conversion value is a common source of errors that are hard to spot later.

When POAS isn't needed

POAS isn't a universal fix and shouldn't be forced everywhere. If you have uniform margins — every product in the account carries roughly the same margin (say, 30% ± 3%) — then a ROAS target already implicitly encodes profitability. In that case, the extra POAS infrastructure is overhead without real benefit: development time, tracking risk, and the ongoing maintenance of margin data cost more than they return, since the resulting bidding decisions would be practically identical to what you'd get from a well-calibrated ROAS target based on that same average margin.

A practical rule I use: if the margin gap between the most and least profitable product category is under 10 percentage points, POAS rarely changes budget decisions compared to a well-calibrated ROAS target. Above that, POAS starts to genuinely change recommendations — and it's worth investing in the implementation.

There's also a middle option I often recommend to small and mid-size accounts unsure whether they need the full build: run a one-off custom column analysis (done manually, outside Google Ads, in a spreadsheet) at the product-category level. If that one-time analysis shows margin differences shift the profitability ranking of products relative to their ROAS ranking, that's the signal it's worth investing in a lasting POAS setup. If the rankings match, ROAS is a good enough proxy and you can save the development budget for something else.

What changes in bid strategy when the algorithm sees profit

This is the part that surprises clients the most: when the tROAS algorithm optimizes toward a conversion value that's actually profit (not revenue), bidding behavior changes fundamentally — not just the numbers in the report. This isn't a cosmetic change to the metric you look at on a dashboard; the actual function the algorithm tries to maximize changes at the level of every single auction, thousands of times a day.

In practice, that means two auctions that would previously have looked identically valuable under a Maximize Conversion Value goal (say, both worth €100 in conversion value) can now get entirely different bids — if one comes from a product with a 50% margin and the other from a product with a 10% margin, the algorithm will bid more aggressively for the first. That's a fundamental difference from a standard tROAS setup, where both auctions would be treated exactly the same since their conversion value (revenue) is identical.

• The algorithm starts favoring high-margin products, even if they generate lower revenue per transaction. A product bringing in €50 revenue at 60% margin (€30 profit) becomes "worth more" to the algorithm than one bringing in €80 revenue at 15% margin (€12 profit) — even though the second product would look better under a pure ROAS/Maximize Conversion Value goal.
• Budget naturally reallocates toward ad groups and campaigns tied to more profitable product lines, with no manual intervention — the algorithm does this automatically because it's optimizing the signal you gave it.
• The target needs recalibrating — if you previously had a tROAS target of 400% based on revenue, once you switch to profit-based conversion value, that same numeric target no longer means the same thing. It should be set based on the POAS break-even (1×) plus a margin of safety — e.g. a target POAS of 1.3–1.5× for a healthy buffer above break-even.
• Volatility rises in the short term — profit per order has more variance than revenue (margin fluctuates with discounts, promos, and product mix), so the algorithm needs a somewhat longer learning period to stabilize its predictions.
• Client reporting gets more direct — instead of "ROAS is 4×, great," the conversation becomes "POAS is 1.6×, meaning every € invested returned 60 cents of clean profit after all costs." That's a conversation clients with a finance background grasp and appreciate immediately.

What this means in practice: when you switch to profit-based bidding, the first report after the change often looks "worse" by the old standard — total revenue can dip slightly because the algorithm deliberately pulls back spend from low-margin products it used to push aggressively. That's not a regression, it's exactly what you asked for — less revenue from poorly profitable sales, more budget where every € of ad spend generates real profit. Across the accounts I run, this transition typically means a mild dip in total revenue in the first month, followed by growth in total profit over the next 2–3 months as budget fully redistributes. That's why an average POAS of 1.78× is a more valuable data point than any single ROAS number — it tells you what the client actually earns, not how much turnover the account generates.

FAQ

Related guides