Performance agency for e-commerce: how it really differs from a classic PPC agency

A PPC agency manages campaigns, a performance agency manages profit

A classic PPC agency gets a budget and a task: set up campaigns in Google Ads or Meta Ads to generate as many conversions as possible at a given CPA. A performance agency for e-commerce works with different inputs, it layers conversion data with margin, returns and cost price per product. The difference isn't which platform is used, it's what gets fed into that platform.

Google, Meta and TikTok all now offer value-based bidding, an automated strategy that doesn't optimise for the number of conversions but for the value a business assigns to them. According to [Google Ads Help](https://support.google.com/google-ads/answer/15099424?hl=en), this is a subset of Smart Bidding that maximises conversion value within a budget or target ROAS. The question most e-commerce businesses never ask is a simple one: what value is being sent to the system, order revenue, or profit after costs?

This is exactly where PPC work and performance work part ways. Classic campaign management leaves the conversion value field equal to revenue. A performance agency recalculates it as gross profit, meaning revenue minus cost price, shipping and expected returns, and it's this figure that gets sent to the algorithm.

What the data says about ROAS and profit

Google's own material on value-based bidding states that businesses moving from target CPA to target ROAS saw a median 14% increase in conversion value, even though return on ad spend stayed roughly the same, according to [Google Business: Increase your ROI with Value-based Bidding](https://business.google.com/uk/resources/articles/increase-your-roi-with-value-based-bidding/). That's proof the algorithm can do more when given a more precise signal about value, not just a conversion count.

Search Engine Land also flags a weakness in ROAS calculated from revenue alone. Across a catalogue with varying margins, optimising for revenue value can push the algorithm to favour high-revenue but low-profit products, or as the author puts it, "an overemphasis on high-revenue but low-margin products", according to [Search Engine Land: Target ROAS in Google Ads](https://searchengineland.com/target-roas-key-considerations-431770). We've seen exactly this in accounts we've taken over.

The answer that's getting increasing attention is POAS, profit on ad spend. Instead of revenue, gross margin after costs gets fed into bidding. According to [Search Engine Land: Value-based bidding](https://searchengineland.com/value-based-bidding-boost-google-ads-437240), this metric ties spend directly to profit, not turnover. It's not a new algorithm, it's a different input into the same algorithm.

A condition most e-commerce businesses overlook: value-based bidding needs at least 15 conversions in 30 days at the conversion goal level, otherwise the signal is too noisy, as stated in [Google Ads Help](https://support.google.com/google-ads/answer/15099424?hl=en) recommendations. Below this threshold, you're measuring noise more precisely, not performance.

Margins across a catalogue commonly vary threefold, ROAS doesn't see it

In catalogues we've reviewed for new clients, gross margin between categories commonly varied threefold. Budget then drifts towards high-revenue, low-margin products, exactly the mechanism Search Engine Land describes above. A business ends up watching ROAS climb and profit fall in the very same month.

With Zlatá Putňa, we shifted the controlling metric from revenue to profit per order and reworked the feed so the algorithm understood margin at product level, not category level. The result was a 45.8% increase in revenue and a 574% increase in visibility, without budget rising proportionally. The full case is available on the [Zlatá Putňa reference page](/referencie/zlata-putna).

Domintell was a different kind of problem, the e-commerce site had weak traffic, not poor margins. A combination of a corrected feed, technical SEO and performance campaigns brought a 272% increase in visitors. The figure shows that performance work for e-commerce isn't only bidding, it's also the state of the data bidding is built on.

Where experts disagree

The Ehrenberg-Bass Institute has long argued that shifting budgets into short-term performance activation at the expense of brand building damages campaign effectiveness over the longer term, as described in [Ehrenberg-Bass Institute on short-termism in marketing](https://marketingscience.info/marketers-obsession-with-roi-and-short-termism-undermining-growth/). Part of the industry disagrees, calling some of the institute's conclusions too rigid a rule to apply equally across every sector.

For e-commerce, this debate is a somewhat different question than it is for branded CPG marketing. An e-commerce business needs cash flow every month, not just mental availability a year from now. That's why performance work in e-commerce tends to concentrate on the lower and middle funnel, where profit can be measured directly. That doesn't mean the upper funnel doesn't exist, it's just harder to attribute to one specific order.

Our position

The consensus in expert discussion holds that the decision mainly comes down to choosing a bidding model, target ROAS versus POAS versus maximising conversion value. Based on our own data, we argue that model choice is secondary. The difference between models is smaller than the difference in quality of the input data a business feeds into them. The accounts we've taken over had too many manual exclusions and too little order in their feed and conversion data, not a bad model.

Google, Meta and TikTok buy media brilliantly. But they know nothing about your cost price, margin and return rate unless you tell them. This is exactly the gap our [AI engine within performance services](/sluzby) closes, working as a transparent layer over the platforms, not a replacement for them.

The +300% ROAS figure we cite in our references is an upper bound for a catalogue with high-quality margin data, not an average. Anyone presenting it as an average is misleading you. The controlling metric for an e-commerce business should be profit per order after returns and shipping, not revenue, for exactly the reason Search Engine Land describes above.

Practical steps if you're considering a performance agency for e-commerce

The first step isn't choosing an agency, it's a data audit. Check whether your feed carries price, margin and stock status at product level, not category level. Without that, no agency, not even one with the best model, has anything to optimise.

The second step is volume. If a campaign generates fewer than 15 conversions in 30 days, value-based bidding has nothing to learn from, according to [Google Ads Help](https://support.google.com/google-ads/answer/15099424?hl=en) recommendations. In that case, it's wiser to first improve the product, feed and creative.

The third step is the agency's payment model. A percentage of budget rewards spending more, not earning more. That's why for e-commerce, a model tied to results, not spend levels, makes more sense. Say the price out loud at the first meeting, a [free consultation](/rezervacia) is a good place for that.

The fourth step, if you're unsure whether advertising is genuinely working or you're just seeing a seasonal effect, is a geo test on part of the market. It costs time and part of your territory, but it's a fair price for certainty.

When this doesn't apply

If a catalogue has uniform margins across all products, target ROAS from revenue naturally tracks close to profit, and recalculating for profit adds little on top of more complex data management.

If an account generates fewer than roughly 15 conversions per 30 days per campaign, complex optimisation is pointless. You're measuring noise more precisely. In that case, it's better to invest in the product, feed and creative than in the bidding model.

If a business hasn't cleaned up its costs for product, shipping and returns, moving to full POAS will fail on the input data, not the algorithm. That work needs doing before the strategy change, not during it.

And if the website or the product itself has nothing worth selling, optimising site speed or bidding is a waste of time. Fix the offer first, then the performance.

Frequently asked questions