Why last-click ROAS misleads almost everyone

Brand search converts at several times the rate of everything else and costs a fraction as much. Any account that blends brand into its headline ROAS is reporting a number that mostly measures how many people already knew about it. Scale the budget and that number falls, and it looks like the new spend failed when in fact the old spend was never doing the work.

So we separate brand from non-brand in the first week, and report them apart permanently. It is an uncomfortable conversation once and a clear account afterwards.

The feed is the campaign

In Shopping and Performance Max you do not choose keywords – the product feed decides which queries you are eligible for. A title written for a product page is rarely a title written for search, and the gap between the two is the single largest source of wasted Shopping spend we find.

Practically: titles that lead with the terms buyers use, attributes complete enough to qualify for the queries you want, and custom labels carrying margin band, stock level and seasonality so bidding can favour what actually makes money rather than what merely sells.

Performance Max without losing the account

PMax works. It also, left alone, absorbs your brand traffic and reports it as incremental. The structure we use keeps brand in its own Search campaign so PMax cannot claim it, splits asset groups by margin band rather than by whatever the catalogue tree happens to be, and uses exclusions deliberately rather than as an afterthought.

We also insist on the reporting that PMax does not volunteer – search term and placement visibility, product-level performance – because "trust the algorithm" is not a strategy when the algorithm is optimising towards a conversion value you set carelessly.

Signal quality is now the differentiator

Consent banners, iOS restrictions and cookie decay have quietly removed a chunk of the conversions these platforms use to learn. Server-side events through the Conversions API and Enhanced Conversions put a meaningful share of that back, and an account with good signal simply outperforms an identically managed account without it. This is unglamorous plumbing and it is where we start.

Questions people actually ask

What ROAS should we target?
The one your margin supports, which is a different number for every catalogue. A 3x ROAS is excellent at 70% gross margin and a slow loss at 25%. We work back from contribution margin and set the target from there rather than adopting a figure from a case study.
Is Performance Max unavoidable now?
For most retailers, effectively yes – it holds inventory that Shopping and standard Search no longer reach on their own. What is avoidable is running it as a single undifferentiated campaign that quietly eats your brand traffic and reports it as new revenue.
How do you handle the product feed?
As a ranking asset, not a data dump. Titles carry the terms people actually search, attributes are complete enough to qualify for the queries you want, and custom labels encode margin and stock so bidding can act on profitability rather than on revenue.
Do you run Meta as well as Google?
Usually both. They do different jobs – Google harvests demand that already exists, Meta creates it – and the interesting question is the split between them, which changes with season, catalogue and creative supply.
How do you deal with returns?
By getting them into the conversion value where the platform allows it. A category with a 40% return rate is not as profitable as the dashboard says, and bidding that ignores this systematically overspends on exactly the wrong products.
What about first-party data?
Enhanced Conversions and the Conversions API, plus customer lists segmented by actual value rather than by "all purchasers". Signal quality has become the difference between accounts that scale and accounts that stall, and it is unglamorous work almost nobody has finished.

Free

Get the ecommerce account audit

What are you after

One of these. Whichever you pick, no call is required to get an answer.

What should we look at

Pick one or more. Each audit is done by hand, not by a tool.

We cannot audit anything without it.

Genuinely useful to us – half our traffic is impossible to attribute otherwise.

Free · one reply within a working day · no call needed

Start with the audit.

We go through your accounts, tracking and funnel, then send back what we would change and in what order. About five working days, no charge, no call required to get it.