US ecommerce advertising has three characteristics that do not apply elsewhere at the same intensity: your attribution has been partially broken since 2021 and most people still plan as though it is not, Google has a product that spends your budget without telling you where, and a quarter of your annual revenue arrives in six weeks under auction conditions nothing else in the year prepares you for. Tool choice follows from those three, not from feature lists.
Your reported numbers are modelled, not counted
Since app tracking transparency, a meaningful share of iOS conversions are not observed — they are estimated. The platforms do not present them differently from observed ones, so your dashboard reads as though everything were measured.
Two practical consequences that change how you should run a US store.
Platform totals disagree with your books
Meta and Google will together claim more revenue than your store recorded. Both are counting the same sales under different attribution windows, and some of it is modelled. This is normal and not a bug to chase.
Small campaigns report worse than they perform
Modelling needs volume. A campaign with twenty conversions a month gets a rougher estimate than one with two thousand, which systematically understates smaller tests and makes them look like failures.
💡 The single most useful fix is server-side conversion tracking — Conversions API on Meta, enhanced conversions on Google — sending the order from your server rather than only from the browser. It recovers a chunk of what the browser loses. Ask any vendor whether they support it, because a tool optimising on browser-only data in the US is working from partial information and cannot tell you so.
And the ground-truth measure worth more than either platform: your blended figure. Total ad spend divided by total orders, taken from your store. It attributes nothing to anything, which is exactly why it cannot be gamed.
Performance Max: powerful, and opaque by design
Google pushes it hard at US ecommerce and it genuinely works for some stores. The honest version of the trade-off:
What you gain
One campaign reaching Search, Shopping, YouTube, Display, Discover and Gmail, optimised together against your feed. With enough conversion volume it frequently beats hand-built campaigns.
What you give up
Visibility. You see far less about which surface spent your money and which search terms triggered it. Diagnosing a bad month is substantially harder.
The trap
It will happily spend on brand searches — people already looking for you — and report them as conversions. That inflates reported return while buying customers you already had.
Two things to do if you run it: add brand terms as negatives at the account level where the structure allows, and keep a separate hand-built Search campaign for your highest-intent non-brand terms so you retain one channel you can actually see into.
Below roughly thirty conversions a month it is generally the wrong choice. It has too little signal to optimise on and you have given up visibility for nothing.
Q4 breaks your benchmarks
Worth planning for rather than being surprised by, because it is the period that decides the year for most US stores.
- 1Auction prices rise sharply from late October. Your cost per acquisition target from September is not the right target in November, and holding to it will keep you out of the auction during the only weeks that matter.
- 2Learning phases are expensive in Q4. A campaign launched on Black Friday spends its most expensive days learning. Build and warm campaigns in October so they enter November already optimised.
- 3Creative fatigues faster at high frequency. Have the next set ready rather than generating it under pressure — this is the one period where a creative tool producing twenty variants genuinely pays for itself.
- 4January reverts. Do not read a December cost per acquisition as your new normal, and do not read January as a collapse.
How catalogue size changes the answer
This determines which category of tool fits you more than anything else.
Under ~50 products
A feed adds little. Hand-built campaigns to your best sellers, with a carousel giving each card its own destination, will outperform catalogue automation. Full-stack builders are the right category.
50–1,000 products
The sweet spot for catalogue automation. Madgicx and AdScale earn their fee here, joining spend to revenue. Requires a clean feed and verified order data.
Thousands of SKUs
Feed management becomes its own discipline — titles, attributes, exclusions, custom labels. This is specialist territory, and a general campaign builder is the wrong tool regardless of how good it is.
First-party data is the durable advantage
While platform signal degrades, the data you own does not. A US store with a real email list and purchase history can build customer lists, suppress existing buyers from acquisition campaigns, and seed lookalikes from high-value cohorts rather than from everyone.
Suppression alone is worth doing this week: excluding people who bought in the last thirty days from your prospecting campaigns stops you paying to acquire customers you already have, which is a quiet and common leak.
Ask any tool whether it can build audiences from your customer data or only from platform targeting. The second is the commodity; the first is what nobody else can copy.
The prerequisite that outranks everything above
None of this matters if the conversion event does not fire where the ad lands. Open the exact destination URL with the pixel debugger running and confirm the event appears, with a value, deduplicated between browser and server. We ran a campaign to a 4.62% click-through rate that recorded zero product views, because the traffic arrived on collection pages where that event did not exist.
Campaigns you can see into
Adyft builds Google and Meta campaigns in your own ad accounts, with carousels that send each card to its own product and everything visible before it runs. 14-day free trial, no card.
Start free — no card needed