"Google Doesn't Work for Our Ecom Product" Is a Diagnosis of Your Setup, Not Your Market

"Google doesn't really work for our kind of product."

I've heard this from founders for 16 years. It's almost never a statement about Google. It's a statement about one badly built test, run once, years ago, that nobody has re-examined since.

Here's what actually happened in most of these accounts. Someone turned on a campaign, usually PMax, because it's the path of least resistance. No keyword strategy, or broad match doing the job of one. Traffic dumped on a cold product page. Tracking never audited, so the numbers were fiction from day one. Three or four weeks later the ROAS looked ugly, the founder killed it, and a belief was born.

But look at what was actually tested. Not "does Google work for this product." What got tested was: "does an untargeted campaign with broken tracking, pointed at the wrong page, judged on Meta's timeline, make money in three weeks?"

The answer is no. The answer would be no for every product ever made.

Who Actually Built These Broken Accounts

Here's the part the other "Google doesn't work" articles skip, because it's uncomfortable for half the industry.

Most DTC brands didn't set up Google themselves. Their Meta agency did. Meta is where the brand grew up, Meta is where the agency's skill is, and when the founder asked "should we do Google too?", the agency said yes, because saying no means someone else gets the line item.

Worth separating two different problems here. A trusted, high-delivery Meta account (the kind our best ad account providers ranking covers) fixes account bans and restricted delivery. It says nothing about whether the team running your Google spend actually understands intent-based buying. A Meta-first agency can have a perfectly healthy account and still build a broken Google campaign, those are unrelated skills.

Now follow the incentives. Nobody has to lie for this to happen:

A Meta-first agency running Google gets rewarded for brand capture, not cold acquisition. Brand search converts the demand Meta already created. It produces a beautiful ROAS number with zero skill and zero risk. Cold non-brand scaling (Shopping structure, feed work, intent mapping, query sculpting) is hard, takes 90 days to prove, and can visibly fail. Which one does a Meta shop pick?

So the account gets capped at brand. Sometimes literally: brand search plus a loose PMax that's 70% brand anyway. The Google line item exists, the dashboard looks green, and the channel never gets tested for what it's actually for: acquiring customers who have never heard of you.

And then the CPC inflation starts. Low tROAS bidding on your own brand terms tells Google "pay whatever it takes." You end up paying $3 for a click that cost $0.40 two years ago, to capture a customer Meta already sold. The brand CPC climbs, the blended ROAS still looks fine, and nobody in the room has an incentive to ask why.

This isn't a conspiracy. It's a skill gap wearing a strategy's clothes. Meta buying is audience-first; Google buying is intent-first. An agency that thinks in audiences will always build Google as a retargeting net under their Meta funnel, because that's the only shape of Google they can see.

Then, when the founder eventually asks why Google isn't scaling, the verdict comes back: "Google doesn't really work for your kind of product." The channel takes the blame. The execution walks free.

The Pullback Test Grift Built on Top of It

There's now a second act to this play, and it's worse.

The same Meta-first operators have started selling "pullback tests": turn off your Google spend, watch revenue hold, conclude you were overspending, pay them to keep it off. And in a brand-capped account, the test works. Of course it does. If 80% of your Google spend was brand capture, pausing it just moves those clicks to your organic brand listing. Revenue holds. The test "proves" the spend was waste.

But the test only proved what the setup guaranteed. Pause brand capture, and organic brand absorbs it, because an organic substitute exists. Pause non-brand Shopping in a properly built account, and there is no substitute. Your non-brand organic slots belong to Amazon, marketplaces, and affiliate listicles. That click doesn't come back to you for free. It goes to the competitor who kept bidding.

The largest pullback test in history already ran. Amazon zeroed out Google Shopping globally in July 2025: the company with 4 billion product searches a month, Prime lock-in, and the strongest direct traffic on earth. Within 30 days it was back at ~74% impression share in every international market, "as if nothing had happened." (The US is the strange exception, still at zero a year later, unexplained by either company, and worth its own article.) If the brand with the most organic substitutes on the planet looked at its own pullback data and turned Shopping back on everywhere else, what exactly does a mid-sized DTC brand, with none of those substitutes, think its pullback reveals?

It reveals one thing: that your account was built for brand capture. Which is the disease, not the diagnosis.

What a Fair Test Actually Looks Like

If your Google history is one Meta-agency setup and one pullback verdict, you have never tested Google. Here's what a real test is:

• ▶Audit tracking first. One true conversion action, no GA4 double-counting, no add-to-carts posing as purchases. The bidding optimizes toward whatever number you feed it, feed it fiction and it confidently loses money.
• ▶Split brand from non-brand, and judge them separately. Brand gets a high tROAS ceiling so it stops inflating its own CPC. Non-brand is the actual test.
• ▶Clean the Merchant Center feed before a penny goes to Shopping. Titles, GTINs, product types. Shopping is a feed-quality auction pretending to be an ad platform.
• ▶Match page to intent. Transactional queries to PDP. Problem-aware queries to an advertorial. A symptom search landing on an add-to-cart button is a bounce you paid for.
• ▶Run a discovery layer. DSA or a broad-match sandbox with tight negatives, so the test covers real demand, not the keywords someone guessed in an afternoon.
• ▶Judge on new-customer acquisition over 90 days. Not blended ROAS at week three. Blended ROAS in a brand-heavy account is a mirror, not a metric.

Run that and you'll get a true answer. It might still be "this is a hard channel for us." Fine. At least it'll be real, instead of a belief you inherited from an agency that was never equipped to disprove it.

Three Questions That Settle It in Five Minutes

Before you accept "Google doesn't work for us" ever again, pull these three numbers from whoever ran the account:

1. 1.What percentage of spend went to brand terms, including brand queries inside PMax?
2. 2.What was the new-customer ROAS, as opposed to blended?
3. 3.What did the search terms report surface that wasn't already on the keyword list?

If the answers are "most of it," "we don't split that," and "we didn't look," you don't have a channel verdict. You have a brand-capture operation that was never pointed at cold demand, judged by a metric designed to flatter it, and shut down by a test designed to confirm the shutdown.

That's not Google failing your product. That's your product never having been on Google.