Google Shopping
Amazon Left Google Shopping for 31 Days. Here's What That Tells You About Your Business.
On July 22, 2025, Amazon pulled out of Google Shopping. No announcement. No wind-down period. Just gone.
By August 23, they were back in most markets -- 74% market share restored. But the US stayed dark. That's the most important data point in this entire story. Amazon didn't have a technical problem that got fixed in some markets and not others. They ran a deliberate experiment, and the US result told them something different from every other market. They chose to stay out.
Amazon had already cut US Google Shopping spend by around 50% in May 2025. The July exit completed the test. They wanted to know exactly what they'd lose -- and in the US, the answer was: not enough to justify the spend.
Most brands don't have that data. Most brands wouldn't dare run that experiment. And that gap in confidence is worth thinking hard about.
Incrementality testing is the discipline of measuring what you'd actually lose if you turned off a channel. Not attribution credit. Not modelled conversions. Actual revenue that wouldn't have happened without the ad.
Amazon ran a 31-day incrementality test on one of the world's most competitive ad platforms -- in public -- and came back with enough confidence to stay dark in the US while restoring spend everywhere else. That's not guesswork. That's measurement infrastructure most eCommerce brands simply don't have.
Ask yourself: if you paused Google Ads for 30 days starting tomorrow, what would actually happen to revenue? Not what your attribution dashboard says -- what would actually happen? If you don't know, you're making budget decisions based on faith.
That's the first thing Amazon's exit reveals. Not that Google Ads is bad. That knowing your incrementality is the difference between operating with conviction and operating with anxiety. Running the test matters -- but so does running it correctly. A geo-holdout needs a clean control region and enough time to separate signal from normal variance. The insight comes from the design, not just from pulling the lever.
When Amazon left, the auction data showed the expected bump. Optmyzr tracked July 23-29 against the prior week: impressions up 5%, clicks up 7.8%, CPC down 8.3%. On the surface, a windfall.
Then the conversion data came in. Conversion rate dropped 7.2%. ROAS dropped 4.4%. Conversion value fell 5.5%.
The traffic was real. The intent behind it wasn't what brands hoped.
The shoppers who typed "buy [product]" into Google that week had been expecting to land on Amazon. Fast shipping. Easy returns. Trusted checkout. Competitive pricing. When they landed somewhere else, most of them left. The clicks were displacement traffic, not incremental demand.
This plays out differently by category, which is instructive.
| Category | Conversion Change | Conversion Value | ROAS | Verdict |
|---|---|---|---|---|
| Electronics | +81.3% | +10.9% | +7.1% | Winner |
| Sporting Goods | +20.7% | -9.9% | Declined | Volume Trap |
| Apparel | Up | -9.5% | Declined | Volume Trap |
| Home & Garden | Up | -7.5% | Declined | Volume Trap |
Electronics saw conversions jump 81.3% -- Best Buy and comparable retailers can actually compete on the dimensions Amazon shoppers value: price-matching, same-day pickup, real storefronts. They absorbed the displaced intent because they had a credible alternative offer.
Most other categories fell into what Optmyzr calls the volume trap: more clicks, worse economics. Categories where non-Amazon retailers have a legitimate competitive position absorbed the displacement. Categories where they don't -- they got clicks they couldn't close.
The uncomfortable part: most brands saw the CPC dip and their first instinct was to scale budgets and capture the "opportunity." Some did. Most of them got burned.
The problem isn't bid strategy. It's that the traffic was conditioned for something they couldn't deliver. Google Shopping doesn't send you demand. It sends you intent -- and intent is shaped by the ecosystem around it. When Amazon dominates a category, the mental model for that category is built around Amazon's operating standards. You're not just competing on product. You're competing against an expectation set you didn't create and can't match on short notice.
That's not an ad account problem. It's a business positioning problem. And you can't bid your way out of it.
The Electronics exception shows the rule clearly. Best Buy won not because they outbid competitors, but because they had the physical infrastructure, brand recognition, and price guarantees to satisfy the same shopper intent. The ad was just the door. What was behind the door had to match what the shopper came in expecting.
This extends to the fundamentals most brands skip over. Amazon's product listings meet high feed quality standards -- accurate titles, structured attributes, clean images. When displaced shoppers hit a competitor's Shopping ad with a weak feed, the mismatch compounds. The ad gets the click. The listing loses the conversion.
For the brands that could compete -- and for those that couldn't -- the CPC relief was short-lived. Target, Etsy, and Wayfair moved into the vacuum within days. By the end of the first week in most categories, CPCs had largely normalised, though the timeline varied. Apparel saw sustained CPC drops of up to 10% for a week or so; other categories recovered faster.
When Amazon returned on August 23, impression share compressed again for competitors across those markets.
The "opportunity" lasted maybe a week in favourable categories. For most brands, it was measured in hours. Small brands that chased volume into this window -- without checking whether their offer could actually convert the intent driving it -- bought clicks at a discount that still didn't return. A cheap click that doesn't convert is just a cheap click.
The takeaway isn't "run an incrementality test" as a one-off exercise. It's about building the operational clarity to make channel decisions with conviction rather than reacting to auction fluctuations.
Know your own incrementality. Pick a region where you have decent volume. Cut spend by 40% for four weeks. Track what changes in revenue against a comparable region where spend stays flat. You don't need Amazon-scale infrastructure -- you need a clean comparison and the discipline to hold the test long enough for the signal to separate from noise. The answer will either confirm your spend is working or reveal that you've been overpaying for traffic that was already coming.
Understand what your traffic is actually shopping for. The Electronics vs Apparel split from this data is a useful diagnostic lens. Ask honestly: can your brand compete on the dimensions that conditioned the intent? Speed, price, returns policy, trust signals -- these aren't copywriting problems. If your offer can't meet the expectation, improving your bids won't help.
Don't confuse availability with opportunity. Cheaper CPCs mean less competition in the auction. That's only useful if the intent behind those clicks is convertible for your offer. Volume at a discount is still bad business if the conversion economics don't work.
Competitive moat is upstream of ad strategy. If you can't compete on the dimensions your target shoppers use to decide, improving your campaign structure is rearranging furniture. Fix what the channel exposes before you scale into it.
Amazon ran a 31-day test and came back with a market-by-market decision. Most brands spend 31 months in the same auction without ever questioning whether they're winning or just participating.
The 31-day test Amazon ran is available to you at a smaller scale. Run it.
Common questions about incrementality testing and Google Shopping strategy.
Incrementality testing measures how much revenue you would lose if you stopped running a particular channel or campaign. Rather than relying on attribution models -- which assign credit to ads that may have influenced a purchase -- an incrementality test creates a control group that doesn't see your ads and compares their behaviour to an exposed group.
The most common approach for eCommerce brands is a geo-holdout: pause or significantly reduce spend in one geographic region while maintaining spend in a comparable control region, then measure the difference in revenue outcomes. The gap between the two groups represents the incremental value your ads are actually generating.
When Amazon exits the auction, the shoppers who would have clicked Amazon ads still search -- but they land on competitor listings instead. The problem is that many of these shoppers were specifically seeking what Amazon offers: fast and free delivery, easy returns, trusted payment, and competitive pricing. When they land on a retailer that can't match those attributes, they don't convert at the same rate or value.
This is what Optmyzr's data describes as the volume trap: clicks increased because the auction is less competitive, but the quality of that traffic is lower because the intent behind it was shaped by Amazon's operating standards, not the competitor's actual offer.
Electronics was the standout category, with conversions up 81.3%, conversion value up 10.9%, and ROAS up 7.1% during Amazon's absence. The reason is that large electronics retailers like Best Buy can genuinely compete on the dimensions that conditioned the intent -- price-matching policies, in-store pickup, established brand recognition, and comparable return policies.
Most other categories experienced the opposite: more conversions in absolute terms, but lower conversion value and declining ROAS. Apparel (-9.5% conversion value), Sporting Goods (-9.9% conversion value), and Home and Garden (-7.5% conversion value) all fell into the volume trap. The pattern reflects how well non-Amazon retailers in each category can actually replicate the offer that conditioned the search intent.
The CPC relief was short-lived in most categories. When Amazon stepped back, rival platforms including Target, Etsy, and Wayfair increased their Google Shopping activity to fill the gap. This competitive backfill meant CPCs normalised within a few days to a week in most categories. Apparel was the exception, seeing CPC drops of around 10% sustained for longer than other verticals.
When Amazon returned on August 23, 2025, impression share compressed again for competitors across international markets. The window available to benefit from reduced CPCs was narrow -- measured in hours for some categories, a week at most for the best-positioned ones.
Amazon had already reduced its US Google Shopping spend by approximately 50% in May 2025, before the full July exit. The US remaining dark after the August 23 restoration suggests the incrementality test returned a different finding for the US market than for other markets -- specifically, that the revenue Amazon would lose by staying off Google Shopping in the US is low enough to justify the saving.
Amazon's brand recognition and direct traffic in the US is likely strong enough that many shoppers go directly to Amazon rather than discovering it through a Google Shopping ad. The same dynamic may be less pronounced in international markets where Amazon's brand penetration is lower, making Google Shopping a more important discovery channel.
I audit your account, identify where spend is going to waste, and give you a clear picture of what's actually driving revenue -- and what isn't.
Let's Talk