eCommerce PPC Agency

WordStream’s 2026 study covers 13,474 US search advertising campaigns from April 2025 to March 2026, using medians rather than means, with at least 52 active campaigns per subcategory. The retail categories share a shape.

Shopping, collectibles and gifts convert at 4.01%, with a $4.14 cost per click and a $49.40 cost per lead. Apparel, fashion and jewelry convert at 4.50%, at $4.44 per click and $97.51 per lead. Furniture converts at 2.99%, at $3.97 per click and $106.70 per lead. The all-industry median sits at 8.18% conversion, $5.42 per click and $66.69 per lead.

Every one of those conversion rates is roughly half the all-industry median. Furniture is barely a third.

That’s the nature of considered consumer purchases with a lot of comparison shopping and a lot of browsing that was never going to end in an order. But it has a consequence most eCommerce advertisers underestimate: because each click is cheap, the cost of a low conversion rate stays invisible until it shows up as a monthly number.

A law firm paying $9.87 a click notices immediately when the phone doesn’t ring. A brand paying $4.14 doesn’t notice anything in particular, because no individual click hurts. The discipline that expensive traffic forces on other categories has to be imposed deliberately here.

Three practical consequences.

Watch cost per order, never cost per click. A campaign whose CPC drops 20% while conversion rate drops 30% has gotten worse and will look better in most dashboards. Cheap clicks are frequently cheap because they’re low intent.

Segment by intent before you optimize. A branded search, a specific-model search, a category search and a broad discovery impression are four different products at four different real costs. Averaging them produces a blended number that hides the one that’s losing money.

And accept that some of the traffic isn’t the point. In a category converting at 4%, most sessions are research. The real question is whether that 96% is being reached cheaply enough to be worth having and brought back reliably through remarketing and lifecycle campaigns.

Here is the structural fact that separates eCommerce PPC from every other kind, and it’s the thing most brands under-resource by an order of magnitude. 

Shopping and Performance Max don’t match on keywords you write. They match on the data in your product feed. Which means the highest-leverage work in an eCommerce paid program isn’t in the ads interface at all — it’s in Merchant Center. 

Google’s own documentation is direct about it. The product title, it says, is “one of the most prominent parts of your ad or free listing,” runs to a 1–150 character limit, and is used to connect products with customer searches. Google may also dynamically rewrite titles by extracting attributes from your landing pages and images to better match a query — meaning the platform is actively trying to compensate for weak titles, and doing a better job when the source data is good. 

The guidance that follows from that is specific and almost nobody implements it fully: 

Put the most important details first. Google notes users typically see only the first ~70 characters on most screens. A title beginning with your brand name and ending with the attribute someone actually searched for is a title optimized for the wrong reader. 

Give variants distinguishing detail. Size and color belong in the variant’s own title. A feed where forty variants share one title is a feed competing with itself. 

Describe bundles fully, and label subscriptions as subscriptions — Google asks for the word “subscription” and the duration. 

And fix the unglamorous fields. GTINs, brand, condition, availability, price accuracy, product type and Google product category. Disapprovals and stale availability data don’t announce themselves in the Ads interface; they just quietly reduce what you’re eligible to show for. 

We’d rather spend the first month of an eCommerce engagement in the feed than in the bidding. It compounds, it improves organic free listings at the same time, and unlike a bid change it doesn’t stop working when someone else raises theirs. 

Performance Max is where most eCommerce budget now sits, and the standard complaint about it has been opacity. That’s genuinely improving, and the practical response is to use the new reporting rather than to keep arguing with the format.

Google announced a set of steering and reporting changes for 2026, published January 9, 2026. First-party audience exclusions let you exclude existing customers and push the campaign toward new-customer acquisition rather than paying to re-buy people who were coming back anyway. For a subscription or repeat-purchase brand this is one of the highest-impact settings available. Network segmentation in placement reporting, under the “When and where ads showed” tab, shows which networks actually served your ads. Budget reporting now includes end-of-month spend projections and the effect of daily budget changes. Audience reporting has expanded to include age and gender breakdowns by segment.

None of that turns PMax into a manual campaign, and it isn’t meant to. What it does is make three long-standing questions answerable: are we paying to acquire customers we already had, where is the inventory actually running, and is the budget going to land where we planned.

The operating posture that works is to give the algorithm good inputs and clear exclusions, then read the reports it now provides rather than trying to micro-manage placements it won’t hand over. Good inputs mean an accurate, well-structured feed. It means well-defined conversion actions with sensible values.

The last structural point, and the one that most often separates a program that looks successful from one that is. 

Target ROAS optimizes toward revenue per dollar of spend. It knows nothing about your margin. So a bid strategy hitting a 4× ROAS target across a catalog with margins ranging from 15% to 70% will systematically over-invest in whatever converts easily and cheaply, which is frequently the low-margin, heavily discounted end of the range. 

The fix is to get margin into the system. Feeding profit rather than revenue as the conversion value — or supplying margin data at the product level so the platform can bid toward contribution rather than turnover — changes what the algorithm chases. It’s a data plumbing exercise rather than a media one, and it’s the highest-return unglamorous project available in most eCommerce accounts. 

Two related habits worth building at the same time. Separate branded from non-branded, so the ROAS on people who were already looking for you doesn’t flatter the number for people who weren’t. And measure new-customer acquisition separately from total revenue, because a program that looks efficient may simply be harvesting repeat purchases you’d have received anyway — which is exactly what PMax’s new first-party exclusions are there to address.

Apparel, footwear and jewelry catalogs run deep on variants, which puts nearly the entire job on the feed. That means variant-level titles, accurate sizing and color attributes and asset groups that segment by commercial logic rather than by site navigation.

Apparel’s 4.50% conversion rate and $97.51 cost per lead mean fit and returns content on the landing page does as much for paid performance as anything in the ads account.