The $40,000 Mistake That Lives In Your Product Feed
A home goods retailer came to us spending roughly $40,000 a month on Performance Max with a blended return on ad spend stuck at 2.1x. The team had spent weeks tweaking budgets, swapping headlines, and adding video assets. None of it moved the number. When we pulled their Google Merchant Center feed, the cause was obvious in ten minutes: half the catalog had titles like “Throw Blanket” with no brand, no material, and no size. Google had nothing to match against the long, specific queries that convert, so it spent the budget on cheap, broad impressions.
We rewrote the titles, fixed the missing identifiers, and added custom labels for margin tiers. Within six weeks the same budget returned 4.3x. The campaign settings barely changed. This is the single most important thing to understand about Google Ads for ecommerce: in Performance Max and Shopping, the product feed is the campaign. Everything else is a dial you turn after the feed is right.
Why The Feed Is The Real Lever, Not The Settings
Performance Max for retail is fundamentally a feed-driven system. According to Google’s own Performance Max for retail documentation, the campaign requires a linked Merchant Center account, and the product feed supplies the data that drives ad creation and targeting across Shopping, Display, YouTube, and Search inventory. The assets you add (headlines, images, video) expand the formats Google can build, but the feed is what tells the system what you sell and which queries you deserve to appear for.
The spend distribution makes this concrete. For healthy ecommerce accounts, feed-based Shopping placements typically account for the large majority of where the budget actually goes, and analyses of Performance Max accounts consistently show product feed quality is the biggest single lever on results. Store Growers’ Performance Max ecommerce guide puts it bluntly: a weak feed means weak Performance Max results, and no amount of campaign structure fixes that. You should expect roughly 60 to 80 percent of a healthy campaign’s spend to land on Shopping placements, which means feed quality directly governs the majority of your media budget.
For decision-makers evaluating an agency, this is a useful filter. If a prospective partner opens with talk of bid strategies and asset rotation before they have audited your feed, they are optimizing the steering wheel of a car with no engine. The same feed-first discipline applies to organic visibility, which is why we treat technical SEO and feed hygiene as two sides of the same data-quality problem.
Feed Optimization: Where The Money Actually Moves
Titles Carry The Most Weight
Product titles are the highest-leverage field in the entire feed because Google matches them against search queries. Per Google Merchant Center’s title specification, the title attribute accepts up to 150 characters, but shoppers typically see only the first 70 or so depending on their screen. That means you front-load the details that differentiate the product and matter most to buyers.
A title like “Dress” wastes the field. A title like “Reformation Gabrielle Linen Midi Dress, Sage Green, Size 8” tells Google the brand, material, style, color, and size, so it can surface your product for the exact long-tail query a ready-to-buy shopper types. Google’s guidance is specific about what to avoid: no promotional text such as price or shipping, no all-caps for emphasis, and no gimmicky symbols. A practical title formula for most catalogs is Brand plus Product Type plus key attributes (color, material, size, model number), ordered by what a buyer searches for first.
Identifiers, Images, And Descriptions
Beyond titles, three things make or break feed approval and performance:
- Product identifiers. The required attributes for every product are id, title, description, link, image_link, availability, and price, and for most products brand plus a GTIN or MPN. A GTIN (the global trade item number behind a UPC or EAN barcode) is required when the manufacturer has assigned one, and missing identifiers suppress eligibility and depress match quality.
- Images. Use high-resolution photography on consistent backgrounds, free of watermarks and overlaid promotional text, which violate Google’s policies. Image quality directly affects click-through rate, and click-through rate feeds the algorithm’s confidence in your products.
- Descriptions. Write descriptions that mirror how customers actually describe the product, with natural attributes and use cases, rather than copying a terse manufacturer line.
Custom Labels Are Your Profit Steering Wheel
Custom labels are the quietest power feature in the feed. They let you tag products by margin tier, bestseller status, seasonality, or price band, which then lets you segment campaigns and set different ROAS targets against different economics. Without them, the algorithm tends to chase your existing bestsellers and ignores the high-margin products you actually want to grow. The same attention to structured, query-aligned data that drives feed performance also underpins strong on-page optimization on the product pages those ads point to.
Asset Groups And Listing Groups: Structure Without Over-Engineering
Once the feed is clean, structure comes next, and the most common error is over-segmentation. Google’s retail documentation recommends that each asset group target a distinct set of products, with listing groups organizing which products belong where. A sensible starting point is three to seven asset groups per campaign, organized by category, theme, or margin tier rather than by individual SKU.
A tactic worth knowing: you can build feed-only asset groups that exclude creative assets, which pushes Google to concentrate budget on Shopping placements rather than spreading it across lower-converting Display and YouTube inventory early on. Adding asset groups with images, video, and text later unlocks the broader inventory once you have conversion data to support it.
The hard rule on structure is conversion density. Performance Max needs roughly 20 to 30 conversions per month per campaign to exit the learning phase and bid efficiently. Splitting a catalog into five thin campaigns of ten conversions each will lose to one consolidated campaign of fifty conversions every time. A reasonable scaling ladder looks like this:
- Under $3,000 per month: a single Performance Max campaign covering all products.
- $3,000 to $10,000 per month: one main campaign, optionally splitting hero products once they clear 30 conversions a month.
- $10,000 to $50,000 per month: two to three campaigns segmented by margin tier.
- $50,000 and up: three to five campaigns with fuller segmentation.
One more structural safeguard: use brand exclusions so Performance Max does not claim conversions from your branded search terms, which a dedicated branded Search campaign should handle far more cheaply. Without exclusions, Performance Max often inflates its reported ROAS by harvesting cheap brand traffic that would have converted anyway.
Setting ROAS Targets That Reflect Real Margins
Target ROAS bidding is where strategy and accounting meet. Google’s Target ROAS documentation explains that the system sets bids to maximize conversion value while aiming for the average return you specify, using real-time signals like device, location, and time of day. Google recommends a minimum of about 15 conversions in the past 30 days for Search and Shopping before target ROAS bidding has enough data to work well, and many practitioners prefer 30 or more conversions a month for stable Performance Max performance.
The practical mistakes are setting a target too high too fast, and applying one target across products with wildly different margins. Mixing a 60 percent margin product and a 10 percent margin product in the same campaign forces the algorithm to average their economics, which usually means it quietly defaults to whatever already sells. Separate them, and set each a target tied to its breakeven ROAS.
A disciplined approach:
- Start within about 20 percent of your current actual ROAS rather than picking an aspirational number.
- Raise the target in 10 to 15 percent increments, allowing one to two conversion cycles to stabilize between changes.
- Anchor every target to gross margin, so your minimum profitable ROAS is the floor, not a guess.
Worth noting for 2026: Google has been simplifying its bidding labels, with “Maximize conversion value with a target ROAS” being relabeled simply as “Target ROAS.” The mechanics are unchanged; only the naming is cleaner.
The Hybrid Setup Most Strong Accounts Run In 2026
The current consensus among performance teams is not Performance Max alone. The stronger pattern is running Performance Max alongside Standard Shopping in parallel campaigns. Standard Shopping gives you precise, query-level bidding control to protect branded and bestselling products, while Performance Max handles broader prospecting and scales across Google’s full inventory. This hybrid keeps the transparency of Standard Shopping where you need control, and the reach of Performance Max where you want growth. For stores combining paid acquisition with broader demand creation, this pairs naturally with a full-funnel ecommerce demand generation strategy rather than a single-campaign bet.
Measure The Feed, Not Just The Dashboard
Because Performance Max reports at a more aggregated level than older campaign types, your measurement has to reach into the feed and the Merchant Center diagnostics, not just the Google Ads dashboard. Track disapproved and limited products, missing identifier rates, and ROAS by custom label so you can see which margin tiers and which product groups are actually profitable. Feed optimization is never finished; titles, images, and labels should be tested and refined continuously, and the compounding gains usually outweigh anything you get from fiddling with bids. Rigorous reporting and analytics is what turns those feed signals into decisions instead of guesses.
The takeaway for any business evaluating where to invest: in google ads ecommerce performance max campaigns, the product feed is the lever, asset groups are the structure, and ROAS targets are the throttle. Get the feed right first, and the rest of the machine finally has something worth optimizing.
Sources
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.