The feed health checklist I run before every account audit

When an ecommerce account underperforms, campaign structure gets blamed first. Sometimes that is correct. But Shopping and Performance Max can only work with the products Google can approve, understand, and match to relevant demand. If important products are missing, titles are vague, variant data conflicts, or landing pages disagree with the feed, changing bids is working on the visible symptom instead of the input.

That is why I review the product feed before I prescribe a rebuild. The purpose is not to earn a perfect diagnostics score. It is to identify which data problems limit commercially important inventory, which attributes weaken matching, and which operational gaps could undo future campaign work. A feed audit should end with a prioritized action list, clear ownership, and a way to verify that each change reaches Merchant Center.

I use the checklist below as a repeatable framework, then adapt it to the catalog. Apparel variants, custom furniture, supplements, electronics, and replacement parts do not need the same title logic or identifiers. Good feed work combines platform requirements with the way real shoppers describe and compare the products.

Start with eligibility and catalog coverage

My first question is basic: which sellable products can actually serve? I compare the ecommerce catalog with Merchant Center totals for active, limited, pending, expiring, and disapproved items. I break the result down by destination, country, feed label, category, and business importance. A small overall disapproval rate can still be serious if it contains most of the best sellers or a launch the team is actively promoting.

I group diagnostics by root cause rather than working through a flat list of affected items. One broken shipping rule may affect hundreds of products. A theme setting may cause price mismatches across every discounted variant. A feed rule may overwrite the brand field each night. Grouping the symptoms around a cause makes the work faster and reduces the chance that tomorrow's sync reverses today's manual fix.

Prioritize risk and revenue separately

Account-level policy or website issues come first because they can interrupt the entire program. After account safety, I rank item problems by commercial effect: best sellers, high-margin categories, seasonal products, launch inventory, and items that historically attract meaningful demand. This keeps a team from spending a day polishing low-priority warnings while a small number of important products remain ineligible.

If the account is suspended or a large part of the catalog disappears, use a recovery sequence instead of making scattered edits. My guide to Merchant Center recovery explains how I trace the source and handle appeals safely.

Confirm identity, variants, and product facts

Google uses identifiers and consistent attributes to understand what a product is. I review brand, GTIN, MPN, condition, item group ID, color, size, material, pattern, gender, and age group where they apply. I do not invent a GTIN to clear a warning. If a custom product legitimately has no identifier, the feed should represent that accurately and use the other available attributes to describe it well.

Variant structure deserves its own sample check. The item group should connect genuine variants without collapsing different products into one family. Price, availability, image, color, and size should reflect the exact landing-page selection. If every variant points to a generic parent page or a sold-out option, the feed may technically process while the shopper experience and conversion rate deteriorate.

Trace the source before editing the output

An attribute can originate in the store catalog, an app, a scheduled export, a supplemental source, a transformation rule, structured data, or an automatic item update. I pick representative products and trace each value through that chain. A manual edit in Merchant Center is rarely a durable fix if the primary source publishes the wrong value again a few hours later.

• Use real manufacturer identifiers; never substitute internal SKUs for GTINs.
• Check that item group IDs join true variants and remain stable across updates.
• Match variant URLs, prices, availability, images, size, and color to the selected landing-page option.
• Remove discontinued products intentionally and confirm new launches arrive on schedule.

Write titles for identification and matching

A product title should help a shopper identify the item and give Google useful matching information. The best attribute order depends on the category. Brand, product type, model, material, size, color, compatibility, quantity, or another differentiator may deserve an early position. I review search language, product-page headings, category norms, and the feed's available data before applying one template across a large catalog.

I look for titles that repeat keywords, bury the distinguishing attribute, use unexplained internal codes, or rely on promotional claims. I also check truncation. A title can be factually complete and still be weak if the first visible portion makes several variants look identical. The goal is clarity, not a maximum character count.

Descriptions support the title; they do not rescue it

Descriptions can add materials, use cases, specifications, care details, and other product facts, but the most important identity information should not be hidden there. I remove boilerplate that appears on every item and make sure claims match the landing page. If an attribute matters to eligibility or frequent shopper comparison, I prefer a structured field over a sentence that Google must infer.

Make taxonomy useful to both Google and the business

Google product category provides standardized context. Product type reflects the merchant's own hierarchy. Custom labels create reporting and campaign controls. I review all three, but I use them for different jobs. The taxonomy should be accurate enough for matching and stable enough for the paid search and merchandising teams to use without rebuilding it every month.

Useful custom labels often describe margin band, bestseller status, season, inventory depth, promotion, lifecycle, price tier, or another business characteristic. I only add a label when it supports a real decision. A label that cannot change a budget, bid, report, test, or merchandising conversation creates maintenance without control.

Avoid segmentation for its own sake

It is possible to turn clean product data into a campaign structure with dozens of tiny groups. That usually reduces useful volume and makes management slower. I want enough taxonomy to isolate different economics or priorities, not a separate campaign for every label combination. The feed should increase optionality; it should not force unnecessary complexity.

Audit landing pages, destinations, and policy signals

The feed review continues onto the website. I verify price, sale price, availability, currency, condition, shipping information, and the selected variant on both desktop and mobile. I look for redirect chains, blocked pages, out-of-stock defaults, intrusive overlays, missing contact or return information, and structured data that conflicts with the visible page. These problems can affect approval, conversion, or both.

I also check destination settings. Products may be approved for one program or country but missing from another. Feed labels, languages, currencies, shipping services, and target countries need to line up with campaign targeting and the site's ability to fulfill the order. A technically active product is not useful if it is eligible in the wrong market.

Connect product IDs to measurement

The product identifier used by the feed should be reconcilable with conversion and analytics data. I check whether purchase events send the expected item IDs, quantities, currency, and values. If the commerce platform reports one identifier while Google Ads and Merchant Center use another, product-level analysis becomes unreliable and dynamic remarketing can break.

I do not expect every platform total to match perfectly. Attribution, time zones, returns, consent, and processing create differences. I do expect the direction and item mapping to make sense. Before automated bidding receives a new value rule or profit signal, I verify that the underlying transaction data is stable enough to support it.

Check update speed and operational ownership

A feed can look healthy during an audit and fail during the next price change. I document how often product data updates, how sale windows are handled, what triggers automatic updates, and who owns fixes in the store, feed app, Merchant Center, and ad account. Fast-moving inventory may require API updates or more frequent fetches; a stable catalog may not. The schedule should match the business.

I also look for monitoring. Teams should know when active item counts drop, disapprovals spike, best sellers disappear, or price mismatches grow. The alert should point to a commercial problem, not just create more notifications. This is one reason I treat feed health as an ongoing PPC responsibility rather than a one-time setup project.

My reusable feed audit sequence

1. Export catalog and Merchant Center counts by status, market, category, and business priority.
2. Resolve account-level policy and website risks before lower-impact item warnings.
3. Trace identifiers, variants, price, availability, and images back to their source of truth.
4. Review title and description patterns using representative products and real query language.
5. Validate taxonomy and custom labels against decisions the team actually needs to make.
6. Test landing pages, destinations, shipping, and mobile variant behavior.
7. Reconcile feed product IDs with purchase-event and analytics item data.
8. Assign owners, deadlines, verification steps, and monitoring for every material issue.

Only then do I judge campaign structure

Once eligibility and product data are trustworthy, campaign recommendations become easier to defend. I can see whether a category lacks demand, loses auctions, has weak economics, or was simply absent from the feed. I can group products by a meaningful business dimension and test changes without wondering whether a background sync changed the inventory halfway through.

Feed health also shapes how I handle Performance Max cleanup. If you want to discuss feed, tracking, campaign structure, and wasted spend before committing to ongoing management, book a discovery call. We can identify the account questions worth examining first.