What to Fix First in Your Product Feed
Product data problems rarely announce themselves as product data problems. They show up as Merchant Center disapprovals, a Shopping campaign that used to convert and quietly stopped, or a customer ordering something the site already sold out of.
Feeds, the Merchant API, and Product schema all touch that problem, and most growing stores eventually use all three. None of them fixes the same part of it, which is why most guides on this topic land on “you need all of it” and stop there. True, and not much help to a team with one developer sprint and a Shopping campaign that plateaued months ago.
Something else changed since most of what is written about this topic went live.
External guidance
If your store still calls that API directly, the countdown is over. Check your Merchant Center diagnostics and API error logs today, not next sprint.
Most stores are not the ones exposed. External guidance
Find the Highest-Cost Problem First
Before buying a tool or scoping a migration, find the specific thing that is losing you visibility, spend, or trust. This table is the fast version; each row links to the section that explains it.
| Your situation | Do this next | Why |
|---|---|---|
| Running Shopping ads and seeing disapprovals | Audit price, availability, and identifier matching against your live pages | A common cause of disapprovals is a feed that disagrees with the site, not a missing tool |
| Shopping campaigns are live but performance has flattened | Rewrite titles and fill in missing attributes | A thin or generic title limits which searches your listing can match |
| Product pages exist with no structured data | Add Product and Offer markup | Markup makes a page eligible for search enhancements; it does not guarantee one or move rankings by itself |
| Still calling the Content API directly, in-house | Confirm migration status today; Google’s extension request form buys time, not a pass | Direct Content API calls have been seeing progressive errors since September 1, 2026 |
| Selling across several marketplaces, or past a few hundred SKUs | Evaluate a feed management platform | Manual spreadsheet updates start missing changes once more than one channel and more than one person touch the catalog |
| Prices or inventory change several times a day | Look at a real-time connection, not just a faster fetch schedule | A scheduled file fetch always leaves a gap between a change and Google seeing it |
What Feeds, Schema, and the API Each Do
The most common form a product feed takes is a scheduled file: titles, prices, images, and inventory levels, generated from your platform and fetched by Google Merchant Center on the schedule you set. It is the baseline way products get into Shopping and Performance Max, and it needs no developer to maintain once it is running. The tradeoff is timing. If a flash sale starts at 9 a.m. and your feed last fetched at 7 a.m., Google is still showing the old price for two hours, which is exactly the gap where disapprovals and unhappy customers start.
The Merchant API is the other form: your systems create and update product data programmatically instead of waiting for the next fetch. It replaced the Content API, and actual latency still depends on how your integration is built. It is worth the added complexity when price or inventory moves faster than a fetch schedule can reliably reflect: flash sales, dynamic pricing, or inventory pulled from several suppliers during the day. A store with weekly pricing and no flash sales does not have that problem yet, so the API is solving something that has not happened to it.
Product structured data, Product and Offer markup written as JSON-LD, is a different job. It makes a page’s price, availability, brand, and identifiers machine-readable.
External guidance
Markup earns eligibility. It does not buy a guaranteed rich result, a ranking change, or a specific click-through number, whatever percentage a sales deck attaches to it.
Markup also feeds AI shopping tools, though not the way it gets described in most feed articles. ChatGPT, Perplexity, and Gemini each pull from a different mix of merchant feeds, crawled pages, and citations, and the differences are specific enough to deserve their own piece rather than a paragraph here. If AI visibility is the reason you opened this page, Getting Products Found in AI Search covers what each platform uses and what the correlation research does and does not support.
Which to Prioritize With What You Already Have
Feeds Without Schema
Common setup: Shopping is running, products show up, and schema never came up because the feed seemed to cover it. It does not. Your feed gets products into Merchant Center for paid campaigns. Schema is a separate signal that makes the same product pages eligible for rich results in organic search, and markup is one of several inputs AI shopping tools draw on, covered in the AI section above.
Adding basic Product and Offer schema rarely needs developer time. RankMath, Yoast, and Schema Pro handle most of it automatically on WordPress and WooCommerce, and Shopify has apps that do the same. If you have feeds and no schema, this is usually the highest-impact, lowest-effort item on the list.
Schema Without Feeds
Less common, usually a store that grew through organic and SEO and never built out paid Shopping. Schema alone will not get products into Shopping ads or Performance Max; both require a feed submitted to Merchant Center. If paid Shopping belongs on the roadmap, a feed is required infrastructure, not an optional add-on.
Shopping ads put a price, an image, and your listing directly in front of someone already searching, which is why most retailers run it as its own channel rather than a substitute for organic. If you want that channel, build the feed. Schema will not stand in for it.
Neither Feeds Nor Schema Yet
Start with a feed if paid acquisition is the near-term priority. It is the faster path to measurable revenue, and a Shopping campaign can launch quickly once the feed is approved and the account is configured. Start with schema if organic is the priority and paid spend is not yet part of the plan; it compounds without an ongoing media budget behind it.
Most stores eventually need both. If resources only allow one this quarter, let the answer follow whichever channel, paid or organic, is actually driving growth today.
Audit Your Feed Before You Add a Tool
Start by naming where each field actually comes from. A common pattern in the accounts we audit: price comes from the ecommerce platform, inventory from an ERP, specifications from a spreadsheet nobody updates on a schedule, and promotions from whatever marketing tool ran last. When two of those disagree, whichever one fed the export last wins by accident, not by design. Pick one owner per field before auditing the data that field produces.
The feed and the page a shopper lands on after clicking the ad are one system, not two. A feed that lists $49 and a page that charges $54 is the same defect whether Google catches it as a disapproval or a customer catches it after checkout.
Many stores can recover real performance by fixing the feed they already have before replacing it with a platform. Work through this list against a sample of your top revenue SKUs, not the whole catalog on the first pass.
Source of truth
- Name which system owns price, inventory, title, images, and shipping data, and confirm nothing downstream can overwrite that source without going through it.
- Compare advertised price, availability, shipping, and condition against the live product page and checkout for each sampled SKU, not the page as it looked when the feed was last generated.
Merchant Center
- Pull diagnostics by item issue, affected product count, and severity, and note which issues recur month over month.
- Check identifier coverage: GTIN, MPN, brand, and whether
identifier existsis set correctly for products that legitimately have none. - Check Google product category and your own product type taxonomy against how a buyer would actually search.
Product quality
- Read titles for buyer language: size, material, compatibility, and the specific use case someone would search for, not just the product name.
- List attributes missing that matter for the purchase decision: dimensions, compatibility, age group, material, or technical specifications, depending on the category.
- Check variant handling for duplicate listings, broken parent-child relationships, or a missing item group ID.
- Confirm image quality and coverage against Google’s image requirements, including additional images where the category benefits from them.
Lifecycle and freshness
- Check how out-of-stock, discontinued, preorder, bundle, and subscription products are represented in the feed.
- Compare your feed’s update frequency against how often price and inventory actually change.
Most of this lives across a handful of systems, not one dashboard. Merchant Center holds diagnostics, product status, and account issues. Google Ads holds product-level cost, conversion, and search-term data. Your ecommerce platform holds the product record itself. An ERP, PIM, or supplier feed is where a source-of-truth conflict usually starts. Application logs are where an API error or a failed scheduled job actually shows up, days before anyone notices the campaign underneath it.
Watch the Feed After You Fix It
An audit is a snapshot. A feed drifts again the moment someone changes a price, adds a variant, or launches a promotion without checking what it touches downstream, so monitoring is part of the fix, not a separate project.
- Check Merchant Center diagnostics on a set schedule, not only when a campaign underperforms.
- Track disapprovals by issue type and affected SKU count, not just the total, so a recurring issue does not hide inside a shrinking number.
- Set an alert for a spike in out-of-stock items, price mismatches, or rejected products, especially during a sale or a platform migration.
- Spot-check your highest-revenue SKUs after any catalog update, pricing rule change, or new supplemental feed.
- Name one person responsible for feed health. It does not have to be whoever built the integration, but it has to be someone, or the job defaults to whoever happens to notice a drop in Shopping traffic.
Platform, Custom Build, or Neither
Most stores never need this decision. A native feed from Shopify, WooCommerce, or BigCommerce, cleaned up with the checklist above, covers a single channel and a stable catalog. The decision gets real once you sell through several channels, manage thousands of SKUs, or pull data from more systems than your team can reconcile by hand.
| Compare on | Feed platform | Custom build |
|---|---|---|
| Setup time | Days to weeks, using existing connectors and templates | Weeks to months of discovery, development, and QA |
| Who can maintain it | Usually a marketer or merchandiser | Usually a developer |
| Cost shape | A predictable subscription | Upfront build cost plus ongoing engineering time |
| Fit for unusual rules | Strong for common transformations, strained by proprietary logic | Can model pricing, bundling, or fulfillment rules no platform anticipates |
| Who owns breakage | The vendor, for the platform itself | Your team, for everything |
| Best fit | Several channels, limited developer time, complex but common rules | Proprietary systems, real-time requirements, a team that can support it long term |
The table starts at “platform versus custom” because that is where a real decision exists. Most stores are not there yet:
| If this sounds like you | Likely fit |
|---|---|
| One channel, a stable catalog, no complex rules | A native feed, cleaned up and monitored |
| Several channels, common-but-complex rules, limited developer time | A feed management platform |
| Proprietary pricing or fulfillment logic, real-time requirements, a team to maintain it | A custom build |
Feed management platforms (Channable, DataFeedWatch, and Feedonomics are the three we see most often) earn their subscription when the constraint is channels and speed: more than one destination for the feed, rules that are complex but not unique to your business, and a marketing team that needs to change a title or a label without filing an engineering ticket.
Build only when the constraint is genuinely yours: pricing, bundling, or fulfillment logic that does not exist anywhere else, a real-time requirement tied to your own inventory system, or product data that already needs to serve more than the ad feed. Whoever builds it also has to still be around to maintain it. A custom integration with no named owner a year after launch is a liability, not an asset.
Compare total cost, not the invoice. A platform’s real cost is the subscription plus whatever it still cannot do for your catalog. A custom build’s real cost is development plus the maintenance, monitoring, and API updates nobody budgets for the year after launch. A rough way to run the math: setup cost, plus monthly fee, plus your team’s hours on it each month at their blended rate, plus a realistic estimate of what a typical feed error costs you in paused spend or lost sales, multiplied by the months you are comparing. A $400-a-month platform that takes two hours of upkeep at a $75 blended rate and catches one avoidable disapproval a quarter is closer to $700 a month all-in than the sticker price suggests. Run the same arithmetic on a developer’s build estimate before assuming custom is cheaper because the invoice is smaller.
Most established stores land on a hybrid. The ecommerce platform, ERP, or a PIM stays the source of truth for what a product actually is, and a feed platform or a thin custom connector handles the channel-specific transformation on top of it. That split keeps one place responsible for the product data, while the feed layer stays free to change per channel without anyone touching the source.
Write Product Titles People Search For
A working title formula, roughly in this order, weighted by what actually distinguishes the product in your category:
Brand, product type, the one thing that sets it apart, the specification a buyer filters on, size or capacity, color where it matters.
| Weak title | Working title |
|---|---|
| Green Tent Model X | AlpineTrail 2-Person Backpacking Tent, 3-Season, 4.5 lb, Green |
| Coffee Maker 12 Cup | BrewCraft 12-Cup Programmable Coffee Maker, Thermal Carafe |
| Dog Bed Large | PawsRest Orthopedic Dog Bed, Large, Washable Cover |
Do not apply that formula the same way across every category. Dimensions matter most for furniture, compatibility for parts and accessories, material and care for apparel, and technical specifications for electronics. Pull the top five searched terms in your category from Search Console or your ads account before you template anything.
Custom labels are the part most feed audits skip. Adding one costs no ad-platform fee, only the work of getting the underlying attribute into your product data in the first place. They do not improve a listing’s eligibility or relevance on their own. Their value is campaign control: grouping products so bids and budgets reflect how the business actually makes money, not just how the catalog is organized.
| Custom label | Example values | What it lets you do |
|---|---|---|
| Margin band | High, medium, low | Bid on profit instead of revenue alone |
| Inventory status | Overstock, normal, low stock | Pull back on a listing before it oversells |
| Seasonality | Year-round, spring, holiday | Shift budget by calendar without rebuilding campaigns |
| Price band | Under $25, $25–$75, $75+ | Structure bids and creative by price tier |
| Lifecycle stage | New, core, clearance | Separate merchandising decisions by stage |
| Return-risk band | Low, medium, high | Avoid treating a high-return category as equally profitable to one that rarely comes back |
| Shipping-cost band | Free, subsidized, high cost | Account for fulfillment cost in what you are willing to pay per click |
| MAP status | Restricted, flexible | Keep price-protected products out of campaigns that would violate the agreement |
Common Issues in the Feed Audits We Run
Treating feeds, schema, and the API as either-or choices. Schema does not replace a feed. The API does not remove the need for schema. Each does a different part of the job.
A feed set up once and never revisited. A platform’s default export is functional and rarely optimized, and feed quality drifts as the catalog changes underneath it.
Manual spreadsheet updates past the point they scale. They work for a small, single-channel catalog. Past a few hundred SKUs or more than one sales channel, changes start slipping through.
Schema that stops matching the page. A price change on the site with no matching update to markup creates a mismatch Google and AI tools can both detect: the markup says one price, the page shows another.
Assuming the platform’s default feed is good enough. It runs. It is rarely tuned to the searches your buyers actually type.
Your Next 30 Days
Do not start by buying a platform or scoping a Merchant API migration you do not need yet. Start by finding the issue with the largest revenue impact.
- Export Merchant Center diagnostics and list the issues affecting the most products, ranked by how often they recur.
- Review your top 50 revenue SKUs for title quality, identifiers, category mapping, price accuracy, and image coverage.
- Spot-check structured data on a sample of product pages against what the page actually shows a visitor.
- Confirm whether your team calls the Content API directly. If it does, check migration status today.
- Decide whether your real constraint is feed quality, schema coverage, fetch frequency, or channel complexity, and let that decide the next project, not the order this guide happened to list them in.
If you want a second set of eyes on a Merchant Center account or a feed that has drifted, that is the kind of audit we run.