You have good products, but some don't appear in ads, others have unsuitable images, and the price shown isn't always the one on the site. Before you change the budget, check the information sent to Merchant Center. The feed is the link between the store's catalogue and how the products can be presented in Google.
A good feed describes each product correctly and stays synchronised with the landing page. Check identifiers, titles, variants, images, prices, availability and the requirements that apply to the category. Correct errors in the data source and track eligibility after the update. Feed quality helps relevance and presentation, but doesn't guarantee profitable sales on its own.
Contents
- The role of the feed
- Identifiers and variants
- Titles and descriptions
- Images, price and stock
- Diagnosis and priorities
- Calculation examples
- Maintenance process
- FAQ and checklist
What the feed is and why its source matters
The feed is a structured source of information about products. It can be supplied through an integration, a file or another mechanism accepted by the platform. Before you choose the tool, decide where the price, stock and characteristics come from and who is responsible for their correctness.
Google explains the role of product data in Shopping. For attributes and conditions, use the Merchant Center specification. Requirements can depend on country, category and product type; a generic list doesn't replace checking your own case.
Don't treat the export as a neutral copy
The internal catalogue can contain abbreviations, supplier codes or incomplete fields. These may be enough for inventory management but unclear to the buyer. The transformation of the data must keep the truth about the product and improve legibility, without introducing assumed characteristics.
Document the transformation rules. If a title is built from brand, model and size, check what happens when a field is missing. A rule applied to thousands of products can quickly multiply a small mistake.
Identifiers and variants are the basis of consistency
Keep the product's identity stable
An internal identifier must allow the same item to be tracked between updates. Changing it without reason can complicate reconciliation and history. Before a new migration or integration, document the mapping between old and new values.
Commercial identifiers such as GTIN or MPN must not be invented. Check the supplier's information and the requirements that apply. If the product doesn't have a certain identifier, follow the official rules for that situation, instead of filling in an arbitrary value to apparently eliminate an error.
The variant in the ad must be the one on the page
Colour, size, material and configuration can change the price and availability. Open the link sent for each representative type of variant and check the actual selection. An ad for the blue variant that opens the black variant by default creates a mismatch even if the general page is correct.
Also test unavailable variants. Don't let a promoted product look purchasable if the option shown is no longer in stock. If the store allows special orders or pre-orders, the information must be presented consistently and in line with the relevant requirements.
Titles and descriptions must help the choice
Write the title in the order of useful information
A good title allows the product to be recognised. Depending on the category, the brand, type, model, size or compatibility can be decisive. There is no universal order for all products. Start from how the customer tells alternatives apart.
For example, an internal name such as “Model X premium” can hide the fact that the product is a filter compatible with a certain appliance. A descriptive wording can clarify this information, if it is verified. Don't add “the best”, discounts or non-permitted commercial claims in fields meant for identifying the product.
The description must be specific
Explain the material, the use, the dimensions and the relevant limits. Don't fill the description with a list of search phrases. A text that repeats the same category in ten variations can be less useful than a few precise sentences about what the buyer receives.
When you use information from the supplier, check that it applies exactly to the variant sold. Don't turn a characteristic of a range into a promise for all models. Keep an approval process for the fields that can create important commercial expectations.
Tie optimisation to a hypothesis
If you change titles, note the problem you are after: model ambiguity, a missing dimension or confusion between variants. Evaluate eligibility, relevant traffic and commercial results. An increase in clicks doesn't prove on its own that the change improved the result.
For the store's Google Ads campaigns, the feed and the campaign structure must be analysed together. Still, keep the difference between a data error, a commercial choice and the cost of distribution. Each calls for a different kind of intervention.
Images, price and stock must tell the same story
Check the image of the real product
The photo must represent the item and the variant offered. Check the resolution, the cropping and any elements that violate the applicable requirements. A spectacular image that is ambiguous about the package contents can attract unsuitable clicks.
If the photo shows accessories that aren't included, the page must clarify the contents, and the main image must be assessed carefully. Don't assume the user will later read a small note to correct a wrong impression created by the ad.
Synchronise commercial changes
A promotion has a start and an end. Check the moment of the update in the store, in the product source and in Merchant Center. Don't schedule the ad separately from the price and then assume the systems will be aligned to exactly the same second.
Likewise, stock must be updated at a pace compatible with the speed of sales. For products that sell out quickly, infrequent synchronisation can generate mismatches. Check the real mechanism, the update log and any integration errors.
Fix it in the right source
A manual correction in an intermediate system can be overwritten by the next import. Identify the owner of each field and the place where the change must be made. Keep a copy of the rules and test a sample before applying to the whole catalogue.
If the integration starts from Shopify, store development and maintenance must be coordinated with ad management. The catalogue can also feed Meta campaigns, but each platform's requirements must be verified separately.
How to prioritise errors without getting lost in the catalogue
| Situation | What to check | Practical priority |
|---|---|---|
| Product rejected | The exact reason and the attribute affected | Fix before promoting |
| Different price | The source, the moment of synchronisation, the variant | High, affects the offer |
| Unsuitable image | The product and the real contents of the package | High for the promoted selection |
| Vague title | The information needed to choose | Optimise after correctness |
| Wrong stock | The link with inventory and the frequency | High for active products |
| Complete data, weak sales | The offer, demand, page, costs | Separate commercial diagnosis |
Start with the problems that prevent display or create an incorrect offer. Then work on clarity and relevance. Don't give the same urgency to a missing secondary attribute and to a wrong price for the campaign's main product.
Distinguish the number of errors from the impact. A problem affecting ten important products can matter more than one affecting hundreds of items without stock or without commercial priority. Document the criterion, so it doesn't look like an arbitrary choice.
Three hypothetical examples for analysing the feed
Hypothetical example 1: catalogue coverage
You have 800 variants you intend to promote. Of these, 680 are eligible at the time of checking. The coverage of the selection is 680 / 800 × 100 = 85%. The denominator is the planned selection, not the company's whole catalogue.
The remaining 120 variants must be grouped by cause and commercial value. The percentage doesn't automatically estimate lost revenue. For that you would need demand, stock, contribution and a credible scenario about their performance.
Hypothetical example 2: clicks versus orders
In a 30-day test, before a change you have 1,000 sessions and 25 sessions with a retained order each: 2.5%. In a later period you have 1,200 sessions and 24 such orders: 2%.
More traffic didn't mean more orders in this example. We can't attribute the difference solely to the titles without a controlled test and checking the other conditions. Analyse relevance and mix, not just the growth in volume.
Hypothetical example 3: economic priority
Group A has 30 retained orders, an average contribution before media of 90 RON and media of 1,800 RON. That leaves 30 × 90 − 1,800 = 900 RON before fixed costs. Group B has 40 orders, a contribution of 45 RON and the same media: nothing remains.
If both have data problems, prioritisation can also take this economics into account, together with stock and demand. It doesn't mean B must be excluded automatically, but that order volume is not the only criterion.
A maintenance process that prevents errors from coming back
Set an owner for the source and one for checking in the platform. Define what happens after a price change, a product import or an integration update. A feed is not a deliverable done once and forgotten.
At every important change, test a representative sample and track processing. Keep the list of recurring errors and their cause. If the same attribute breaks every month, the problem is probably in the process or integration, not in a lack of one more manual correction.
In the first week, take stock of the sources and the selection. In the second, fix the eligibility blockages and commercial mismatches. In the third, test the clarity of titles and images. In the fourth, evaluate mature results and document the rules worth keeping.
For a coordinated analysis of the catalogue and campaigns, talk to BlueAds. Prepare examples of affected products and the history of changes; these help more than the general claim that Shopping isn't working.
Frequently asked questions
Does a good feed guarantee sales?
No. It supports the correctness and relevance of the presentation. The offer, demand, competition, page and costs remain important. Eligibility is a useful condition, not a commercial promise.
Can I make up a GTIN if I can't find one?
No. Check the manufacturer's data and the rules that apply to a product without an identifier. An arbitrary value doesn't solve the underlying problem and can create incorrect information.
Do I have to change all the titles at once?
Not necessarily. You can start with a coherent selection and a clear hypothesis. Keep the history and track what changes, without automatically attributing every variation to the change you made.
Why does the error come back after I correct it?
An import or a rule may be overwriting the change. Identify the field's main source and correct the process, then check the next synchronisation.
Can one feed serve several platforms?
The commercial source can be shared, but the mappings and requirements differ. Check each destination separately and don't assume that acceptance in one system confirms compatibility with all the others.
How often do I update stock?
At a pace compatible with the dynamics of the products and the mechanism available. Check the actual delays and errors. There is no universal frequency that suffices for any catalogue and sales speed.
What do I check first when products don't appear?
Eligibility, the reasons for rejection, the data source and inclusion in the campaign's selection. Only then analyse the budget and distribution. A catalogue error isn't corrected by bidding.
Final checklist
- Identifiers are real and consistent.
- Variants open the right pages.
- Titles describe the product without exaggeration.
- Images represent the real offer.
- Price and stock are synchronised.
- Errors are fixed in the right source.
- Priorities include commercial impact.
- There is an owner and a log of changes.