Product data feed before the season: Google Shopping, marketplaces and the typical sources of error

A product data feed isn't just a chore for the marketing department; it's the file through which a significant portion of your revenue flows in the fourth quarter. If an attribute is missing, your product won't just be shown less effectively—it won't be shown at all. This article outlines the six most common errors we encounter and the specific cost of each one.

The previous post about hosting under Last.org discussed whether your shop can handle the influx of visitors. Today, we'll address the earlier question: whether the visitors will even come at all. Because a large portion of Christmas traffic doesn't originate in your shop, but rather on Google Shopping, Amazon , eBay, or price comparison sites—and what appears there is determined by your feed.

The annoying thing about this file is that it either works or it doesn't. A rejected product doesn't get a prominent error message. It simply disappears from the results, and you only notice it because of a number that nobody analyzes in November.

Why this becomes particularly important in August

Because feed problems take time—not to fix, but to have an effect. If you add a missing attribute today, it can take hours or even days, depending on the platform, for the product to be displayed again. If you make a change in November, you lose the most valuable days of the year stuck in a waiting loop.

In addition, there's the market trend: According to an analysis by the German E-Commerce and Distance Selling Association (bevh), online marketplaces are the fastest-growing segment of German online retail. The figures are available in the bevh's analysis of e-commerce growth . Those who aren't properly listed there are missing out on the fastest-growing channel.

The six sources of error — and what they cost

1. Missing or incorrect GTIN

The classic, and also the most expensive. The GTIN — formerly known as EAN — is the globally unique product number. Without it, no platform can match your product to an existing catalog entry.

The consequences vary depending on the channel: On Google Shopping, visibility decreases because your offer cannot be included in the product comparison. On Amazon and eBay, the offer is not even accepted in many categories. Anyone wanting to know how the numbering system works and where valid numbers come from can find the basics at GS1 Germany on the GTIN as a product identifier —the organization that assigns these numbers in Germany.

2. Prices that don't match the shop's offerings

If the price in the feed differs from the price on the landing page, the product is rejected. This sounds trivial, but it happens constantly—and for a systematic reason: The feed is generated once a day, but prices change continuously, especially during discount promotions.

This is precisely where the Black Friday trap lies: You lower prices at midnight, but the feed doesn't update until 6 a.m. During these six hours, the feed and the shop's prices diverge, and the platform rejects the offer. Therefore, check now how frequently your feed is generated—and whether the frequency can be increased for promotional days.

3. Shipping costs and delivery times are missing or incorrect.

Both belong in the feed, not just on the product page. If the information is missing, the platform uses default values—and your offer will appear more expensive or slower than it actually is. In a price comparison where three providers are within two euros of each other, this is precisely what matters.

A particularly common overlooked feature is shipping cost tiers based on weight or zone. Entering a flat rate here results in lost visibility for lighter items and higher costs for heavier ones.

4. Images that do not meet the requirements

Too small, with a watermark, with a logo, with text in the image, or with a frame—any of these characteristics can lead to rejection. Google describes the requirements for images and all other attributes in the Merchant Center product data specification.

The mistake almost no one notices: If you include Christmas banners in your product images in December—such as "Only until December 20.12th!" or a discount ad—you're violating the image guidelines. The products will then be removed at the exact moment they're supposed to sell. Promotional graphics belong on the page, not in the product image.

5. Categories and product types are poorly maintained

Two distinct fields that are often confused: The Google product category comes from a fixed list provided by the platform, while the product type is your own custom structure. Both should be filled in—the category determines the competitive landscape in which your product appears, and the product type helps you manage your campaigns.

Anyone who lumps all their items into a single category is competing with the platform's entire product range instead of with the relevant niche. This isn't an error that the system will report—it simply costs money silently.

6. Inventories that are updated too late

Availability is one of the attributes that becomes obsolete the fastest. If a sold-out item continues to be advertised, you pay for clicks on a disappointing product—and risk account restrictions if this happens repeatedly.

The opposite scenario is more expensive and less frequently noticed: An item is listed as "unavailable" in the feed, even though the stock has long since arrived. It isn't displayed, incurs no costs , and doesn't generate an error message. It's simply invisible. Be sure to explicitly check for this scenario.

The title is the most important attribute — and the worst maintained.

E-Commerce News - Tips & Tricks 🤖Is your product data ready for AI agents?🛒

If I could only improve one thing about a feed, it would be the title. It determines visibility more than any other field because the platform uses it to infer which search queries your product matches.

The typical shop title is "Winter jacket model Anna". That's a name, not a search term. Nobody searches for that. People search for "Women's lined waterproof winter jacket, black, size 38" — and anyone who doesn't include those terms in their title won't be found.

A reliable search order for most product ranges is: brand, product type, key feature, color, size. So not "Anna — Jacket", but "Nordwind Women's Winter Jacket, lined, waterproof, black, size 38". It reads more awkwardly, but it will be found.

There are two limitations you should be aware of: Platforms shorten long titles in the ad—the first 60 to 70 characters are what count. And advertising language doesn't belong in it. "Bestseller," "Buy Now," or "% Discount" in the title will lead to rejection by most providers.

The test, which lasts three minutes

Take five of your most important products and read only their feed titles. Would you search for them yourself? If not—and in most shops the answer is no—you've found the most effective task of the summer. Titles can usually be generated using rules in the feed tool, from fields that are already populated: brand plus category plus attribute plus variant.

What “high-quality data” actually means

There are mandatory attributes and recommended attributes. The mandatory fields determine whether your product even runs at all. The recommended ones determine how often it appears in relevant search results—and that's precisely where the leverage lies that most people overlook.

Color, size, material, condition, age group, gender: Every additional attribute is another way to be found. Someone searching for "women's red winter jacket, size 38" in December will only find listings where this information is included in the feed.

How closely these requirements are related to machine-readable data in general was described in our article on structured and consistent product data — the same fields that carry the feed also feed the responses of AI systems.

The feed check: seven points for an afternoon

This is the short version that allows you to assess the health of your feed in a few hours — without additional software.

First: How many items are active in the shop, and how many are in the feed? If the number differs, you have your first task. Second: How many items are currently rejected, and why? The platform's diagnostic overview provides the reason and number. Third: How many items are missing their GTIN? Fourth: Do the price and availability match between the feed and the product page in a random sample of five items? Fifth: How frequently is the feed generated? Sixth: Are shipping costs specified, and if so, are they tiered? Seventh: How many items have more than just the required attributes?

Marketplaces follow different rules.

What applies to Google doesn't automatically apply to Amazon or eBay. The requirements for article numbers are stricter there, the category logic is different, and titles are subject to their own length and format specifications.

Practical advice: Maintain a clean database in your ERP or PIM system and generate channel-specific feeds from it—instead of maintaining separate data for each channel. As soon as the same information is manually maintained in three different places, it will diverge. This isn't a matter of discipline, but of time. We've described the role of inventory management in this context in more detail in our section on ERP and PIM integration.

Anyone considering marketplaces for the first time will find the fundamental considerations in our article on multi-channel distribution and marketplace integration.

A feed is a process, not a project.

The difference between shops that run smoothly in December and those that constantly experience outages rarely lies in skill. It lies in whether someone is checking regularly.

Three habits are all it takes. First: Check the diagnostic overview once a week and note the number of rejected items. Don't analyze it—just note it. An increase from twelve to two hundred will be immediately noticeable. Second: After every major product import, compare a sample of five items between the shop and the feed. Third: Before every discount campaign, check when the feed will next be generated.

This takes less than half an hour per week and prevents the vast majority of cases where someone discovers in December that a third of their product range has not been available for three weeks.

Google's guide to high-quality product data serves as a benchmark for what "well-maintained" specifically means — it describes not only the mandatory fields, but also how a platform distinguishes good datasets from mediocre ones.

Who is responsible?

A question that often goes unanswered in smaller teams because the feed is technically tied to the shop, content-wise to the product range, and financially to the marketing department : Designate one person to report the number of rejected items weekly. Not the person who solves everything—just the one who notices the problem. That's usually enough.

Conclusion: The feed is a distribution channel, not an export channel.

Most shops treat the product data feed as a technical minor detail—a file that simply gets generated. In reality, it's the interface to the channels through which a significant portion of demand flows in the fourth quarter.

The good news: This work only needs to be done once. Cleaning up the data in August will benefit every channel and every campaign —and also wherever AI systems analyze product information. The bad news: This work can't be made up for in the week before Black Friday.

Where are you stuck?

How many of your articles are currently rejected — and do you even know that number? It's the one metric that almost no one ever looks at and that can be found in five minutes.

And the question that I expect to elicit the most interesting answers to: Which reason for rejection surprised you the most? For us, it was a customer whose entire product range was removed in December because someone, with good intentions, had photoshopped poinsettias into the product images. Write it in the comments—cases like that aren't covered in any documentation.

If you'd like to know how your product data is performing technically: Our free visibility check analyzes any website without registration. And if you'd prefer to review the feed together, you can reach us via the contact page.

This post is part of our Q4 series. Every Tuesday, a new post will be published here to help you prepare for the peak season. Next week: Practical load testing — how to find the limit before the customer does.