Products disapproved in Merchant Center: the causes that matter

You open Merchant Center and there is a number next to "products not approved". You click, and Google tells you what is wrong in a short phrase: a missing value, a value that does not match, an image that is not allowed. What it does not tell you is what that is costing you, and more importantly, that the disapproved product is not your worst problem.

Disapproval is the visible symptom. The damage is wider

A disapproved product is out: it does not show in Shopping ads or free listings, and that is obvious enough to end up on a list. It is annoying, and it is also the easy part, because at least you know where to look.

The expensive part is different. Among the products that are approved, a great many are shown rarely or never, because the data behind them is not enough to tell Google what they are. No warning, no list, no red number. They are simply not seen, and there is no way to notice from the dashboard.

The causes, in the order we actually meet them

Missing or wrong identifiers. Barcode, brand, manufacturer part number. These are the fields that tell Google which product this is, not what it looks like. Without them the product is not matched against what others sell, and in categories where comparison decides the sale it ends up at the edges. It is the most common cause, and the dullest: the codes almost always exist, they are just in another column of the back office.

Price or availability that does not match the page. Google does not take the feed's word for it: it reads the page and compares. If the feed says 49 and the page says 59, or the feed says in stock and the page says sold out, the product drops. This is almost never a mistake, it is timing: a promotion starts on the site and the feed catches up hours later. The more often you change prices, the more often this happens.

Images with overlays. The "-30%" badge, the "New" ribbon, the frame with your logo, the placeholder somebody used while waiting for the real photo. All of them make the image ineligible. Usually nobody added them on purpose: they come from the theme, or from a promotions app that stamps them on afterwards.

Wrong or restricted category. Some categories carry their own rules, and some require attributes that are irrelevant elsewhere. A badly classified product can land in a category whose requirements it does not meet, and get disapproved for a reason that looks absurd until you look at the classification.

Misrepresentation. The only one on this list that is not about a product but about the account, which is why it is the only one worth being afraid of. It is about trust: contact details, return terms, clarity on who is selling. When it triggers you do not lose a product, you lose the channel.

Why a large catalogue cannot be fixed by hand

A hundred products is a morning's work. The problem is that the hundred products fixed by hand are almost always the same hundred: the bestsellers, the ones somebody looks at. They are also the ones that already work.

The rest of the catalogue stays as the supplier sent it, and not because nobody thought of it: because the arithmetic does not work. If fixing one page costs twenty minutes of a person's time, four thousand products is a thousand hours. Nobody has them. So the tail of the catalogue, which is where the demand no competitor has claimed sits, stays untouched forever.

That is what changes when enrichment costs close to nothing per product: you do not fix the first hundred better, you also fix the other three thousand nine hundred.

What happens when the data is right

On a multi-brand fashion boutique with thousands of items we rewrote the product pages from catalogue data, attribute by attribute, and rebuilt the feed so that the shopping listing and the product page said the same thing. Over the following five months, against the same months a year earlier, Google Shopping impressions grew 133.3% and clicks grew 302.3%.

That is not only a visibility number: clicks grew more than twice as fast as impressions, which means the listings were not just shown more, they were chosen more often. The full case, with the month-by-month chart.

Where to start tomorrow morning

Before touching anything, two checks that cost ten minutes and change the order of your priorities.

One: how many products you have in Merchant Center against how many you have in Shopify. If the first number is much lower, disapprovals are not your problem: part of the catalogue is not even getting as far as being judged.

Two: how many of the approved products got at least one impression last month. The ones at zero are the silent tail: technically fine, practically invisible. It is usually the largest slice, and the one nobody is working on.

Frequently asked questions

How long does Google take to re-approve a product after a fix?

It depends on when the feed is read again and on the kind of problem. Fixes to feed data show up at the next update; anything needing a review takes longer. The more useful thing to know is this: fixing the single product puts it back in play, but if the cause is in how the feed is generated, it will come back.

Do all products need a barcode?

Not all. Handmade goods, made-to-order items and own brands often do not have one, and there are correct ways to declare that. The problem is not lacking the code: it is leaving the field empty without saying why, because then Google has to guess.

Should I remove products that never get shown from the feed?

That is the instinctive reaction, and it is almost always the wrong one. A product that is not shown is usually not a bad product, it is a badly described one. Removing it deletes the symptom and gives up the demand. Work out whether the cause is the data first, because if it is, it gets solved once for the whole catalogue.

Does one disapproved product hurt the others?

A single disapproved product is about that product. Account-level problems such as misrepresentation are different: they are about the whole shop. Which is why it helps to keep the two apart in your head. One is maintenance, the other is a risk to the channel.

Does fixing the feed also help with being cited by AI?

It helps, but it is not the same job. A clean feed is structured data: identifiers, price, availability, category. An answer engine quotes what it finds stated plainly on the page, so measurements, materials, compatibility, what is included. The two feed each other, because they come from the same catalogue data, but a clean feed with thin product pages covers only half the work.