Four readers look at every product you sell. Most stores write for one.

The record travels, the page does not

Your product page stays where you built it. Your product record leaves the building. It is compiled into a feed for Google, pushed into your social catalogues, indexed by the search bar on your own store, and read by the assistants people now shop with. Four readers, one record, and each of them needs different parts of it filled in.

Notice what the four have in common. Not one of them reads your layout, your fonts or your photography. They read fields. So a catalogue can look immaculate on the storefront and still be unsellable everywhere else, which is exactly the failure that never shows up in a report: no error, no warning, just a competitor shown instead of you.

You merchandise the page and inherit the record. Only one of them leaves the store.

What a record that sells actually holds

A complete record is not a long one. It is about twenty fields, and they fall into four jobs: identify the product, describe it, classify it, and state its commercial facts. Each job has a reader waiting for it, and each blank has a consequence. ◆◆WHAT A BLANK SAYSAn empty field is not neutral. Software reads it as a fact about the product: not available, not applicable, not worth showing.The rule: state it, or it isn't true

Identity is the one layer we have already covered in full: what a GTIN is, why the Shopify barcode field is the whole ball game, and the three ways stores get it wrong. If your identifiers are thin, start in the UCP guide and come back. The rest of this piece is the three layers after it, which is where the sales are actually leaking.

The title, written for the wrong reader

The title is the single highest-leverage field in the record, because every reader uses it and most of them see nothing else first. It is also the field most likely to have been written by a marketer for a banner, or by a supplier for a warehouse. Both fail. The test is simple: could a stranger describe this product out loud using your title, and would software recognise the same product from it?

The formula underneath is dull and it works: brand, then product name, then what the thing is, then the details that separate this variant from its siblings. Front-load it, because titles get truncated by every reader at a different length. Then leave out the three things that cost you matches: promotional words (“new in”, “sale”, “free delivery”), all caps, and internal codes. Google will disapprove products for the first two. An assistant will simply not match them.

The attributes nobody fills in

Attributes are the fields that answer “which one”: colour, size, material, pattern, gender, age group, capacity, fit. They feel like admin, which is why they sit empty. They are in fact the difference between matching a broad request and a specific one, and specific requests are the ones with a buyer attached.

Two of these are not optional and stores routinely miss it. If you sell apparel or accessories, Google requires colour, size, gender and age group on each variant, and products missing them are held back from Shopping and free listings rather than shown with a gap. The others are technically optional, which is a trap: optional to submit, decisive to match. “Waterproof canvas”, “relaxed fit”, “1.5 litre” and “stainless steel” are the words in the question, so they need to be words in the record.

Category is the attribute that sets the terms for the rest. Your store’s own product type or taxonomy value tells every reader which family the product belongs to, and it decides which further attributes are even expected of it. File a hiking boot under “footwear” and you have told the truth and nothing useful. File it under men’s hiking boots and the whole vocabulary of that category opens up.

Images: the field with two audiences

Images are the one part of the record that both a person and a machine read, and they read it completely differently. A person judges the product. A machine reads the file: whether there is a main image at all, whether it clears the minimum size, whether the variant that a customer picked has its own picture, and what the alt text says the picture shows.

Three habits cover most of it. Give every variant its own image, or a colour filter becomes a wall of the same photo. Write alt text that states what is in the frame rather than repeating the title, because a size chart or spec sheet that exists only as a designed image is invisible to every reader but the human one. And keep promotional overlays off the main image: badges, prices and “sale” flashes get products disapproved in feeds, and they date the moment the promotion ends.

Freshness, the field that goes stale on its own

The other three leaks are things you never wrote. This one is different: it is a field that was correct when you loaded it and quietly stopped being true. Price and availability are the two that matter, and they are also the two every reader checks hardest, because sending a shopper to a sold-out product is the worst outcome any of them can produce.

The punishment is not a message, it is a slow demotion. A feed with a price that does not match your product page gets the product held back. An assistant weighing two stores with the same product leans on availability to break the tie. So treat freshness as plumbing, not admin: your platform should push stock and price changes to your feeds automatically, and if it does not, that is the thing to fix before you fill in another field by hand.

Your search bar is running the same test

Here is the part that makes this work pay for itself before any feed does. The search bar on your own store reads the same fields. So do your filters. When a customer types “black leather boots size 8” into your store and gets nothing, that is not a search problem, it is four empty attributes returning an honest answer. ◆◆THE READER YOU OWNYour own search bar runs on the same product fields as the feeds. Fix the record once and the store gets easier to shop as a side effect.The rule: one record, every reader

Which means you already own the most direct audit there is. Your zero-result and low-result search reports are a list of things customers asked for in their own words and your catalogue could not answer. Read them as field requests, not as search misses. The same goes for your filters: a filter that returns almost nothing is an attribute that is mostly blank, and a filter customers use heavily is an attribute worth completing across the whole range.

An afternoon, a spreadsheet, and three reports

You do not need a tool to find out where you stand. Export your catalogue to a spreadsheet and count the blanks in each column. That number, your fill rate per field, turns a vague worry into a work list ordered by how many products it affects.

Then work in the order the readers care about, not the order the spreadsheet is in: identifiers first, then titles, then the required attributes for your category, then the optional ones your own search reports asked for, then images, and freshness set up as automatic rather than remembered. On a small range that is a couple of days of honest work. Across thousands of variants, kept accurate as the range turns over, it stops being a task and becomes an operation, which is the point at which it belongs to someone permanently. That is the wall we hit on our own store, so we built the thing that holds it.

One record, paid for four times

Most growth work has a single beneficiary. Spend on ads and you get traffic while the budget lasts. Fix the record and four things improve at once: your paid shopping campaigns get more products eligible and better matched, your free listings and organic search pick up the same specificity, your own store becomes easier to shop because the filters finally work, and the assistants people now buy through can put you forward with confidence instead of hedging.

That last reader is the newest and the reason the old chore has become urgent. For years, clean product data was something you did for Google Shopping and could put off. Now it decides whether an agent can sell your products at all, which is the argument we made at length in how AI agents shop your Shopify store and shipped against on this site in what we made readable by AI. Both of those guides end up in the same place: the record.