---
title: "Your product data was written for a page, not for a question"
url: "https://mercantir.com/blog/product-data-written-for-a-page"
description: "Catalogue copy is optimised to be scrolled past. Ask it something specific and there is nothing structured underneath to answer from.
"
---

# Your product data was written for a page, not for a question

August 19, 2026·2 min read·Mercantir

Open any product page. There is a photograph, a paragraph of description, a specification table and some marketing copy. All of it is written to be looked at by a person deciding whether to scroll further.

Now ask a question. Will this fit that. Is it suitable for outdoor use. What is the actual difference between these two models. In most catalogues nothing structured exists to answer from, and the answer only appears because a person reads the description and infers it.

## The gap between description and knowledge

**Attributes are inconsistent.** The same property is called different things across categories, or lives in free text for some products and in a field for others, because it was entered by different people over several years.

**Relationships are missing.** What this fits, what replaces it, what it needs to work, what it is an upgrade from. This is the most valuable information in a catalogue and it is usually only in the heads of the people who work in the shop.

**Context is absent.** What a product is actually for. Two items with identical specifications can be right for different jobs, and the person who knows that is on the counter, not in the database.

**Negative information is nowhere.** What a product will not do, what it is unsuitable for, what people return it for. Commercially uncomfortable and enormously useful, because half of buying decisions are eliminations.

## Why this matters more than it did

For a person browsing, the description was enough — they filled the gaps with judgement and asked in the shop when it mattered.

Anything answering on their behalf has no judgement to fill gaps with. A search that ranks, a support tool that answers, a model summarising your catalogue — all of them work from what is structured, and where nothing is structured they infer from marketing copy, which is written to persuade rather than to describe.

The failure mode is specific and bad: a confident answer assembled from adjectives. It sounds right, it reaches a customer, and the return comes back three weeks later.

## What structured product knowledge looks like

**Consistent attributes with meaning.** Not just a specification table but typed, comparable values, so that larger, compatible and suitable-for are computable rather than adjectives.

**Explicit relationships.** Fits, replaces, requires, upgrades-from. Written down rather than implied by category.

**Use context.** What this is for, in the words customers actually use rather than the words the manufacturer uses.

**Stated limits.** What it does not do. The single most useful field nobody has, and the one that prevents the most returns.

## The unglamorous conclusion

This is data entry, and there is no version where it is not. Some of it can be extracted from existing descriptions and supplier data, and a lot of it exists only in the heads of experienced staff, which means capturing it is a process rather than an import.

That is why it is worth starting before the channel arrives. A retailer with structured product knowledge has better search, fewer returns and better support today, and is ready when something starts asking questions on a customer’s behalf. A retailer without it will be asked to produce it in a hurry, which is the expensive way to do anything.

-   product data
-   catalogue
-   structured data
-   search

## Keep reading

-   [One question, four systems](https://mercantir.com/blog/one-question-four-systems)
-   [What agentic commerce actually needs before it can work](https://mercantir.com/blog/what-agentic-commerce-actually-needs)