How to write product descriptions when your supplier data is poor

If you sell products you did not make, your descriptions probably start as someone else's text. Suppliers and distributors send fragments, spec dumps or marketing copy written for a catalogue, not for a shop. Pasting it in unchanged leaves you with thin pages that look like everyone else's.

This guide shows a practical order of work: spot weak data, separate facts from filler, decide which source wins, organise the facts, and only then write. The same steps work by hand, in a spreadsheet, or with an AI tool.

1. Spot weak supplier data

Before you rewrite anything, check what you actually have. These are the usual warning signs:

• Fragments instead of sentences, such as a string of specs with no context.
• Mixed or missing units: inches in one product, centimetres in the next, or numbers with no unit at all.
• The same text on many other shops. Search a distinctive sentence from the description in quotation marks. If many other shops show up, your page offers nothing of its own.
• Claims without facts, such as "premium quality" or "best in class", with no material, size or test behind them.
• Contradictions between title, description and variants, for example a different model number or colour name.
• Awkward machine-translated phrases and leftover codes from the supplier's system.

Write down which of these show up in your catalogue and how often. It tells you how much work each product needs.

2. Separate facts from filler

A fact is specific and can be checked: a dimension, a material, a power rating, what is in the box, which devices it works with. Filler is everything else: adjectives, promises and slogans.

Copy the supplier text into a sheet with two columns, facts on the left and filler on the right. Keep the facts and drop the filler. The example below shows the idea.

Every statement in the rewritten text comes from the supplier line. Nothing was added, such as room size or sound character, because the data does not say so.

3. Decide which source wins

Suppliers are rarely your only source. When two sources disagree, you need a rule for which one counts, otherwise every product turns into a debate.

A simple order of trust is the product itself first, then the manufacturer, then the supplier. Add a "source" column to your sheet so you can see where each fact came from and re-check it when the supplier updates their feed.

4. Put the facts in order before you write

Writing goes faster, and stays consistent, when the facts already sit in a table. Decide on a fixed set of attributes per product type. For a speaker that could be driver size, impedance, power range, finish and what is included. For clothing it could be material, fit, sizes and care.

Use the same units and spelling everywhere. If your theme supports it, store the structured facts in product metafields so they can be shown as a specification table and reused. If not, a spreadsheet and your product CSV export are enough.

Then fix the shape of a description for each type: one opening sentence that says what the product is, a short list of facts, and what is included. A fixed shape makes hundreds of descriptions read like one shop.

5. When AI helps, and what to check by hand

AI is good at turning a fact table into readable sentences in a consistent voice, across hundreds of products. It is not a source of facts. Unless you give it the data, it fills gaps with plausible details, and plausible is not the same as correct.

Whatever tool you use, check these by hand:

• Every number and unit against the data sheet
• Model names, sizes and colours
• Compatibility statements such as "works with ..."
• Anything about safety, materials, certifications or health
• Any sentence with a detail that was not in your fact table

Start by checking every number in the first ten products. That shows you which kinds of mistakes your tool makes. After that, check a sample from each product type instead of reading everything.

A short checklist

• Mark weak data: fragments, mixed units, duplicated text
• Split facts from filler and keep only the facts
• Pick one source of truth per fact and note it
• Fix attributes and units per product type
• Write from the fact table, then check the numbers by hand