Episode 11: Five Ways to Stop Losing Sales to Bad Product Data

Product data is the most important pillar of any ecommerce strategy, and almost nobody treats it that way. With AI search and product discovery moving as fast as they are, it matters more every month.

The losses are quiet. Nobody files a ticket saying the specs were inconsistent so they bought elsewhere. The sale just does not happen.

So here are five places we watch sales leak out of businesses, and what to do about each. All five can be tested this week. None need a platform project to start.

Three layers of failure

Whatever the size of the catalogue, product data fails in three ways, and most businesses have all three at once.

It is broken: missing, mixed up, or out of date. It is invisible, not available to channels, to SEO or to generative engines. And it is not built for AI, so it does not match how buyers now search.

The numbers are stark. Around 90% of buyers abandon a purchase when product information is missing or wrong. Customers are three times more likely to buy from a competitor with better structured data. And AI tools skip your product entirely if there is nothing structured to read. That holds in B2B as much as B2C.

It applies to vast catalogues and short ones alike.

1. Your specs do not match across channels

Picture a procurement manager doing final checks before raising a purchase order. Your website is open, alongside a PDF and a distributor portal. The weight differs on all three. They check a second supplier selling the same product and it is consistent everywhere.

They will not email to ask which figure is right. They will quietly decide whether they trust you.

These inconsistencies are almost never somebody failing at their job. Something was updated in one place and never followed through. That is a process failure, not a people failure.

The fix does not start with a PIM project and twelve months of board sign-off. It starts with a list of every channel your product data lives on. Pick five products, open every tab, record the differences. Then build a master record, even if it begins as a Google Sheet, with a last verified date against each field.

To make the case internally, show your CEO one product across three channels with two of them wrong, then the same product after the fix. Nobody argues with that.

2. Missing information kills the sale mid-purchase

This one lands deep in the buying process, once somebody has chosen your product. Technical categories suffer most. Buying a pushchair, dimensions matter, and when they are not on the page the logical next step is to email or phone.

Most people do not. They go to the site that shows the spec.

The fix already exists inside your business. Sales and customer service hold a detailed record of what buyers ask, and nobody has connected the two halves. Theming that volume of enquiry used to be a project. With AI it is an afternoon.

Take those questions to your product pages and answer them from what is published. If you cannot, your customers cannot either. Write the answers into the descriptions, the specs or an FAQ block, in the phrasing customers use, and push the update to every channel.

Then make it a loop. Pull the last 20 inbound enquiries each week, check the themes, update the listings. Customer service is the most untapped product knowledge in most businesses, and it tells you what your returns really mean.

3. Worse competitors outrank you

Most people reading this can name a competitor selling the same products, with worse service and a weaker brand, sitting above them in the results.

They are not winning because Google likes them more. They are winning because Google understands them more.

Their data is structured, their pages answer the questions buyers actually ask, and their attributes are filled in. Search used to be about engines understanding your brand. Now it is about them understanding your products.

Run a competitor gap audit on your top three keywords. For whoever ranks above you, compare word count, spec count, structured data, image count and review volume.

Then close it in three places. Structured product data, because that is what gives ChatGPT and Perplexity enough context to know what the product is. Full answers to real questions, phrased as buyers phrase them: what is best for this use case, what is the difference between these two. And reviews collected with use case language, by asking how customers used it and which feature mattered.

4. Your data is right, and unreadable

Some teams have done the work. Specs accurate, descriptions thorough, and it still does not perform. The problem is presentation, and it fails twice. Humans cannot find the point through the fluff. AI cannot parse the page at all.

Buyers spend around eight seconds scanning a listing. AI tools spend zero seconds on an unstructured one.

Try this. Show a listing to somebody who does not know the product, for five seconds, then hide it. If they cannot say what it does, neither can anybody else.

Four things fix most of it. Structure the title as brand, name, key differentiator and variant, rather than a keyword sprawl or a bare part number that means nothing to a human. Lead the description with the biggest benefit in plain English, structured specs beneath. Fill every attribute field the platform offers, because comparison engines pull from those first and a blank field is the same as not existing. And write real alt text, because AI reads it, and product image three tells it nothing.

Attributes are where the biggest wins hide. We have seen conversion transform on attribute completion alone.

5. Nobody owns it

This is the hardest of the five, and the reason the other four keep coming back.

Fixes without ownership decay. You run the project, the numbers improve, and a year later everything sits where it started.

Product data touches most of the business, so it needs buy-in across departments rather than heroics in one. One customer completed a large enrichment programme where the improved data never technically reached the website. All that work, invisible to customers.

Somewhere in your business is a person who could own this. A person, not a team. In a small business it might be you. In a larger one, the best owners are cross-functional and well networked, and know everyone from sales through to supplier onboarding and category management.

Find them, give them the authority, and put a light process around them. Governance does not mean a committee. An inbox or a Google form is enough, as long as data does not change without passing through a defined route. Then track a quality score, because improvement nobody can see does not get funded twice.

The test is a single sentence. Write down who owns data accuracy at your company. If you cannot, that is the gap.

The 90 day version

Each of the five has a tactical fix you could run in a week. Structural change takes longer, and we use the same three phases with customers.

Phase one is targets and sources. Define the KPIs you intend to move: completeness, accuracy, time to list, return rate, inbound enquiries. Then consolidate every source into one place. ERP, legacy systems, supplier spreadsheets, marketing assets, images, manuals, SharePoint. Until that is done you do not know what you are fixing.

Phase two is the target model. Build the taxonomy first, then classify the products, then define the right attributes per category. Without that, there is no definition of a gap.

Phase three is filling the gaps. Attribute values, media, descriptive content, FAQs, until every product is populated against the model from phase two. How long that takes is a resourcing question, with 90% completeness a reasonable bar. Sequence by category and start with a high value one.

The part most teams get wrong is treating it as finished. It is a programme, not a project. Whether the catalogue holds 2,000 products or 100,000, next month's arrivals need the same discipline.

Start here

Take your five best sellers this afternoon. Open every channel they appear on and compare them line by line.

You will find a mismatch. It will not be anybody's fault, and it points straight at whichever of these five is costing you most.

Fix that one today. Then decide who owns making sure it stays fixed.

Listen to the full episode

Episode 11 of Product Data Weekly is available now. For more episodes and the weekly newsletter on operational issues inside product data and ecommerce teams, visit productdataweekly.com.