Better Product Data, Better Customer Experience
Most customer experience programmes start with the page. A new product page template, a faster checkout, a chat widget, a redesign of the category pages. Those things matter. But by the time a customer sees any of them, most of their experience has already been decided by something less visible: the product record sitting behind the page.
Our co-founder learned this the hard way during a decade in digital trading at Tesco. Leadership spent a lot of energy on category pages, convinced they drove trade. In practice, Google rankings, shopping feeds, paid search and email decided whether a customer arrived at all, and every one of those channels ran on product data. The page was the last layer, and it was the one everyone was optimising.
That lesson applies even more now. If you want a better customer experience, start with better product data.
Every touchpoint reads the product record
Think about the routes a customer takes to a product. They search on Google. They tap a Shopping ad. They use your site search, then narrow the results with filters. They read the product page, compare two items, ask a question in chat, or ask an AI assistant to recommend something. They get a recommendation email a week later.
Each of those moments is generated from the same source: titles, attributes, descriptions, images and classifications. When that record is thin, inconsistent or wrong, every channel inherits the problem. A beautiful product page cannot fix a product that never appeared in the search results.
How thin product data damages the experience
The customer cannot find it
Filters only work on data that exists. A shopper who selects “waterproof” or “suitable for underfloor heating” will never see a product where that attribute was left blank, even if the product qualifies. The same is true of site search, Google Shopping and marketplace feeds. From the customer’s side, the product simply does not exist.
The customer cannot decide
A customer with an unanswered question has three options: guess, contact you, or leave. Missing dimensions, vague materials, no care instructions, no compatibility information. Each gap pushes some customers to a competitor whose page answered the question, and pushes others into your customer service queue.
The customer gets the wrong thing
Inaccurate data is worse than missing data. A wrong colour name, an incorrect size or a copied supplier description that does not match the actual product sets an expectation the product cannot meet. That shows up as returns, poor reviews and lost trust, which cost far more than the original sale was worth.
The AI assistant skips you
Customers increasingly ask AI assistants to shortlist products for them. These assistants match on structured facts and avoid recommending products they cannot verify. We covered this in detail in why agentic commerce fails on thin product data. The short version: a sparse record gets left out of the answer.
Better means deeper and more accurate, not longer
A common response to “our product content is weak” is to make descriptions longer. Length rarely helps. A 400-word description that never states the dimensions is still a poor experience.
Better product data has three qualities:
Depth. The record answers the questions customers actually ask before buying: size, material, fit, finish, care, compatibility, what it is suitable for. Your returns reasons and customer service logs are the best guide to which fields matter.
Accuracy. Every value can be traced to a reliable source (the manufacturer, a spec document, the product itself) rather than inferred or invented.
Consistency. The same thing is described the same way across the catalogue. “Oak”, “oak”, “Natural Oak” and “Oak Wood” might mean the same finish to your team, but to a filter they are four different options. Our guide to product data enrichment covers this cleanup in more depth.
How merchi.ai builds a better product record
merchi.ai is built to fix the record at the source, at catalogue scale, with people staying in control of what gets published.
It starts with research. Web Research finds specifications, manufacturer documents and other trustworthy sources for each product before any content is written, so attributes are grounded in evidence rather than guessed from a product name.
Your schema defines which fields every product must have and how they should be written, so depth and consistency are enforced across thousands of SKUs, not left to whoever happened to load the product. Where an industry classification applies, Standards Packs map products to it automatically.
Accuracy is visible rather than assumed. Confidence scoring shows how reliable each generated field is, so your team reviews the uncertain values and approves the rest quickly. Every piece of AI content also carries a provenance record showing where it came from.
The outcome is a product record that serves every channel at once: richer filters, better search matches, product pages that answer the question, feeds Google can use, and structured facts AI assistants can trust.
What it looked like for one retailer
A UK flooring retailer had around 1,000 products sitting unpublished because the data was incomplete. Creating content manually took around 45 minutes per product, so the backlog never moved. Once those products were researched, enriched and published, customers could finally find, filter and compare them. Online revenue grew 976% year on year, driven by organic search rather than paid spend. The full story is in the case study.
The revenue number gets the attention. The experience behind it is simpler: customers found the products they were looking for, and the pages told them what they needed to know.
Where to start
You do not need to fix the whole catalogue at once. Take your 50 highest-traffic products and compare their records against the questions customers ask in chat, in reviews and in returns reasons. Every unanswered question is a gap in the experience, and the pattern usually points to the same five or six missing fields across a whole category. Fix those first, then scale. If you want to understand the cost trade-off, see AI vs manual product data.
Talk to us about your product data
If customers struggle to find products through your filters, your service team answers the same product questions every day, or returns point to descriptions that do not match the product, book a call to see how merchi.ai improves the product record behind your customer experience.
Or start a 30-day free trial and run it on your own product data.
Frequently Asked Questions
How does product data affect customer experience?
Product data feeds every customer touchpoint: search engines, shopping feeds, site search, filters, product pages, recommendations and AI assistants. When it is incomplete or inaccurate, customers cannot find products, cannot answer their questions before buying, and are more likely to receive something that does not match their expectations.
Can better product data reduce returns?
Yes, where returns are caused by mismatched expectations. Accurate dimensions, materials, colours and fit information help customers choose the right product first time. Your returns reasons will show which fields are causing the most problems.
What is the difference between deeper and longer product content?
Longer content adds words. Deeper content adds answers. A deep product record contains the structured attributes customers rely on to filter and compare, plus a description that explains them clearly. Length without facts does not improve the experience.
How does merchi.ai keep AI-generated product data accurate?
merchi.ai researches each product from trusted sources before generating content, enforces your schema across every SKU, and gives every field a confidence score so your team can review uncertain values before publishing. Each piece of content also carries a provenance record.
