Product Page UX: Stop Losing Customers to Unanswered Questions

A customer who has reached one of your product pages is most of the way to buying. They found you, they found the product, and now they have questions: will it fit the space, what is it made of, does it work with the one they already own, what will delivery cost and when will it arrive. In a store, they would ask someone. On your site, the page either answers or it doesn’t, and when it doesn’t, the customer opens a new tab and asks Google. Whoever answers the question gets a say in where the sale lands.

Nobody hears about these failures. In Baymard Institute’s product page testing, almost no participants would use a chat feature to ask about a product even when one was offered, so an unanswered question doesn’t become a support ticket. The customer just leaves, and the page carries on failing the same way for the next one.

This post is for ecommerce managers and heads of digital. Some will recognise the problem immediately, because “improve product content” has sat on the roadmap for years. For others it is invisible, with the symptoms booked against other causes: returns, product questions in the support queue, paid traffic that lands and leaves. Either way, the argument here is the same: most of it is not a content problem in the copywriting sense. The gaps have a structure to them, the structure points back at where your product data comes from, and that changes both who has to fix it and what fixing it costs.

Where thin product page content comes from

Nobody decides their product pages will say so little. When a store is built or a catalogue is loaded, the product content comes from whatever source already holds it, and that is almost always a supplier feed, an ERP export, or both.

A supplier feed carries what the supplier needed it to carry: a title, a price, a barcode, and usually one paragraph of marketing copy written for a trade catalogue rather than a shopper. The ERP holds dimensions and weights because freight calculations need them, but those fields were mapped for the warehouse, not the website, and often never make the trip. The manufacturer’s full specifications exist as a PDF datasheet, so the page links to the PDF. The imagery is the supplier’s single cut-out shot on a white background, because that is what came with the feed.

At ten or twenty thousand SKUs, the enrichment everyone agreed to do after launch never starts, because nobody has the hours to fill in that many products by hand.

That inherited copy carries a second cost. Every other retailer who buys from that supplier received the same paragraph, so the identical description is live on a dozen competing sites. For a customer comparing tabs, your page reads like everyone else’s. For a search engine deciding which of those near-identical pages to rank, you have given it nothing to prefer.

What the customer sees

Baymard’s current benchmark rates 52% of desktop sites and 62% of mobile sites as mediocre or worse on product page UX, built on more than 30,000 manually reviewed usability scores. Their consistent finding is that pages fail through an accumulation of medium-severity gaps rather than one catastrophic defect. The gaps below are the ones that recur, and each one maps back to a specific hole in the data pipeline.

A description written for someone else

Copy inherited from a supplier is easy to spot once you read your own pages as a shopper. It leads with the brand’s positioning rather than the product, uses the trade’s jargon, and runs out after two sentences. It was written to sell the product to a retail buyer who already knows the category, which is exactly the person your customer is not.

Baymard found 10% of major ecommerce sites have descriptions insufficient for users’ needs, and their testing shows what insufficient costs: participants abandoned products they had already added to the cart when the description couldn’t confirm suitability, and others filled the gaps with guesses, which surfaced later as frustration and avoidable returns. When several products in a row came up short this way, participants concluded the whole catalogue was poorly documented and left to research on a competitor’s site.

Structure matters nearly as much as substance. Baymard found users slow down and explore each feature when a description is broken into highlights, meaning each feature paired with an image or icon, yet 78% of sites rely on plain text blocks and bullet lists even on their five best-selling products.

Specifications that stop mid-table

Open a spec table on one of your own products and count the rows that are blank, marked “N/A”, or missing entirely against what the manufacturer publishes. A half-empty table is worse than no table, because it tells the customer you don’t know your own products, and it usually means the attribute exists somewhere upstream and stopped there. Dimensions sit in the ERP because freight needs them. Materials and capacity sit in the PIM, or in the PDF, or in a spreadsheet the buying team keeps.

The PDF datasheet link deserves its own mention. It answers the question technically while failing the customer practically, especially on a phone, where a multi-megabyte download abruptly ends the research. Baymard’s testing of spec sheets found 50% of sites present them in ways users struggle to scan, and a PDF is the least scannable format there is.

One photo from the supplier

The single cut-out shot on white answers one question, which is what the product looks like from the front. It answers nothing about size, texture, or what the product looks like in use. Baymard found 42% of users try to judge a product’s size from its images, which is close to impossible without an in-scale image showing the product against something familiar, and 28% of sites don’t provide one for their own best sellers. In their testing, more than half of participants went straight to the image gallery before reading a word of the page.

For furniture, homewares, luggage and anything else where size expectations drive returns, this is the cheapest returns-reduction work available: photograph your top sellers next to a person or in a room, and the “smaller than I expected” refunds drop.

The answers that aren’t product data at all

Two of the questions customers most reliably bring to a product page have nothing to do with the product: what will delivery cost, and what happens if I need to send it back. Baymard found 64% of users looked for shipping costs on the product page before deciding whether to add the item to the cart, and 43% of sites show nothing on the page, not even an estimate or a calculator. Their benchmark also finds 44% of sites keep returns information off the product page entirely.

In Australia the delivery estimate is even more important, as times and costs can vary wildly for metro and regional areas. Showing a real estimate on the page requires postcode-aware rates and delivery times from your carrier integration, which is why this gap is often a backend integration concern rather than related to the copy itself.

Copy problem or data problem

The same split we described for category navigation applies here: the two halves have different owners and different price tags.

The copy half is real and bounded. A competent writer can rewrite the descriptions for your top two hundred products, structure them into scannable highlights, and lift the pages that carry most of your revenue. That is weeks of work, it needs no engineering, and you should do it.

What a writer cannot do is populate ten thousand spec tables, and nobody should ask them to. That half is a pipeline problem: attributes that exist in the ERP or the supplier’s data and never reach the storefront, attributes populated inconsistently across the catalogue, specifications trapped in PDFs that were never extracted into structured fields. The fix runs through your product data layer, which is one of the clearer arguments for a PIM if you don’t have one, and through an integration that carries every field the page needs on every sync instead of the minimum the original mapping specified.

The same gaps in your product data are about to matter in a second place. AI shopping agents research products by reading the same pages and feeds your customers do, and an agent evaluating your catalogue cannot squint at a photo or infer a missing dimension the way a patient human sometimes will. A blank attribute is simply a product that never enters the comparison. The work described here and the work of becoming legible to agents are the same work.

Diagnosing your own product pages

You don’t need a usability study for a first diagnosis. Everything below takes about an hour, using only your own store, your support inbox and your returns report.

• Pick your five best sellers and read each description aloud. Would a salesperson say that sentence to a customer standing in front of the product?
• Take one full sentence from a description and search Google for it in quotes. Count the competitor sites carrying the identical paragraph.
• Open the spec table on twenty products in one category and count blank or missing rows against the manufacturer’s published specs.
• Find a product whose dimensions you know are in the ERP, because freight uses them, and check whether they appear on the page.
• Open a best seller on your phone and try to work out how big it is without reaching for the tape measure.
• Try to find the delivery cost to a regional postcode without adding the product to the cart.
• Search your support inbox for your top product’s name. Every repeated question is a page answering too little.
• Pull your returns by reason and read the “not as expected” and “wrong size” rows as page defects rather than customer error.

What to fix first

In rough order of return on effort:

1. Put shipping estimates and a returns summary on the product page. The template change is small and the demand is measured at 64% of customers.
2. Fix how specifications render: hide empty rows and replace PDF links with structured tables for the products where the data exists.
3. Rewrite and restructure the descriptions on the products that carry your revenue.
4. Photograph your top sellers in scale.

The catalogue-wide job comes last, because it is the largest and it reaches back into your systems: deciding where each attribute lives, extracting what is trapped in PDFs and spreadsheets, and building the pipeline that moves all of it to the page and keeps it there as the catalogue turns over. It is also the piece that stops the problem regrowing, because pages enriched by hand on top of a broken pipeline drift back to thin content within a couple of buying cycles.

If you would rather hand the diagnosis to someone else, our UX audit is led by a Baymard-certified UX Professional and covers product pages as one of its core areas: you get a report of where your pages fall short and what to change. Whether the fixes land in the template or back in your product data, they are work we do ourselves, so get in touch and we can look at it with you.