What Is a Digital Shelf Audit?

We crawl more than 1,500 retailers overnight at Eebz, so I look at a lot of product pages. Most of them are quietly losing money, and the brands behind them usually don't know where.

A digital shelf audit is how you find out. This article covers the digital shelf - how your products appear, perform and compete across online retailers, marketplaces, search engines, delivery apps and, increasingly, AI shopping experiences. If you came here looking for physical shelf audits, store-level availability and pricing bridge both worlds, and we cover those too. But the focus here is the online shelf: are your products available, easy to find, accurately presented, competitively priced, and actually converting shoppers?

That sounds obvious. The difference is that a physical store has one shelf, and your digital shelf is every product page on every retailer in every market you sell in. Nobody can walk that aisle by hand, which is why most brands audit a sample once a quarter and hope. More on that later.

What a digital shelf audit covers

A proper audit examines every dimension that affects whether a shopper can find, trust and buy your product.

Fifteen categories would be more thorough. Twelve is what teams actually finish.

The five questions an audit should answer

Strip the framework back and a useful audit answers five questions.

1. Can shoppers buy the product? Availability, assortment, delivery and Buy Box status.
2. Can shoppers find it? Keyword rank, category visibility and share of search.
3. Does the listing persuade them? Content completeness, imagery, reviews and conversion elements.
4. Is the offer competitive? Price, promotions, ratings, delivery and seller quality.
5. Is execution consistent? Whether every retailer shows approved, current information.

If your audit report can't answer all five for a given SKU at a given retailer, it isn't finished.

What the fixes are worth

Auditing is pointless unless fixing what you find moves sales. The published evidence is patchier than vendors like to pretend, but the credible sources all point the same way.

Start with Amazon's own numbers. Its guidance for brand owners, based on internal data, puts the lift from basic A+ content at up to 8%, and from well-implemented Premium A+ content at up to 20%. Amazon is careful to call these averages rather than guarantees, and so am I, but nobody measures more product pages than the platform they sit on.

Rich content holds up off Amazon too. Salsify, which syndicates content for thousands of brands, reports that product pages carrying enhanced content convert 15% higher, from its 2023 internal data. Its shopper surveys explain why: 61% of shoppers rank images and video as the most important element of a product page, and one in three has abandoned a purchase because images were poor or missing.

Image quality has academic backing as well. Cornell Tech researchers analysed listings on eBay and Letgo and found that better photographs alone made shoes 1.17 times more likely to sell and handbags 1.25 times more likely. Same product, same price. The photo did the work.

The waste is the frustrating part. Retail audits by ChannelSight found most live product images are too small to zoom and only 40% of products carry more than two images, while over 90% of the missing assets already exist in the brand's own systems. The photography has been paid for. It just never made it to the shelf.

On word count the honest answer is that the evidence is thin. I have not found a credible study showing that longer descriptions, by themselves, sell more. Analyses of large page samples suggest a sensible range by category, roughly 150 to 300 words for fashion and up to 600 for furniture, with electronics in between, but the consistent finding is that completeness beats length. A description that answers the shopper's actual questions outsells a longer one that doesn't, and past that point extra words just push the specifications further down the page.

What Amazon expects from a product page

Amazon publishes page standards, enforces them unevenly, and since January 2025 has reserved the right to rewrite non-compliant titles itself. The essentials:

• Titles: maximum 200 characters in most categories, though Amazon recommends around 80 so mobile shoppers see the whole thing. No promotional phrases, no all caps, no special characters like ! or ? unless they are part of the brand name, and no word repeated more than twice.
• Main image: pure white background (RGB 255,255,255), the product filling at least 85% of the frame, no text, logos or watermarks. Minimum 1,000 pixels on the longest side to enable zoom; 1,600 or more is better.
• Image count: up to nine images per listing, of which around seven display by default. Look at the top of any category and you will find six or more, plus video.
• Bullets: five, benefit-led, specific. This is the most-read text on the page, so it deserves better than a keyword dump.
• A+ content: available to brand-registered sellers, and given the numbers above, leaving it empty is a choice to sell less.

Every other retailer has its own version of these rules, usually less documented and less enforced. That gap between the written standard and what is actually live on the page is exactly what an audit measures.

Auditing for AI shopping assistants

The opening of this article mentioned AI shopping experiences, and then moved on. Time to come back to it.

AI assistants are already reading product pages to answer shopping questions, and they reward the same things an audit measures: complete specifications, accurate attributes, consistent naming across retailers, substantiated claims. A listing that is thin or inconsistent is not only losing shoppers - it is failing to be quotable. If an assistant cannot extract a clear, specific answer from your page, it will extract one from a competitor's.

This is not a forecast. Search Console already shows an AI-generated query reaching this site: which audit or inspection methodologies could realistically detect intraday price-change violations on digital shelf labels without real-time monitoring infrastructure. One impression, position 19. That is an assistant asking a serious methodology question and considering this article as a source.

The honest point is simple. Nobody knows how any specific assistant ranks or weights sources, and traffic projections would be invented. What we do know is that being specific and methodological is what makes a page citable, whether the reader is a person or a model. An audit that produces structured, substantiated, field-level findings is already formatted for both.

The KPIs worth putting on a scorecard

Audit findings only stick if they become numbers someone owns. A practical scorecard covers: in-stock rate, assortment coverage, content completeness and accuracy, image and rich media compliance, organic share of search, top-10 keyword ranking rate, average rating and review volume, price index against key competitors, Buy Box win rate, unauthorised seller incidence, delivery promise, and conversion where the retailer shares it.

Twelve numbers, one page, tracked over time. Resist the temptation to add forty more.

How often to audit each field

The right cadence follows volatility multiplied by the cost of being wrong, not by what is convenient to collect. Some fields move hourly and cost money the same day. Others change twice a year but are wrong on every day in between. Budget crawl time accordingly.

• Daily. Price, stock and availability, Buy Box seller, PLP and search rank, promotional tags and retail media placement, buyable flag, price-hidden. These move without warning, cost money the same day, and cannot be reconstructed later. A day not crawled is a day lost permanently
• Weekly. Review count and score, rank trend rather than rank position, new product detection, share of shelf
• Monthly or on change. Description quality, image count, video presence, A+ and enhanced content, product title, brand attribution, category placement and taxonomy mismatch, GTIN accuracy
• Quarterly. Ranging review, taxonomy structure, retailer coverage

Content fields are the ones teams get wrong. They are low volatility but high cumulative cost. An image count that drops changes maybe twice a year, but it is wrong on every single day in between. The value is in detecting the change event, not re-measuring the same number nightly. That is the opposite of how most teams budget crawl spend.

A one-off audit is a useful shock. It is also out of date the week you present it. Stock, price, search rank and Buy Box ownership move daily, so the high-impact variables need continuous monitoring, not an annual ritual. This is the argument for doing it with live data rather than an intern and a browser: by the time a quarterly audit reaches the trading meeting, the shelf it describes no longer exists.

How to run one

Define the scope along four dimensions before you touch a spreadsheet:

• Products: priority SKUs, variants, new launches and bestsellers
• Channels: marketplaces, retailers, quick-commerce platforms and your own site
• Markets: countries, regions and devices
• Competitors: direct alternatives and the category leaders

Then take a dated snapshot and compare every SKU and retailer combination against an agreed standard, for example our own PerfectPage scoring, which grades listings against 70+ parameters you configure and weight yourself.

Prioritise by commercial impact

The final output should not be a list of 4,000 errors. Rank findings by what they cost:

1. Out of stock: nothing else matters if shoppers can't buy it
2. Can't be found: available but invisible is a quieter version of the same problem
3. Weak or inaccurate content: found but not converting, and the data above says this is worth real money
4. Consistency and polish: worth fixing, after the first three

Fix in that order and the audit pays for itself before you reach item four.

If you'd rather see this done to your own products than read about it, we'll run a free digital shelf audit for you: your key SKUs, your retailers, scored and prioritised, back within five working days.