Completeness score: measure and improve your product listing quality

A "complete" product listing on Amazon is not the same as a complete listing for a B2B reseller. The completeness score is the tool that allows you to objectively measure the quality of each product listing according to its target channels — and to identify exactly what is missing.

Why measure catalogue completeness?

Most retailers intuitively know that their product listings are “incomplete”. But without measurement, it is impossible to prioritise, delegate or track progress. The consequences of a poorly populated catalogue are clear:

• Reduced conversion rates— shoppers abandon listings without descriptions, dimensions or high-quality images
• Marketplace rejections— Amazon, Mirakl and most marketplaces reject products where mandatory attributes are missing
• Poor search ranking— without a structured title or key attributes, products do not appear in the platforms’ internal search results
• Customer returns— missing or inaccurate information (incorrect dimensions, unspecified materials) leads to costly returns

The Pixee PIM Completeness module transforms this vague issue into an actionable dashboard.

How the completeness score works

The score is calculated per listing, based on the rules you define for each channel or catalogue segment. An identifier being present is one thing, its being valid is another — that is the job of the EAN Manager, which checks barcodes. The score takes three factors into account:

• Presence— is the attribute filled in? (title, description, EAN, brand, category, price, stock)
• Quality— does the value comply with the defined constraints? (minimum title length, description > 150 characters, at least 3 images, dimensions in millimetres rather than centimetres)
• Channel compliance— are the attributes specific to the target channel present? (ASIN for Amazon, GTIN for Google Shopping, EPREL ID for household appliances sold in the EU)

Each attribute can be assigned a weighting in the score calculation — a missing EAN carries a heavier penalty than a missing second image.

Define your completeness rules by channel

This is where the module’s power really shines. You can configure different completeness profiles according to your channels and segments:

• Amazon profile— title between 80 and 200 characters, minimum of 5 bullet points, 7 images including 1 with a white background, valid ASIN, mapped Amazon category, weight and dimensions mandatory
• B2C shop profile— long description > 300 words, minimum of 3 images, technical specifications with key technical attributes, canonical URL
• B2B PDF catalogue profile— manufacturer’s reference, precise dimensions in mm, weight in grams, customs code, country of origin, available certificates
• EU compliance profile— CE declaration, WEEE marking where applicable, REACH safety data sheet, EPREL ID for household appliances

The same product may score 95% for the B2B profile and only 60% for the Amazon profile — which tells you exactly what to enhance before launching a campaign on the marketplace. The attributes that this particular marketplace expects are covered in our article on selling on Amazon from a PIM.

The completeness dashboard

Pixee PIM displays completeness at several levels to facilitate prioritisation:

• Overview— average catalogue score by channel, trends over time, breakdown by bracket (0–50%, 50–80%, 80–100%, 100%)
• Category view— the most incomplete categories appear first, so the team can focus its efforts where the impact is greatest
• Supplier view— compares the quality of data received by supplier, useful for supplier performance reviews
• View by product code— for each product, a detailed attribute-by-attribute breakdown of what is present, what is insufficient and what is missing

Block the publication of incomplete product sheets

The Completeness module integrates with the publication workflow. You can configure minimum thresholds below which a record cannot be published on a given channel:

• Score < 70% → record blocked with status "To be enriched", not visible on the channel
• Score 70–90% → record published but marked "To be improved", included in weekly alerts
• Score > 90% → listing published and validated

These thresholds can be configured per channel. You can set higher standards on Amazon (threshold 85%) than on your internal shop (threshold 60%) depending on your business priorities.

Who fixes what, and in what order

Measuring achieves nothing if nobody works through the list. The score becomes valuable when it feeds an explicit correction circuit: the editorial workflow in Pixee PIM moves each record through Draft → Review → Approved → Published → Archived, with assignment by role and a completeness gate at the entrance to publication. A record below the threshold does not clear the stage, and its owner knows it is coming back to them.

The order in which you work then matters as much as the tool:

• Blocking attributes first— the ones without which the channel refuses the record. Fixing ten invalid EANs unblocks ten SKUs; adding a fourth image unblocks none.
• Heavily weighted attributes next— the ones you gave a high weighting because they influence the purchase decision or the channel’s internal ranking.
• Everything else in bulk— fields that are uniform across a whole category are best handled in spreadsheet editing, up to 100 products at a time, rather than record by record.

AI enrichment and completeness

Pixee PIM links the completeness score directly to the AI enrichment engine. From the product view, clicking "Enrich with AI" automatically generates the missing attributes as a priority:

• Short and long descriptions generated from the title, brand, category and existing technical attributes
• Automatic translation of descriptions into target languages (FR, EN, DE, ES, IT, NL…)
• Category and tag suggestions based on semantic analysis of the product
• Completion of missing technical attributes via web search or Icecat where available

The score is recalculated in real time after each update, allowing you to see the impact of each action immediately.

Alerts and automated reporting

Catalogue teams cannot manually monitor thousands of product listings. Pixee PIM sends automatic completeness reports:

• Weekly email report listing the X most incomplete SKUs for each active channel
• Immediate alert when a supplier import introduces SKUs with a score below the configured minimum threshold
• Monthly report on changes in the average score by category and by supplier, for performance reviews

What the score does not measure

A completeness score measures whether a piece of data is present and well formed, never whether it is correct. A record with every field populated shows 100% even if the weight is out by a factor of ten, even if the description refers to the previous generation of the product. This is the structural limit of the exercise, and it is better stated openly to the team from the outset: the score is there to spot gaps, not to certify that the catalogue is true.

Three safeguards usefully complement the measurement:

• Format rules— expected unit, plausible value range, minimum length. They turn part of your accuracy errors into formatting errors, and therefore into detectable errors: a television listed at 300 grams falls outside the range.
• Identifier validation— an incorrect EAN reveals itself through its checksum, an incorrect weight does not. Anything that can be checked mechanically should be.
• Human review by sampling— a handful of records picked at random from each enriched category is enough to spot a systematic drift, whether it comes from a supplier mapping or from a poorly framed AI model.

An illustrative scenario: electronics retailer, 15,000 SKUs

The orders of magnitude below describe a typical situation, reconstructed from setups we commonly encounter. They are not the measured results of a named customer.

Take a consumer electronics distributor with 15,000 active SKUs and 20 suppliers:

• Initial score at launch: around 50% on average on the Amazon profile, 70% on the B2C shop profile
• After six weeks of AI enrichment on the 3,000 priority SKUs (score < 60%): average Amazon score close to 80%, B2C close to 90%
• Expected downstream effect: an Amazon acceptance rate moving from around two thirds to more than nine listings out of ten, and a noticeable drop in returns caused by incomplete descriptions
• Enrichment time: a few hours for 1,000 SKUs using AI, against several weeks manually

Which plans include the Completeness module?

The Completeness module is included in all paid plans (Starter, Growth, Scale). Customised completeness profiles by channel and automated reports are available from the Starter plan onwards. Real-time alerts and integration with the publishing workflow are available from the Growth plan onwards.

Frequently asked questions

Is a record at 100% necessarily a good record?

No. The score confirms that the expected fields are populated and well formed, not that their content is correct. A generic description, an incorrect weight or an image showing a different colourway all clear the check without difficulty. Read the score as a floor: below it, the record is certainly inadequate; above it, the record is merely publishable.

Which blocking threshold should you choose at the start?

The Pixee PIM completeness gate sits at 70% by default: high enough to keep hollow records out, low enough not to freeze an entire catalogue on the day you go live. A very demanding threshold set too early produces the opposite of the intended effect — teams work around it or switch it off. Raise it channel by channel, once the average score has started to climb.

How do you weight attributes without spending weeks on it?

Start from the channel’s list of mandatory attributes: it is imposed on you, there is nothing to arbitrate, so give them the maximum weighting. Then add the two or three attributes that genuinely drive the purchase in your sector — dimensions for furniture, compatibility for spare parts, composition for textiles. Everything else can stay at equal weighting: finer tuning costs more configuration time than it returns in prioritisation.

What should you do with SKUs that will never reach the threshold?

Not every record deserves to be enriched. An end-of-life SKU, with no stock and no rotation, costs more to complete than it will ever return: the useful decision is to pull it from the channel, not to fill it in. The score by category and by supplier serves that purpose too — it makes visible the share of the catalogue that is better archived. You can query that scope in natural language with the AI Copilot, which answers questions such as “which products have a completeness score below 60%?”.