Insights lesenGespräch

Product data

Good data helps sell.

Product Data Quality in Ecommerce: How It Drives Visibility, Sales and Returns

Images, descriptions, dimensions, materials: product data looks unremarkable. Yet purchase decisions, return rates and visibility depend directly on it.

4 min read

Key takeaways

  • Incomplete product information leads to abandoned purchases and avoidable returns.
  • Good product data is also a prerequisite for visibility in Google, price comparison sites and AI search.
  • Data quality needs clear ownership, fixed standards and a leading system.
Contents

Product data is the raw material of online retail. Without it, there is no product page, no filter, no feed for Google Shopping and no recommendation in AI search. Yet in many companies it is maintained on the side: by different people, in different systems, to different standards.

The consequences show up in many places.

69 %of Germans have abandoned a purchase at least once because of incomplete informationAkeneo, September 2025
60.4 %cite inaccurate descriptions or product images that differ from the item as a reason for returnsKPMG, December 2025

Where poor product data costs money

In the shop: fewer purchases

When dimensions, material details or compatibility information are missing, customers don’t buy, or they buy elsewhere. Incomplete data also makes filters and search useless: a product with no color attribute won’t appear when someone filters for “blue”.

After the purchase: more returns

If the product doesn’t match its description, it comes back. In a KPMG survey of German online shoppers, 60.4 % of respondents cite inaccurate descriptions or differing images as a reason for returns. Wrong sizes, at 72.6 % one of the most common reasons, are also often tied to missing or unclear size information.

In search: less visibility

Google Shopping, price comparison sites and marketplaces work with product feeds. If mandatory fields are missing or data is contradictory, products are ranked lower or not shown at all. The same applies to AI systems: they only use data they can understand. When AI reviews assortments, compares prices and recommends products, a great deal of visibility is lost right here.

Every gap in the product data is a question the customer can’t ask, and so they don’t buy.

What makes good product data

Four characteristics of good product data

CompleteAll relevant attributes
Dimensions, material, compatibility, care, scope of delivery
AccurateVerified and up to date
Details match the actual product
ConsistentFixed standards
Same units, spellings and structures
StructuredMachine-readable
Attributes as fields, not just in running text

The last point in particular is often underestimated. If the weight appears only in the description text, no filter, feed or AI can reliably extract it. It has to be maintained as a separate attribute.

How to improve your product data systematically

1. Take stock

How complete is your data today? For your most important categories, check which attributes are missing. Start with the products that generate the most revenue or the most returns.

2. Set standards

Define mandatory attributes, units and spellings for each category. A data model for “T-shirt” needs different fields than one for “cordless drill”.

3. Determine a leading system

Where is product data maintained? In the ERP, in the shop, or in a dedicated PIM (product information management) system? What matters is that there is exactly one leading source for each piece of information. All other systems are fed from it.

4. Clarify responsibility

Who maintains which data? Who reviews it? Without clear ownership, even good standards quickly erode.

5. Use feedback

Return reasons, customer inquiries and reviews show where information is missing. If customers keep asking about the same detail, it belongs on the product page. According to the German retail institute EHI, 75.2 % of online retailers in Germany rely on detailed product information to avoid returns.

6. Use AI deliberately

AI can suggest missing attributes from descriptions, standardize texts and find contradictions. This saves a lot of manual work. The review, especially of technical and legal information, remains with the team.

An example: from running text to attribute

Suppose a shop for garden equipment has so far maintained its lawn mowers like this: one long description text contains cutting width, battery life, weight and grass box capacity. That is readable for customers, but barely usable for systems.

After the changeover, these details are stored as separate attributes in the system. The results:

  • Customers can filter by cutting width and battery life.
  • A comparison table shows differences between models at a glance.
  • Google Shopping and price comparison sites receive complete feeds.
  • AI systems can correctly classify the products for questions such as “Which cordless lawn mower for 500 square meters?”
  • Customer service gets fewer follow-up questions about technical data.

The effort: set up a clean data model for the category once and transfer the existing details. This can be done well with AI support, followed by a review by the team.

A simplified example for illustration.

Conclusion

Product data is not a compliance exercise but a competitive factor. It determines whether customers buy, whether they keep the goods, and whether your products are found in Google, on price comparison sites and in AI search. Those who invest systematically here improve several metrics at once.

Frequently asked questions

Do I need a PIM system for good product data?

Not necessarily. With many products, variants, languages or channels, a PIM usually pays off. But what matters is not the system, it is standards and ownership.

Which product data is most important?

The attributes customers need for their purchase decision and that they filter by. Depending on the category, these are dimensions, material, sizes, technical specifications, compatibility or care instructions.

The more accurately the description, images and size information reflect the product, the less often customers are disappointed. Inaccurate information is among the most common reasons for returns.

Sources

Inhalt

    InsightsIns Gespräch kommen