Retention Edge E28: Your product data is holding you back

Many ecommerce teams are still writing product descriptions by hand.

Sometimes that means tens of thousands of SKUs, with team members manually putting descriptions together. Talk about a bottleneck.

In this week’s Retention Edge pod, we sat down with Vidar Trojenborg, co-founder of Emfas AI, to talk about the future of product data in ecommerce - and why AI-native PIMs are about to replace a whole category of manual work.

Now let’s jump into the top takeaways from our chat with Vidar.

Product data is the unsexy work that drives everything

Every product you sell online needs to be described, categorized, tagged, translated, and optimized. Across every channel. For every SKU.

That work has traditionally been done by hand. Someone sits in a PIM system or a spreadsheet and types out descriptions, assigns categories, writes meta tags, translates content into other languages. Rinse and repeat for every new product.

As Vidar put it, most brands are drowning in administrative work around their product catalog instead of doing what they're actually good at: building the brand, being creative, making good products.

It's the kind of thing that doesn't feel urgent until you realize your product pages are thin, your SEO is suffering, and your content is inconsistent across channels.

And with LLM-powered shopping here, the quality of your product data now matters even more.

AI doesn't need your data to be clean. Just honest

Vidar works with a lot of fashion brands where the design team hands over product notes that are basically random bullets with spelling mistakes.

Material info, design features, fit details, all shoved into an unstructured document that the ecommerce team then has to decode and rewrite.

That handover used to be a real pain point. But LLMs are actually great at this.

You feed in messy, unstructured data - notes from the design team, product images, technical specs - and the AI extracts what matters, organizes it, and rewrites it in your brand voice.

The same goes for product images. If you have a photo of a jacket on a model, the AI can identify the fit, color, design details, even whether it's oversized.

That visual data becomes another input for richer product descriptions.

The takeaway: You don't need a pristine data pipeline to start using AI for product enrichment. You need accurate raw information. AI handles the rest.

Multi-brand retailers have the most to gain

If you're a multi-brand retailer, you're getting product data from dozens of suppliers.

Nike sends it one way. Adidas sends it another. What one brand calls "Marine" another calls "Navy."

Your job is to normalize all of that into a consistent catalog that reflects your brand's standards.

Vidar and Emfas built a feature called deep research - where you give the AI a SKU or barcode, and it goes to the brand's website, pulls all the relevant product information (material, fit, design details), brings it back into the PIM, and rewrites it in your brand voice.

Fully automated. For multi-brand retailers with thousands of SKUs from dozens of suppliers, that's a fundamentally different workflow than manually copying and reformatting supplier spreadsheets.

Rich PDPs are your best bet for showing up in AI search

GEO (generative engine optimization) and regular SEO are largely the same thing. Most of the same inputs move the needle the same way.

But there is one difference worth paying attention to.

When someone searches on ChatGPT, the query has way more context than a typical Google search.

It's not "ski goggles". It's "I'm going on a ski trip to this specific place, I have sensitive eyes, I prefer these brands, and my budget is X."

The LLM matches that intent against the data on your product pages. If your PDP just says "ski goggles, $149", you're invisible.

Vidar's recommendation: add rich FAQ sections to your PDPs.

• Who is this product for?

• How is it used?

• What are the unique selling points?

• What conditions is it designed for?

It doesn't have to be visually prominent. It can sit at the bottom of the page. But as long as the data is there, AI shopping tools can find it and match it against what the customer is actually looking for.

Winning in AI-native ecommerce is not necessarily about sophisticated SEO hacks. It’s about product data that’s as in-depth as your customer’s prompts.

The product manager role is shifting from data entry to AI management

The person who used to spend their day manually updating product descriptions is now becoming the person who manages the AI that does it.

Same domain knowledge. Completely different job.

As Vidar described it, they're going from writing individual descriptions to setting rules, defining brand guidelines, and reviewing AI output. They're prompt engineering their product catalog.

For some content types (alt texts, category assignments, meta descriptions) brands are putting the AI on full autopilot. The accuracy is high enough that manual review isn't worth the time.

For high-visibility content like product descriptions, some brands still review everything. But even that is shifting as they dial in the rules and see consistent output.

The big picture is this: AI isn't replacing the product team. It's changing what the product team does.

And the brands that figure out this new workflow first have a real advantage. They can enrich their entire catalog faster, more consistently, and with more creative control than they ever could with manual processes.

Summing up

Product data has been a bottleneck for as long as ecommerce has existed (particularly for brands with high SKU counts).

AI is removing that bottleneck, and the brands that move on this now will have richer catalogs, better discoverability in AI search, and more time to focus on actually building their brand.

Catch the full episode on YouTube or. Spotify. Wherever you watch or listen, be sure to like, comment, share if you found it useful.

You can connect with Vidar here, and check out his company Emfas here.

I’ll back later this week with more to help you grow & scale your brand the right way.

— Pietro

PS - Want to see what your store would look like as a mobile app? Get a free preview here, or shoot me a DM on LinkedIn.