Plainsight

Torfs · Retail

We get too much customer feedback to read, and it's too valuable to ignore

Plainsight built Torfs a generative AI solution that turns its steady stream of customer feedback into structured insight on price, product and service, while every original comment stays available for deeper analysis.

  • Retail
  • Generative AI on customer feedback
  • Embeddings and vector storage
  • In the top 5 most consulted reports

Uitdaging

Too much feedback to read, too valuable to ignore

Torfs, the Belgian shoe retailer, receives a high volume of customer feedback: about 30,000 reviews a month. That feedback is one of the clearest signals there is on how customers see the price, the product and the service, and where new trends are starting.

The problem is volume. Nobody can read all of it, and a sample read by hand misses the patterns that only show up when you look at everything. Feedback nobody reads can't be acted on.

Wat we deden

From free text to insight you can rank

Every comment becomes a vector.

We convert each piece of feedback into an embedding: a numerical fingerprint of what it means, not just which words it uses. Stored as vectors, comments that say the same thing in different words end up close together, so themes surface on their own.

Structure where you need to rank.

On top of that, the solution derives structured information on the factors that matter most, such as price, product and service. Those can be ranked and followed up like any other KPI.

How we build AI into the way you work →

Nothing gets thrown away.

The unstructured feedback stays available. When a score moves, the team can go back to the comments behind it and read what customers actually said.

Impact

The insight people actually open

Customer perception of price, product and service, structured so it can be ranked and followed up. Every original comment stays available for in-depth analysis.

  • About 30,000 customer reviews a month
  • In the top 5 most consulted reports at Torfs
Generative AIEmbeddingsVector storageCustomer feedback analytics

Veelgestelde vragen

What are embeddings, in plain terms?

An embedding turns a piece of text into a list of numbers that captures its meaning. Texts that mean the same thing get similar numbers, even when they use different words. That makes it possible to group, search and compare large volumes of feedback by what it says rather than by keywords.

Why not just ask a chatbot to summarise the feedback?

A summary tells you what was said once. Torfs needed something to track: structured information on price, product and service that can be ranked and followed up, with the original comments still there to read. Embeddings and vector storage make both possible on the same data.

What is vector storage?

A database built to store embeddings and find the ones closest to each other. It lets you ask for every comment that means roughly the same as a given one, which keyword search can't do.

Does this work for other kinds of text?

Yes. The same approach works on any large volume of free text, such as reviews, support tickets or survey answers.

Een gelijkaardig probleem bij jou?

Contacteer een expert