Online fashion retail interface showing personalized product recommendations.

A global fashion retailer predicts style from data people agreed to share.

Predicting the style of a new visitor, from data contributors agreed could be used for that purpose.

The data

Listening behaviour linked with fashion preference signals, from tens of thousands of contributors.

A global luxury fashion retailer in the Fortune 500 wanted to personalise for new visitors who arrived with no purchase history. Its research unit partnered with OpenDataLabs to test whether listening behaviour and other social signals could predict fashion preference.

What people agreed to

Each contributor gave permission for their data to be used for this purpose.

The record

Each contributor’s consent for inference to be run on their data.

That makes each of the model’s predictions consented inference. The research unit built and validated the model internally, against baseline demographic and behavioural inference methods.

What it made possible

The retailer’s model predicted style preferences more accurately than demographic and behavioural baselines. It can personalise for a new visitor on their first visit.

A new visitor sees recommendations shaped by their taste in the first session.