Fanatics’ Shop the Look: What a 10% Engagement Lift Does—and Doesn’t—Tell You

OpenAI says Fanatics uses Shop the Look to help fans complete game-day outfits. Testing across roughly 30,000 MLB products showed 10% higher engagement, according to the supplied excerpt ¹. The excerpt does not define “engagement” or report conversion or revenue results ¹.

For a retailer evaluating an AI shopping assistant or other AI product discovery feature, that is a reason to consider a test, not a business case to copy. An engagement increase alone does not establish whether shoppers bought more. Your own pilot would need to answer that.

What OpenAI reports about Fanatics’ Shop the Look

The supplied OpenAI excerpt describes Shop the Look as helping fans complete game-day outfits and reports 10% higher engagement in testing across roughly 30,000 MLB products ¹. It does not specify what “engagement” measures, or provide conversion, average order value, or revenue results ¹.

The excerpt also does not describe the test duration, comparison group, assignment method, or statistical significance ¹. Those are useful details to request when assessing a result. Their absence from the supplied excerpt means the 10% figure alone cannot answer those questions.

That does not show the result is wrong. It means the excerpt provides a headline result, not enough detail to assess the test design or downstream business impact. Checking a vendor or partner claim against what it actually reports is a habit worth keeping; Gemini's Overnight ADK Optimizations: What Companies Should Verify takes the same approach to another announcement.

Why engagement alone is a weak measure of AI retail engagement

A rise in an engagement measure does not, by itself, show that shoppers bought anything, spent more, or would not have bought without the feature. Before running a comparable test, define the action you will count as engagement and track it alongside commercial outcomes such as add-to-cart rate, completed orders, and revenue per session.

The supplied excerpt does not report those downstream outcomes for Shop the Look ¹. For your own test, tracking them separately would help distinguish interaction with a feature from commercial results. The same gap between activity and finished outcomes appears in More AI-Generated Code, No Matching Increase in Completed Work, which is a useful companion read on measuring output versus results.

Questions to answer about test design

The supplied excerpt does not provide the following details ¹:

• Comparison group: Was the feature compared with a randomized control, an earlier page, or no recommendation?
• Assignment method: How were shoppers assigned to test and comparison experiences?
• Duration and exposure: How long did the test run, and how consistently did shoppers see the feature?
• Significance: What analysis was used to assess whether the result could reflect random variation?
• Audience and context: Which shoppers and traffic sources were included, and what else changed during the test?

For your pilot, define the counted action, comparison experience, assignment method, test period, audience, analysis plan, and downstream outcomes before interpreting results.

What deploying a Shop the Look-style feature involves

The supplied OpenAI excerpt describes the purpose of Shop the Look but does not detail its interface, recommendation logic, catalog integration, or placement in the shopping flow ¹. Treat it as an example of a use case, not a deployment blueprint.

For your own implementation, first define the shopping task and the product information needed to make recommendations useful. Decide where the experience should appear and how it should handle items that become unavailable. If you are comparing models for the recommendation step, a governed model catalog can help you keep the options and constraints in one place.

Design instrumentation alongside the feature: record what recommendations were shown, what shoppers selected, and what happened afterward. That can help connect feature interactions to outcomes you choose to measure.

Can you reuse the 10% number?

The reported result concerns Fanatics’ Shop the Look testing across roughly 30,000 MLB products ¹. The supplied excerpt does not establish that the result generalizes to other retailers or shopping tasks ¹. Use it as a reason to consider testing a similar idea, not as a forecast for your own results.

Whether a comparable feature is useful for your business is something your own test can assess. Define the shopper task, the baseline experience, and the outcomes that matter to your business before using results to make a rollout decision.

What to measure in your own AI product discovery pilot

Choose a primary discovery metric and define it precisely, for example a click on a recommended item within the feature. Track it separately from add-to-cart rate, completed orders, and revenue per session. A change in the discovery metric without a change in downstream outcomes would indicate that the measures are moving differently; it would not, by itself, establish why.

Set up a comparison group, define the assignment method, and choose the test window before reviewing results. Where the available data permits, examine results by category, device, and traffic source. Decide in advance what evidence would lead you to expand, stop, or retest the feature, using your own baseline rather than the reported 10% figure.

The takeaway

The supplied report provides a signal that Fanatics’ Shop the Look testing was associated with 10% higher engagement across roughly 30,000 MLB products ¹. It does not provide evidence in the excerpt about conversion or revenue impact ¹. The practical lesson is to define the metric, design a comparison, and track the path from recommendation to purchase before deciding whether a similar feature merits a wider rollout.

FAQ

Did Fanatics’ Shop the Look increase sales?
OpenAI reports 10% higher engagement across roughly 30,000 MLB products, but the supplied excerpt does not report a conversion or revenue result ¹.

What did the 10% engagement lift measure?
The supplied excerpt does not define the engagement measure or the underlying action ¹. Get that definition before comparing it with another result.

How was the Shop the Look test designed?
The supplied excerpt does not report the test duration, comparison group, assignment method, or statistical significance ¹. Those remain open questions for evaluating the result.

Can another retailer expect the same 10% lift?
Not from the available evidence. The result is reported for Fanatics and roughly 30,000 MLB products, and the excerpt does not establish that it transfers to other catalogs or shopping contexts ¹.

What should retailers measure in their own pilot?
Measure the defined engagement action separately from add-to-cart activity, conversion, and revenue. The Fanatics excerpt reports engagement but does not provide those commercial outcomes ¹.

Source

¹ “How Fanatics and OpenAI streamline the fan experience,” OpenAI, published October 26, 2026. https://openai.com/index/fanatics-streamline-fan-experience