The insights model

This is for:

Developer

In this article, in this article, we’ll introduce you to the data model behind our product insights.

What are product insights?

Note

This feature is currently in Beta–this means that the functionality is subject to frequent change as we fine-tune the implementation.

Product insights are suggestions we make about a product and a product location that we predict, if actioned, would increase the number of clickthroughs to the PDP and, therefore, the number of conversions.

How does our model predict these insights?

Each of your product insights is the result of a machine learning model that predicts which combination of a product and a PLP would drive the largest increase in revenue if more people on your site clicked-through on that product.

The model determines the likelihood that a visitor to your site will convert after clicking any product on any of your site’s PLPs. It then forecasts the increase in the revenue (based on conversions) you might expect if you could increase the number of clickthroughs for each product on the PLP.

The model also adjusts its forecasts based on other product features. The model learns how both product price, and its pricing relative to other similar products, affect {clickthrough-rates} if its views on the PLP were increased.

We also filter the generated insights before showing them to you. Only insights that yield "enough" extra conversions or “avoid products currently on sale” are worth actioning.

What action can I take on an insight?

As an example, you can action an insight by increasing the prominence of the suggested product on the PLP by positioning the product at the top of the PLP and adding an appropriate badge such as "our choice."

What level of control can we have over the model?

If you already know what your flagship products and bests PLPs are, you can also ask the model to tell you only about the products and pages you haven’t optimized yet. These insights should be not only actionable but relevant.