This is for:


In this article, we will look at the Replenishment model and how it is used to predict visitor behavior. We will provide an overview of how this model is built, its data requirements, and the predictions it can make. We will also provide details of use cases showing how this model can be integrated into your personalization efforts.

In particular, we will cover:

  • What our model predicts

  • What it uses to make that prediction

  • Some potential use cases

  • How you can reach out to Qubit should you be interested in using the model to power your experiences

We presume that you have already familiar with Derived Datasets, the Qubit product used to action this model.


The Replenishment model is a Derived Dataset model designed to predict when a returning visitor–a visitor that has previously visited the site–intends to replenish a product they previously purchased. The model uses the median inter-purchase period of different products to determine when they need to be replenished.

Input data

The model uses the QProtocol fields indicated in the table below. One or both of product_productId and product_sku are required to identify products:

Event Field Example Required



2017-08-12 11:27:03.931 UTC













Depending on QP implementation




Depending on QP implementation






The events detailed above are valid for the ecommerce vertical. For the equivalent fields for other verticals, refer to our Events page.


If one or more of the feature fields are missing or invalid, the model will often continue to operate, but with less input information, which can lead to lower accuracy.


This model requires historical transactions for products to determine their replenishment periods. As such, clients using this model require a sufficiently long history of transaction data. For example, if certain products are typically replenished once every three months, you’ll need at least three months of transaction data to identify this.


Unlike some of our other models, this model does not use any machine learning, simply business logic and behavioral data.

For each visitor and product, we measure the number of days between pairs of purchases–the replenishment period. We then take the median length of all the replenishment periods for that product. This tells us, on average, how long it takes for visitors to replenish a certain product.

Using this information, we can then estimate the replenishment date for each visitor that has previously purchased that product by finding the time of their last purchase and adding on the average replenishment period. In other words, this is when we would expect the visitor to next need to buy the product.

For example, a cosmetics company might find that visitors replenish the site’s best selling lipstick on average every six months. For a visitor that last purchased that lipstick four months ago, we would estimate that their next replenishment date is in two months' time.

Aspects of the model can be configured based on your specific business logic. For example, how many days before their replenishment date we should nudge the visitor. Your strategist will work with you to find the best configuration for the model.


The output of the model is the upcoming replenishment dates for different visitors and products. These are available through Derived Datasets and can be used directly in experiences or segments. Replenishment dates are updated on a regular schedule that can be configured as necessary.

Example use cases

The Replenishment model can be used to power Qubit experiences to provide reminders for visitors to restock products. Use cases include:

  • Send visitors an email when they have product replenishment dates coming up

  • Have notifications on the site reminding visitors to replenish products

  • Badge products that the visitor may want to replenish

For example, a cosmetics brand could use this model to remind visitors to replenish non-permanent make up products such as lipstick.

Getting started

To start using Qubit experiences powered by Derived Datasets, contact your account strategist and ask for the early access submission form. Submit this form specifying the model name as the project you are interested in.