Introduction to product recommendations

Qubit’s Recommendations uses the power of machine learning (ML) to learn patterns in the product catalog and visitor buying and viewing behavior and then create recommendations tailored to the visitor and the moment in the journey.

It includes cross-sell and upsell capabilities and enables merchandisers to leverage data in the Data Store to personalize the user’s journey, influencing engagement, and purchases.

At Qubit we offer strategies, defined as recommendation endpoints, designed around specific business goals. A strategy for example can be focused on increasing engagement or increasing order value.

Qubit’s plug-and-play strategies are continually optimized by Qubit’s Data Science team to learn best practices for your business and deliver insight that can be fed back into your overall business strategy.

The outcome of each strategy is contingent upon the context where the recommendation is placed. The context where the visitor sees the recommendation is most commonly defined by a page type or channel.

We believe it is important to test and evolve each strategy within a particular context, for example page type, as different points in a visitor journey on a site may merit different strategies for product recommendations.

For example, the strategy to attempt to increase order value on a product page may be different to the one later in a visitor’s journey on a basket page. In testing and evolving our strategies, we exercise control over the strategies and context, thus ensuring we deploy strategies at the point in the visitor’s journey where they work best. This maximizes the positive impact of recommendations on a website.

Each of our strategies comes with a specific combination of algorithms, configurations, and fallbacks and with the exception of trending, popular, recent, and new, each one requires information about the currently viewed product on a page. If that information is not being sent, for example, in a setup that calls the Recommendations API directly, recommendations will fall back to most popular products across a site over the last 30 days.

Refer to the following table for details of our recommended strategy use:

Qubit offers the possibility to extend recommendations through the creation of custom rules. Rules help improve product recommendations by adding a business logic layer that is overlaid on top of our algorithmic strategies and enable marketers, who are often best placed and have the deepest product knowledge to fine-tune recommendations to better align them with key strategies and business goals.

See Using Marketer Controls To Further Customize Recommendations for more information.

For those interested in using the Recommendations API to deliver rules specific to each request, see Per-request filters.

Qubit’s Recommendations API offers the possibility to deliver more powerful and personalized recommendations experiences. See Deliver recommendations (Qubit) for a guide to using the API.

We also offer the possibility to deliver product recommendations using one of Google’s recommendations strategies. You can find out more in Deliver recommendations (Google).