Engage use a recommendation engine to drive relevant product recommendations to your store. The type of recommendation vary a bit depending on the page your visitor is viewing. This is mainly because the recommendation model needs input to function properly. E.g. first time a new visitor browse your start page, the model doesn’t know anything about the users behavior and can therefore not provide a personalized recommendation. But as the user interacts with the website the model picks up on behavioral patterns and are then able to provide better recommendations. The following illustrates how information is added to a user in order to provide recommendations relevant in each step of the journey.
The initial visit to the website allows only for use of high level variables such as time of visit or geographic region etc. These indicators are usually considered weak and don’t generally provide enough information to recommend relevant products for an individual user. They may however still outperform the options of not recommending any products.
When the user starts interacting with the website, like browsing products or adding products to cart, the model ingest information that can be used to compare this visitors pattern to earlier visitors and thereby extract possible products of interest for the user based on that pattern.
Once the user reach checkout the model has a pretty good set of information about the user that are used to recommend upgrades or additional products at time of checkout. Since checkout often requires some kind of identification of the user, the visitors historical purchases can also be utilized here if any purchases was made prior to this one. Following the order completion, the user can be re-targeted with product recommendations via email or advertisements based on specific customer segments.
There are a few reasons why recommendation engines generally outperform manual selections of recommendations at scale. The first one is simply the scale and speed of which a recommendation engine can produce relevant recommendations for all products in store, not only a select few. And it can maintain it in near real-time updating it as trends changes or seasons shift.
Secondly, the model introduce less bias on what should be recommended or what “goes” well together. The model simply look at what’s actually been sold together and based on the patterns and behaviours are likely to be sold together next time. Furthermore, the models can learn from their prior recommendations and adjust the next recommendation for a particular product based on the historic outcome. All of which happens automatically each time the model is retrained.
Recommendation engines are a powerful way to increase revenue, and engage makes it easy to use. Get started now