How and When are Results Personalized?

Here at Constructor we’re big believers in applying personalization only when the data shows it’s a good idea.  It's hard for humans to predict the areas where personalization will be most effective. This is why we use machine learning to find the optimal situations to apply personalization. Sometimes this personalization is highly visible and expected while other times it may not appear in places you expect because the data does not support its use in that particular context. What it always does though is lead to an increase in your key business KPIs like conversions and revenue. 

We are constantly collecting clickstream data on what a customer has clicked on, viewed, added to cart, purchased, and a host of other actions available in the clickstream. The system then looks to see if this data is meaningful and predictive of other actions. For example, just because someone clicks on a women's shirt doesn't mean that they plan to purchase only women's shirts.  Or, if they purchased a particular brand of taco sauce that doesn’t mean they are loyal to that brand in every possible context in which the brand is exposed in the result set.

Constructor is constantly analyzing and learning over time so you may see some obvious personalization in places where the data strongly suggests items based on a user’s behavior. Other times you may not see the personalization if the data does not show it as having any effect. 

We use the current level of personalization because the AI has learned to not personalize just for personalization's sake, but to save instances of personalization for those situations where it leads to improved KPIs. We could apply “more personalization” to make the effect more visible in additional, or even all contexts, but AB tests we’ve run have proven this leads to lower revenue for our customers; and we view everything we do through the lens of how it affects our customers’ KPIs.  

We have a set of model parameters that affect personalization’s impact on users.  Our optimization engine is continuously fine-tuning the parameters for every dataset of our customers to achieve the best result in the KPIs that matter to them most. 

Shoppers will receive a personalized experience through our search, browse, collections, autosuggest, recommendations, and quizzes products.  Even the filters and filter options are personalized!

Due to personalization, the order of items on site may not always match exactly what you see in Constructor’s Interact tool. Keep in mind that any slotting rules curated by your team will remain in place and will not be impacted. To further inform the algorithms, we recommend sending us additional data you have on shoppers which you think could be helpful.  In addition, leveraging Constructor’s quizzes product to gather zero party data will drive further personalization.  Zero party data is the most customer-friendly form of raw data there is because it’s data your customers actively want you to have about them!