To provide truly personalised recipes each week, the data fed into the algorithm needs to be of high quality and relevance. During onboarding, new customers are asked a set of questions that shape their ‘taste profile’—protein preferences and general likes & dislikes about meal types. 

Our focus was not only on data quality and relevance but also the framing and language used for the onboarding questions. Putting ourselves into the customer's shoes was critical in order to balance both the algorithm and the customer needs at this point in the new customer's journey. The onboarding flow appears right after a customer has paid—when expectations are at their highest. We focused on striking the right balance between making the customer feel heard and in control, supplying the algorithm with relevant data and realistically representing the power of our recommendations.

In collaboration with our data analysts, we identified the most valuable explicit data points that would serve the algorithm best from day one of a customer’s journey with us. We conducted some qualitative research and further identifying some key factors that customer’s found important when making choices about what to eat from our menu. This combination gave us a strong idea of what to prioritize from both the business and customer perspectives.

During the design phase, we started broad and experimented with various different user flows, focusing on a wide range of customer types. We split out all the data points into different narratives—some more granular than others. However in the context of iterative enhancements, we made the decision to prioritize the most valuable data points to begin with, while still keeping the framing of each step feeling as though the customer was making a strong contribution to their personal weekly menu.

I also wanted to further dig into the idea of continuous data collection. A strong onboarding can only go so far. What’s more powerful is behavioural data and opinions after having some experience with a product. I explored the idea of a ‘check-in’ which would prompt new customers with options to refine their taste profile after their second box had been delivered. Based on what we know about the customer's preferences after two weeks with us, we could tailor this checkin dependant on which segment they fall into (eg. if a customer had ordered all meat recipes for the first two weeks, we could prompt to understand on a more granular level, what type of meat cuts they prefer to eat.). The visual below is of a customer's order page within their account—the check-in UI element being placed after their second box. This was conceptual and remained in a wireframe state to communicate the idea with the team and asses how this may scale into something more systemised that could support ongoing data collection for all customers, not just new.