Improving User Satisfaction Via Better Recommendations

Gobble

SUMMARY

  • Gobble offers a number of features in their meal kit delivery service that make them stand out from their competitors.

  • Hop was engaged to evolve Gobble’s recommendation system as a path toward increasing customer satisfaction and sales, and reducing churn.

  • There are many factors to incorporate for Gobble that are atypical of traditional recommendation systems.

  • Our approach included modernizing the legacy infrastructure and testing a myriad of techniques before moving forward on a system aligned with Gobble’s business aims.

  • Multiple generations of Gobble’s new recommendation system have been rolled out, and work continues to improve the experience for end users.

THE COMPANY

Gobble, a pioneering meal kit delivery service, is founded on the principle of making home-cooked meals accessible, convenient, and enjoyable. Gobble's disruptive business model and innovative approach have set it apart in a crowded industry – unlike traditional meal kit services that require extensive chopping, dicing, and prep work, Gobble's meal kits come with pre-prepped ingredients, reducing cooking time to a mere 15 minutes. Moreover, Gobble’s menu caters to diverse dietary preferences, and their subscription-based model allows customers to skip weeks or adjust their meal plan to accommodate changing circumstances. Overall, the convenience, choice, and flexibility Gobble offers is reshaping the way people approach cooking in their daily lives.

THE CHALLENGE

Gobble approached Hop with an interesting challenge: could we improve key business metrics – customer satisfaction, sales, and churn – by enhancing the quality of their weekly meal recommendations? 

Unlike traditional recommendation systems, Gobble’s needs to account for many unusual factors. Gobble offers only a few dozen products at a time, many of which have never been seen before. There is an influx of new customers every week, and very little is known about them. The system isn’t just recommending next steps for the customer experience that can be skipped if necessary – the recommendations actually often end up being shipped to real customers to eat. Customers have very specific and nuanced dietary restrictions, and sometimes they’re surprised by enjoying a meal they wouldn’t have thought to order for themselves. Lastly, experimenting with new approaches affects the procurement, culinary, and fulfillment teams in significant and expensive ways. And then the nature of the product experience means we don’t know if a change was worthwhile for several weeks.

There was an existing recommendation system that had served Gobble well so far, but it hadn’t kept up with their operational evolution over the years. Exciting product changes required fresh thinking, and the recommendation system was just the first of several key areas in which Gobble was hoping to deploy ML.

THE APPROACH

Hop’s approach was to first add stability and clarity to Gobble’s legacy system, then to develop the next generation of recommendations. Hop engineers collaborated with their Gobble counterparts to modernize the legacy recommendations infrastructure, enabling new techniques to be rigorously A/B tested. Meanwhile, Hop researchers formulated Gobble’s business challenges into a series of specific technical opportunities. 

After experimenting with a constellation of techniques and features, the team settled on principled and promising approaches towards addressing the myriad challenges that Gobble’s recommendation system posed. Such a complex project required going beyond mere best practices to develop a baseline system – Hop’s expertise paired with close client collaboration allowed us to diverge from best practices when we understood it would lead to better outcomes for Gobble. 

THE RESULTS

Our collaboration with Gobble is ongoing. Together we have rolled out multiple generations of recommendation systems to customers, added rigor to the experimentation stack, and modernized the infrastructure underpinning the various ML initiatives within the company. We continue to make progress towards Gobble’s aims to increase customer satisfaction and sales and have begun to apply ML for optimization of internal operations as well.

Are you looking to leverage ML to improve your operations and business metrics? Contact us to learn how Hop can help.