How Gousto is growing its subscriber base using behavioral data created with Snowplow

Background

Gousto takes the stress out of grocery shopping and meal prep by delivering ingredients and recipes to customers’ doors. The business—which has been doubling year on year—had its sights set on further growth.

To achieve their growth goals, Gousto knew they needed to maximize return on ad spend, relentlessly improve retention, and keep delivering delicious meals. And to achieve these three things, they needed advanced analytics. Yet Gousto’s data team was being held back by disparate data sources and an inability to use the growing volume of data to its full potential.

Solution

Using Snowplow, Gousto can now access rich, granular, and highly-structured behavioral data from its various platforms, channels, and campaigns. And this consistent data stream is proving the ideal ingredient to power growth in three key areas.

1. Maximizing return on ad spend with accurate analytics

As a subscription business, Gousto devotes a lot of their marketing budget to advertising across multiple channels, including Facebook, Instagram, Twitter, and Google Ads.

To measure the return on each campaign, Gousto needed to know how long each of the customers acquired on that campaign stayed subscribed. Loyal customers who’ve used Gousto for years drive much more return than those who leave after one or two months.

Gousto needed data detailed enough to be easily segmented by marketing channel and campaign, and Snowplow delivered. Snowplow provided a rich, detailed data stream for each user, showing:

• Which campaign on which platform the user engaged with
• When they subsequently signed up
• Exactly how they browse Gousto’s recipe selection each week
• How many boxes they get delivered
• How long they remain a subscriber

This made it straightforward for Gousto to aggregate the data to calculate the real return on each campaign and use that data to optimize their spending across all campaigns and channels.

2. Relentlessly improving retention with a data-driven approach

Because Gousto is a subscription business, retention is a key driver of growth. The longer they retain customers, the faster they grow.

Gousto’s data science team built a model that predicted how likely it was for customers to remain into the near future, and assigned each user a retention score. They could then run A/B tests on different initiatives to see which ones had the biggest impact on each user’s likelihood to retain—in the most cost effective way.

The first version of their model used random forests. However, they moved to a deep learning model so they could feed it much more data without having to do expensive feature engineering. The deep learning model could also understand all the rich data that Gousto created with Snowplow—which all helped them figure out a user’s likelihood to retain based on:

• Web behavior
• Mobile app usage
• Engagement with email
• Transactional history
• Customer service conversations in Zendesk

The Gousto team A/B tested different campaigns with customers at different scores to understand the sweet spot for intervention, and the most effective intervention at each stage. They learned that customers respond better at different threshold points. Once a retention score got too high or dropped too low, the user was essentially disengaged and would “ignore” any of the attempted interventions.

Without Snowplow data, this would not be possible at all. We tried to do this with transactional data, but that doesn’t give you enough information. Really looking into customer activity is what actually gives you predictive information.”

DEJAN PETELIN | HEAD OF DATA, GOUSTO

3. Using data to surprise and delight subscribers with personalized recipes

Gousto’s service is geared towards people who care about food but are stressed by shopping and meal planning. Subscribers want two things from the service:

• Delicious recipes. If you’re not excited about the meals from Gousto, you’re going to leave.
• Convenience. You don’t want to sift through 30 recipes to find one or two that you like. After extensive testing, Gousto realized 5-10 recipes with a high likelihood to appeal is ideal.

It’s hard to deliver both of these. Reducing selection makes choosing recipes more convenient, but there’s the risk that users don’t see meals they like. Conversely, giving people more selection means they have more meals they’re excited about, but the experience is less convenient.

Personalization solves this tension. By personalizing the recipes customers see, Gousto gives each person an easier choice to make—as they only see recipes they’re likely to enjoy—improving the customer experience and driving retention.

Gousto’s personalization efforts were powered by a combination of behavioral data created with Snowplow and a graph database of their individual recipes built using Neo4j. Feeding the algorithm a combination of recipe data plus rich data from Snowplow about how users engage with each recipe before making their selections was critical. It ensured subscribers who like chicken don’t end up with ten different chicken dishes to choose from, and instead get just the right amount of variety each week.

Results

Gousto now has granular insights into which channels customers visit—and how often—before subscribing, and can optimize their marketing spend accordingly.

With Snowplow feeding data into its deep learning model, Gousto can join behavioral web and mobile app data with their email, Zendesk, and transactional data to predict churn and offer customers the right interventions at the right time.

And, most importantly, Gousto can continue to delight customers with helpful, delicious recipe recommendations that are personalized to their tastes—while still offering up enough variety to keep them engaged.

How you can get started with Snowplow

To learn more about how Snowplow can empower your organization with behavioral data creation, book in a chat with our team today. Alternatively, Try Snowplow is our free, easy-to-use version of our technology, which allows you to create your own behavioral data in under 30 mins.

‍