Using data to personalize culinary experiences for customers at Blue Apron

Pooja Bathia, Senior Data Architecture Engineer, Blue Apron

Aug 13, 2020

Have you ever wondered how Blue Apron manages to deliver millions of “moveable feasts” — delicious, healthy meal kits, complete with step-by-step recipes and ingredients — to customers across the country every week?

For several years, Blue Apron has leveraged the combination of Google BigQuery and Looker to power organization-wide access to data. You can learn more about the details of our migration to BigQuery here and about our tech stack and implementation outlined in our case study with Google Cloud. Recently, the Google Cloud teams reached out to ask us about the new ways we’re leveraging data to efficiently serve and respond to our customer preferences, as well as hear some of the tips and tricks we’ve learned over the years. You can read below and watch this webinar to learn about some of the newest, innovative ways I’ve had the pleasure of witnessing our Blue Apron teams leverage data to enhance our customer experience, along with some useful things we’ve learned throughout our journey with data.

Increasing customer satisfaction

We at Blue Apron are always committed to providing the most positive customer experience. With data, we’re able to take that even further by visualizing the entire customer journey, drilling into and rapidly resolving issues, and learning from feedback and best practices to improve our processes.

In the past, the Blue Apron customer experience team worked primarily out of Zendesk’s customer service platform. Now, with Zendesk data piped into BigQuery and modeled in Looker, the customer experience team is empowered to create and access dashboards where they can easily visualize, drill in, and explore different parts of the customer experience. Our customer service supervisors also use this modeled data to track the performance of their teams. This combination of metrics about our internal teams and customers has helped the customer service teams set long-term goals and monitor their progression towards reaching them on Looker Dashboards.

When it comes to customer satisfaction, one area we monitor closely relates to the issues our customers report. The most common issues reported by our customers are categorized as shipment, recipe, and ingredient issues. To make this data serve as launching points for our teams to take action on, our Data Architecture team developed models in BigQuery and plugged those into Looker, empowering our customer service agents, analysts, and business teams to monitor issues and create action plans for other teams.

Now that they have more visibility into these issues, our customer experience team can quickly identify problematic trends related to different shipping carriers, recipes, or suppliers. Directly from the dashboards, they can further slice and dice the data to ask deeper questions about these trends. For example, customer experience agents can surface the most prominent issues by facility, cities, recipes, ingredients, and other parameters. Images of problem orders are linked to the orders in Looker; if there is a shipment issue, the agent can view photos or videos (such as a delivery box that was damaged in transit) directly from their Looker account. The agents can then share this information with the logistics team, enabling them to work with our shipping partners to remediate the current problem and prevent similar ones in the future. The logistics team can also link customer feedback in Zendesk to trace it all the way through to warehouse and shipping operations to find the root cause of the issues.

Data to please every palate

Apart from providing on-time delivery services, we are also attentive to quality and customer tastes. By leveraging recipe reviews, we’re able to get a deeper understanding of the likes and dislikes of our customers.

Delivering quality through the supply chain

For instance, perhaps multiple customers file ingredient issues for carrots at a certain time. Leveraging the tools we have, we can further investigate this issue, ask questions of the data, and discover why a particular shipment of carrots was unappealing. This helps us track and monitor the quality of our entire food supply chain, and helps us ensure that we continue to meet our Food Safety and Quality Assurance (FSQA) compliance requirements.

Delivering quality through information

Delivering accurate information plays a crucial role in making sure that our customers receive exceptional culinary experiences. As an example, say a group of customers didn't care for a particular recipe because they found it to be spicier than advertised, or because they are vegetarian but forgot to include that detail in their account preference settings. Our culinary team wants to know these things. In fact, they read every single recipe comment each week because they’re so dedicated to understanding every customer experience. And because of the organization-wide data access we have, what could’ve been a complex issue to solve is instead easy for our teams to address and resolve. In this case, using data is able to help with merchandising, enabling our culinary team to ensure that recipes are accurately represented on the website and include information like the level of spiciness or availability of a vegetarian option.

Our culinary experts also team up with the customer experience team to provide customers with the right information and make sure they have a superior experience. For example, if someone complains about not receiving a vegetarian meal, our culinary team shares this information with our customer experience team, who then take the action of updating the customer's profile to reflect their specific needs and tastes. With this level of discovery and sharing available, we’re able to listen to customer feedback and take meaningful actions like updating recipes and planning future meal choices for our customers — which is especially important since we add new recipes every week. We need to quickly know what was a success, what wasn’t, and what improvements can be made. And with data, we’re able to do exactly that.

Wine with your dinner? Honing in on customer tastes with Blue Apron’s eCommerce marketplace

In addition to shipping meal kits, we also have an e-commerce marketplace where customers can purchase products to enhance their in-home culinary experience, including items like aprons, wines, olive oils, and spice blends. One of our most popular items is our extensive selection of wines from well-known vintners that customers can pair with their meals. Using Looker, we’re able to analyze the purchasing trends of our customers, cater to their needs, and identify new opportunities for expansion and product strategies, which recently has included offering à la carte items to our customers.

Prior to using Google BigQuery and Looker, our e-commerce teams relied on manually updated Google Sheets and struggled to get a handle on exactly what products were selling well. After centralizing this data in BigQuery and giving the team access to visualize and explore in Looker, the entire marketplace business could be seen from end-to-end. This gave multiple teams access to things like a simple-to-use wine dashboard, where they could get an accurate picture of what types of bottles sell well and which of our curated bundles are the most popular. Using BigQuery and Looker in this way also enabled us to connect customers across our meal and wine subscriptions, as well as with our marketplace, to get a deeper understanding of what customers want and what’s working best to meet their needs. Adding to this, our Director of Data Engineering, Michael Collis, shares “For the first time, we are able to take insights on customers and act on them. The Looker and BigQuery solution has been instrumental in enabling us to access valued-added derived data so that we can make smart and practical decisions. This is the best way to scale our team so that we provide internal stakeholders with the information they need to drive the business forward.”

Data to pack efficiently

Our small-but-mighty Data Engineering team is always looking for new ways to offer data-driven innovation. Recently, when we sat down with the packaging team to learn more about their workflow, we discovered that they were spending an inordinate amount of time manually tracking numbers in spreadsheets. We knew we could help them with this, so we quickly modeled the data in BigQuery, plugged it into Looker, and I presented them with reports to help solve their current needs while optimizing their process. Now that they’re not spending hours reconciling the numbers, the team is able to use this data to look for ways to further optimize and reduce the packaging needed to protect ingredients.

A tech stack to support the data-hungry masses

In the years since deploying BigQuery and Looker, we’ve experienced many situations like those mentioned above: everyone has an appetite for data. Once we have a team onboarded to Looker, we don't just expose reports and data sets and leave our users to figure the rest out. We make sure they are equipped with accurate tooltips, data definitions, Looker-focused group chats, links to relevant shared folders, and boards to help them wield the full power of Looker. Andrew Rabinowitz, my teammate and our technical lead for Data Architecture, attests to the efficiency gains we were able to realize with Looker: “Once we made the switch, we never looked back. Looker scaled with us and is flexible enough to evolve in-step with our business.”

To continue feeding our data-hungry workforce, we also need ways to empower our teams to perform ad-hoc, what-if analysis and review the results easily. To do this, we leverage Google Sheets Data Connector to allow users to tap into the power of BigQuery for speedy, accurate analysis, while relying on the familiarity of spreadsheets. This has been a game changer: it is so fast and easy for analysts to pull modeled data from BigQuery, add their assumptions, and produce results. We are also excited to start leveraging the recent addition of scheduling for the refresh cadence of these connectors. The sleek integration of BigQuery and Google Sheets is instrumental in breaking data silos and helping us build more trust in data throughout our organization.

This connectivity across tools and teams allows us to better understand our customers and ensure we’re meeting — and exceeding — their expectations with every interaction.

With all of these practices combined, our stakeholders are able to get the most out of Looker and BigQuery. We know it’s important to not just give our users access to data and tools, but also to support them as they become more data-savvy, learn to ask more complex questions, and better integrate insights from data into their workflow. And because we’re able to do this, our entire company is happier and more efficient with access to data.

Next on the menu

With powerful tools like Google BigQuery and Looker in our arsenal, we feel confident about building a strong data culture at Blue Apron. The speed of analysis and reduced time to insight has proven to be influential in uniting different business teams and making data-driven decisions. We are currently working with events streams from Kafka that will enable us to provide more real-time analytics to our data users. With a strong foundation of data warehouse and business intelligence platform already in place, we are excited about the future of our data discovery and insight mining platforms.

A happy Looker customer,
Pooja Bathia
Senior Data Architecture Engineer
Blue Apron


To learn more about how to streamline operations and hone in on customer preferences with data, check out the Looker Solutions for Retail eBook.

Next Previous

Subscribe for the latest posts