The Data River – How Data Flows Through the HelloFresh Production Process
HelloFresh is a company with tech at its core. Technology holds all the strings together: HelloFresh’s tech unit, called HelloTech, not only continuously optimizes the customer experience and makes it more seamless, it also coordinates all operations and drives efficiency and innovation from marketing over menu planning to supply chain, logistics and operations. Today our global technology department is one of our fastest growing teams, with more than 1.000 engineers, designers, data engineers, data scientists, product managers and analysts working from six global HelloTech hubs.
Data is HelloTech’s most precious resource and powers and accelerates every part of how we deliver on our mission to change the way people eat – because first and foremost we operate in a data-driven way. Data is not only a resource for technological improvements and scaling our business, but is also the base of every team's day-to-day work. All business decisions, including our product innovations, are based on the analysis of millions of data-points to build a desirable product for customers. The HelloTech Data Alliance is in charge of fostering HelloFresh’s data maturity and therefore an important driver to deliver our delicious meals to our customer’s homes.
How data serves our customer’s meal kit orders
During the production of our meal kits data plays an important role in every stage of the process. Similar to a river the flow of data starts with a spring (the recipe creation) and after many currents and bends reaches the customer, with each stage having a significant downstream effect. When producing a meal kit there are multiple different stages where data gets generated, processed and used.
In the first stage our recipe development team creates new recipes. This is a continuous process that has a high impact on customer satisfaction. Our customers value the broad and varied choice of inspiring recipe ideas that we provide them with and our aim is to continue to increase their choices. The recipe development team uses historical recipe data (including data from existing recipes, production costs and customer satisfaction scores), new product ideas (including seasonal ingredients like pumpkin) and the available ingredients on stock and in season. Continuous experimentation has allowed the HelloFresh recipe database to grow to more than 14,000 recipes over the last ten years with many influences from different cuisines all over the world.
Every week our customers get to choose from a variety of chef-developed recipes.
HelloFresh uses in-house developed algorithms for optimizing the production process
A couple of months before a certain delivery week the menu planning starts. The recipes are curated into a menu of between 18 and 45 recipes (depending on the market). The menu creation is powered by HelloFresh developed algorithms which strike a perfect balance between maintaining and increasing customer satisfaction (measured in recipe scores), lowering ingredient costs and maintaining diversity in our menu offering. Over the years our tech-team has refined these algorithms to constantly lower procurement costs for HelloFresh while increasing customer satisfaction at the same time.
Another in-house developed algorithm forecasts how often each recipe will be chosen in a certain week. Based on historical order data our tech team can develop a close approximation of the final order list and the quantities of ingredients needed to satisfy customer demand. This allows HelloFresh to order the correct volume of ingredients and add-ons weeks ahead of production, making sure that all items are in stock at the right time. The algorithm also avoids over-purchasing, allowing us to optimize our lean,make-to-order business modelwhich inherently reduces food waste. A recently published research paper underlines that cooking with HelloFresh substantially reduces household food waste by 38 percent in comparison to traditionally cooked dinners.
Early demand data from menu planning and our in-house developed algorithms allow us to predict the real demand for any given week. This leads to less food wastage and lower CO2-emissions compared to the supply chain of a traditional retailer.
We can see how close the forecast was to reality when customers submit their orders. Orders can be changed up to four days before the delivery date (depending on the market and the specific date), which is why the precision of the forecast is so important and our algorithms are constantly refined. While our customers take just a few minutes to choose their recipes, in the backend we combine their choices, their subscription data (e.g. how many active customers there are in a given week) and the delivery data (where and when do orders get shipped). At the end of this step we know exactly how many boxes were ordered and how many meals and how many add-ons were included. This data gets forwarded to production where boxes are packed in one of our production facilities around the globe.
Delivery and avoiding mistakes on the last mile
The next step of our value chain focuses on getting the finished boxes to our customers. In most countries (except for the Benelux) we do not deliver ourselves but we use third party delivery providers. In this step we create the shipping info, shipping label and tracking information. We also track when a box was packed and how long its content will stay fresh to ensurefood safety at all times. We use the tracking info we receive from our delivery providers to constantly update customers on their state of delivery and the expected delivery time.
Finally, we ask our customers to rate the recipes they receive. Their feedback and their historical order data get fed into our databases and the production process begins anew with the recipe creation.
At the end of the process is a delicious meal and a happy customer.