Turning Food Waste into a Byte-sized Problem

The Olympic swimming pool; the double-decker bus; the football stadium. Whilst researching stats to include in this post, it saddened me to discover that these are all units commonly used to quantify the UK’s food waste. As for where it’s all coming from, studies (for example, this study by WRAP) suggest that retail and household food waste accounts for around three quarters of the UK’s overall post-farm-gate total. Given Gousto’s rapid growth, it is now more important than ever for us to consider our impact in terms of food waste. So what exactly are we doing to help? In this post, I will attempt to answer that question, focusing particularly on our strong investment in data science and analytics.

Challenging the Status Quo

In terms of environmental footprint, Gousto wins on several counts when compared to traditional weekly supermarket shops. As our CEO, Timo explained in a previous blog post, for decades, supermarkets have stocked up their shelves and then attempted to shift that stock. This ‘shoot first, ask questions later’ approach has lead to an astronomical amount of food being diverted from dinner plates straight into landfill. In addition to the fact that we deliver perfectly measured ingredients (so no more 2 kilo sacks of potatoes to make a side of wedges), to address the issue of food waste, we also harness the power of data.

How are we doing it?

A strategy we could use to guarantee our factory zero food waste is to simply wait for customers to place their weekly orders and then purchase the exact quantities of those ingredients from our suppliers. Unfortunately, to do so would result in a more lengthy wait for each customer to receive their box. Instead, we make use of Amazon Redshift, a petabyte scale data warehouse. Unlike the large supermarkets — who first built their brick-and-mortar stores and then integrated data platforms as an afterthought — Gousto was built with one from its inception. We use Redshift to store data on everything from the nutritional content of our ingredients, to website clicks to customer order history. This means that from menu to menu, we are able to look at customer recipe choices and use clever analytics to identify trends and explore their potential drivers.

But how does data allow us to reduce food waste? In the same way the MET Office crunches meteorological data to make predictions about next week’s weather, our in-house forecasting algorithm uses data from past orders to predict the quantities of ingredients we will need to fulfil orders over the coming weeks. We currently use a random forest algorithm to predict the uptake of each recipe based on attributes such as cooking time, spice, protein type and cuisine. Doing so ensures we can have the exact ingredients ready to ship as soon as orders come in.

As well as reducing our own food waste, accurate forecasting acts to reduce the incentive for upstream manufacturers to overproduce. When you consider that the latest estimates put the UK’s pre-farm-gate annual food waste at around 2.5 million tonnes, I think we can all agree that businesses like Gousto have a responsibility to get their forecasts right.

We’ve come a long way since 2012, when we offered two weekly delivery slots; you can now choose from fourteen slots spread across all seven days. Despite this leap in operational complexity, since September we have achieved a staggering 80% reduction in food waste1. This was made possible thanks to some impressive collaboration between Gousto’s Planning, Data, Finance and Fulfilment teams — great work!

To help us solve exciting problems like this one, Gousto’s Data Team is growing rapidly. If you’re interested in joining, we’d love to hear from you! Keep an eye out for the latest vacancies on our jobs page or drop us an email at analytics@gousto.co.uk.

Edward Cardy
Data Analyst

Originally published at techbrunch.gousto.co.uk on December 20, 2017.