Business & Innovation
- July 16, 2018How our Discovery Team creates a special menu for every customer
Delivery Hero partners with more than 190,000 restaurants worldwide to offer a tremendous variety of cuisines and dishes across our global platforms. Serving our customers with food that fits their exact taste, lifestyle and mood can be quite a challenge – but Delivery Hero’s Search & Discovery Team is eager to take it on!
We analyse food right down to ingredient level to create personalised recommendations
Instead of providing customers with a static catalogue featuring hundreds of restaurants, we look for ways to make their experience on our platforms personalised and enjoyable. Customers typically order from the first few restaurants in the listing. It is therefore important to ensure that the top restaurants are excellent matches for them.
But how do we know what they are looking for? It’s all based upon rich data and algorithms, says Gugulethu Ncube, Director of Search and Discovery. “We take a very close look at the customer’s taste and order behaviour to adjust the restaurant ranking accordingly,” he explains. “Therefore, we’re analysing and classifying our restaurants on cuisine, dish and even ingredient level.”
Gugulethu and his team engineer algorithms which personalise the customer experience through content-based recommendations. Having started with our foodora platform, they are rolling out smart recommendation algorithms to more Delivery Hero brands. Since the personalisation was launched, foodora’s conversion rate of mobile customers has increased by over ten percent.
Different customer, different taste, different ranking
In a recent blog post, we talked about how the operational performance of restaurants affects the ranking that is provided to the customer. Customers should have an enjoyable experience and not face a negative interaction, which could make them unhappy. Taking this into account, only the best restaurants rank at the top and can be recommended to customers.
Several features are part of the recommendations algorithm, with key attributes being:
- Cuisines and Taste
Through past orders, we learn about customer preferences like cuisines (e.g. Mexican food), taste (e.g. spicy) or dietary restrictions (e.g. vegan). All individual customer attributes will then be matched against our restaurant offerings. - Budget
We will boost restaurants that match the customer’s budget spending habits. For instance, high-budget restaurants are ranked down for customers who frequently order from low-end restaurants. - Food Type and Ingredients
The content-based algorithm also takes into account what type of food has been frequently ordered, including the corresponding ingredients.
In order to ensure data security, no sensitive data is stored and all processes are GDPR compliant. We want customers to have an awesome experience, but we want to do that in an ethical manner and use as few data points as necessary. As Gugulethu adds, “We want the customers to think: ‘These people know me’ – just as in your favorite restaurant.”
How we find the most delicious pizza margherita
We know a decent pizza margherita will satisfy our customers, but an exceptionally delicious margherita will make them really happy. That’s why the recommendations shown to users are not only affected by restaurant performance and individual order history, but also by order behaviour of other customers.
If a customer ordered a pizza margherita from two different restaurants and then re-orders it from one of these places, their margherita probably tasted better. We can be even more certain about that fact as other users follow the same pattern. These data-driven insights allow us to provide more relevant recommendations to customers and to upgrade them to the best possible experience.
“Through our services and recommendations, we can influence what millions of customers eat and how much they enjoy it”, Gugulethu says. Offering personalised recommendations will increase convenience for our customers by saving their time and providing only the highest quality. Gugulethu further adds: “For me personally, the restaurant recommendations given by foodora have become my favorite point for selecting a restaurant when I order food online. Sometimes, I feel like the app knows me even better than I do.”
Share