How Deliveroo uses data to deliver food whatever the weather


Deliveroo has picked the Looker data analytics platform to better map out takeaway trends and maximise capacity and sales at partner restaurants.

The software helps the food delivery giant translate vast quantities of data into actionable insights that can be used across the business, from identifying general market trends to assessing marketing efforts. It can also analyse how weather affects customer orders.

"People order more during rainy days, but by how much is a very difficult question," Colin Lee, a business intelligence analyst at Deliveroo, tells Computerworld UK.

"We have a lot of data scientists and BI analysts to understand the difference, and also to understand what people want when the weather is bad."

Tastes follow the whims of the weather: people tend to order ice cream and cold drinks to cool down on sunny days, and comfort foods such as mac and cheese when it's wet and grey outside.

Deliveroo uses Looker to build different attribution models to understand these effects. It then communicates the findings to executives and restaurant account managers, who can use the insights to quickly make data-driven decisions.

Read next: Deliveroo ditches Amazon's RedShift data warehouse amidst rapid growth

This helps Deliveroo handle swings in demand based on weather forecasts. Preparations can be adjusted to account for variations down to the millilitres of rainfall.

"The tools enable us to get the subtlety and nuance of the equation out better," says William Sprunt, head of business intelligence at Deliveroo.

"We don't need to say it's rainy and we're probably going to get 20 percent extra people ordering. We can actually feed in a number of attributes and see that in the past this has had this effect on our behaviour, and it's likely that we'll see this sort of effect in the future."

Using data at Deliveroo

Deliveroo is data-driven and also data-hungry, an appetite that extends from the tech team to the boardroom.

Before turning to Looker, Deliveroo staff would need to analyse the data by writing a new SQL or Python script. They would then extract the information from the database and into a CSV file and sent it as an email attachment.

Looker allows them to explore it directly through a visual dashboard and share the information with a quick link to a URL.

The product comprises 'Explores' that act as starting points for queries and correspond to a table in a database. Users can drag and drop across the interface to find the information they need and then display it in pre-built visualisations known as 'Looks'.

Read next:Why Deliveroo continues to suffer from server outages

"Looker is most important for giving the stakeholders a way to interact with the data," says Sprunt.

"I wouldn't say that Looker is a pure data science tool by any means. There are plenty of other things that fill that niche in the marketplace a little bit cleaner, but to be able to get a really quick view of data it's really useful."

Multiple Looks can be combined in a single dashboard to identify the connections between pieces of information and add context to results. This can help with decisions such as how to balance the trade-off sales volumes against profitability at any given time.

Deliveroo's hunger for data

Deliveroo tops Deloitte's annual ranking of the UK's fastest-growing technology companies and recruits new staff every week that can benefit from data analysis. Looker makes this job straightforward.

Operational managers can go to Looker to see what's going on across the city, while data scientists can analyse details such as the impact of rain on the orders.

"We don't do this yet, but you can imagine a situation where you can abstract and aggregate that weather data up into a place where you can get a useful country level summary of how many different customers may be affected by rain," says Sprunt.

"Or you could go down to the level of whether it was raining in a neighbourhood at this hour, and have that data all be fed by the same base case."

Read next: How TfL is using predictive analytics to keep the Underground running

The biggest business benefit has been giving analysts more time by removing the need to manually code reports, dump the information into Excel and send it on in emails.

They can use the extra capacity to develop deeper insights from their data.

"It's not enough to say yesterday we had a whole lot more volume than we did the previous week," says Sprunt.

"You now have people free and with the mental capacity to know this was driven by this promotion or this restaurant type or this weather condition, and provide that next level of insight back to the business."

Forecasts can vary immensely between both meteorologists and weather watchers looking out of their windows.

Deliveroo uses Looker to integrate the different predictions into a single source that it can use to guide strategy.

"What you'll end up with if you're not careful is a situation where a bunch of people will download all of this data from Looker and combine it with a weather report which they like for their city," says Sprunt.

"Looker gives you the ability to come together and have a source of truth for weather data. Let's agree on how we pair it up with our order volumes and understand that relationship and then that's a single source of truth for everyone."

Why Looker?

"I feel that Looker occupies a space of its own in the marketplace," says Sprunt. "Data tools tend to either fit into these sort of SQL runner plus visualisation space, or you have to build everything yourself as a BI team. There's not very much I would say that's in this middle ground of flexibility."

Sprunt is also enthusiastic about the support Looker provides and the speed of development of new features, such as dynamic filtering, but admits that Looker isn't for everyone.

Read next: How ZSL is using Cloud AutoML to prevent poaching with automated image-tagging

He advises others be very clear about the end point they want to reach and then pick the best tool to help them get there.

"This is definitely the one which fits us the best, but if you're in a company where everyone is a hacky analyst who loves playing with SQL there may be a different tool for you.

"If you're in another company where people don't even know how to use a pivot table, and what they want is to have a pretty chart every day that's the one piece of information they'll consume, there will also be other tools which will be better. But if you're in this space of democratisation of data I think Looker is an excellent choice."

Copyright © 2018 IDG Communications, Inc.

Shop Tech Products at Amazon