Just Eat orders Google Cloud Platform for its data needs - ditches AWS

Profile picture for user Sooraj Shah By Sooraj Shah November 21, 2019
Summary:
Just Eat selected GCP over AWS and Azure for scale, speed and insights from data – and it’s now reaping the benefits

Image of chicken fries from Just Eat

Popular online food delivery company Just Eat had traditionally been an Amazon Web Services (AWS) shop, with its customer platform and customer facing apps all on AWS.

However, the company’s director of data platforms, Matthew Cresswell, explained that 90% of Just Eat’s data wasn’t being brought together or being used effectively by the organisation.

The challenge is you have 27 million customers wanting food and you’ve got 112,000 restaurants – how do you start to map the two together?

As the company wanted to become more data-driven, using algorithms, it set out to look at the different technologies it could use.  

Cresswell told diginomica:

We evaluated the three big players, AWS, Microsoft and Google and we considered the fact that we had a lot of real-time streaming data that we wanted to bring into the data platform that we wanted to expose, and we also had a lot of unstructured and semi-structured data, and this all led us to Google Dataflow and BigQuery.

The company then embarked on a 24-month project to re-platform its environment to bring all of this data together.

All-you-can-eat data

There has been a significant difference between the AWS product that the company was using before and Google’s products.

Cresswell explained:

Before we migrated, we were using an AWS product and every Monday morning, analysts and data scientists would cause a massive traffic jam and demand to get data, and the average query time would be 800 seconds, so people would start a query, go and get a coffee and then come back and get their results set and then may even have to query again if there’s something wrong.

Now, we’re using BigQuery, the average query time is down to 30 seconds, so not only are people getting data quicker than before, but all of that 90% of data that we weren’t ingesting is also being ingested – so the data estate is a magnitude bigger, and yet we’re still getting lower query times.

This has led to an increase in productivity, an uptick in employees using data – from 60 to 500 - and an increase in the amount of data that employees use.

A cultural shift

Cresswell, whose role it is to look after Just Eat’s data, machine learning, data governance and personalisation, emphasised that the decision to select Google wasn’t based on a strategy towards multi-cloud – particularly as a true multi-cloud strategy would enable the customer product or the data platform to fail over to the other cloud provider.

We looked at this on its merits and we were really going after serverless and the ability to handle real-time data and that’s what led us to BigQuery over things like Amazon Athena, Amazon Redshift, Microsoft SQL and Azure’s offering.

While the integration between different cloud companies is often an issue – this hasn’t been the case for Just Eat. Instead, the biggest challenges in introducing Google Cloud, have been the cultural shift internally, and also the weight of managing expectations.

Cresswell explained:

When you’re going on these transformational journeys, it’s not about a lift and shift because you’re adding more capability. For example, we went from a 30TB data estate to nearly a petabyte in a year, so it’s not comparing like for like. Therefore the challenge is explaining to the business that they need to be taken on that journey so they can start to get value from it as well – the technical piece is quite easy to an extent.

Two years on, and customers may not be aware of what goes on behind the scenes technically, but Just Eat’s work with Google on machine learning frameworks and a single customer view are giving consumers far more tailored recommendations than before.

They’re coming from Google, they’re flowing through AWS into Google, getting enriched and getting transformed and then pushed back.

Cresswell gives two examples of what this means in practice. When consumers open the app they will see a carousel of restaurants, including options for free delivery, no minimum order value, and these change throughout the day and are also personalised for individuals. These recommendations are calculated in Google.

In his second example, Cresswell explained that his favourite Indian restaurant had not been on Just Eat previously, but as he was an employee he remained loyal to his company and ordered from other restaurants that are on the app.

He said:

I didn’t know but my favourite restaurant came onto the platform, so the sales team signed the restaurant up and as a customer I didn’t know that. But as the platform analysed and understood my data, it knew I liked Indian cuisine and that I’m quite adventurous, so it pulled that one signal from 100 million events , picked me as a customer and packaged that into our CRM platform, and sent me an in-app notification on a Sunday to say this restaurant has been onboarded to the platform.

This is an example of the kind of journey Just Eat is starting to see as a result of the introduction into Google.

What’s next?

Cresswell was impressed with Google’s Contact Centre AI product, which has become generally available to customers this month.

The post-order space is just as important [as ordering food], and I think there is a lot of work we can do there to improve the experience and remove anxiety. So, looking for services like Customer Contact Centre AI is a big piece.

In addition, enabling customers to gain more transparency around machine learning use with the introduction of Explainable AI is of interest to Cresswell too.

I think GDPR has raised the bar in terms of how we use the data, tag it and make sure we know where it is. I think customers are expecting that value proposition back in terms of how we’re using the data to give them value and how we’re coming to those decisions. This way, people will feel like what we’re doing is really smart in the way we’re using data to give them their recommendations, rather than it being a black hole and people wondering what we do with the data.