Autodata turns to big data to predict vehicle failures

Limitations with legacy IT

Autodata began a major project to develop portals, known as application programming interfaces (APIs), that would enable its customers build their own applications to make use of Autodata’s vehicle data.

“We had very limited ways to get raw data out of our systems, ” said Brooks in an interview with Computer Weekly. “It was not a real web service, it was not granular, it was not well structured, and it was difficult to use.”

“On one level, it’s quite simple, we are taking data and displaying data. You don’t need a heavy layer of intelligence”

Neil Brooks, Autodata

And because there was no readily available documentation for the data site, it was difficult for developers to create new applications.

“In the past, customers had to contact us, and there was a finite number of people in the organisation who could help them, and it would take time,” said Max Lienard, head of product development at Autodata.

The company also realised it needed to add some security controls to make sure its customers did not accidently crash the company’s computers.

“Historically, if a customer did something unusual like make millions of calls per second, they could bring down our system,” said Lienard.

From Microsoft to open source

The company began by replacing its ageing Microsoft-based hardware and software with modern open source alternatives, including Linux operating system, Apache web server software and the relational database system MySQL, in 2013.

Moving to open source technology had several advantages. The tools work “out of the box” and there are large communities of developers with the skills to build open source applications and trouble-shoot problems. And there was no need for any more sophisticated data analysis tools, said Brooks.

“The things we were doing, on one level, it’s quite simple, we are taking data and displaying data. You don’t need a heavy layer of intelligence,” he said.

The company is running its IT systems partly in the cloud and partly on its own hosted servers, and has a back-up datacentre.

Mashery in the mix

Autodata needed specialist software to create and manage its APIs. It considered a number of alternatives, including developing its own software, before choosing a package called Mashery from US software supplier Tibco.

“It came down to two suppliers, and when they came in and gave us a demo, it was quite close run,” said Brooks. “Mashery had the edge – it was more in tune with our thinking.”

One deciding factor was that Mashery came with an application to create documentation and the capability to create a web portal for developers as part of the package. “It takes a lot of the pain away for developers,” said Brooks.

Autodata created a pilot API in 2014 which provided the company’s mechanics with data on service schedules, eventually covering more than 40,000 vehicle models, including how frequently they should be serviced and how the servicing should be carried out.

Since then, the company has made all its data, including details of car parts, car makes and manufacturers in Europe, Australian and the US, in 17 languages, accessible to garages, service centres and automotive parts manufacturers.

Nearly 100 organisations are using the APIs to buy data on subscription, and a further 50 use the company’s older data services.

API returns

The project has played a key role in growing the company’s revenues by 30% each year for the past two years, and has helped it to increase its operating profits by 31%, said Brooks.

The technology has also simplified work for the company’s 50 IT developers, freeing them up to spend their time developing new products and services, rather than IT maintenance.

“The moment we got our own [internal] APIs up and running, the complexity was stripped away and guys were able to get on with what they were supposed to be doing,” said Brooks. “Over 80% of our people are pushing products forward.”

Agile culture

As with many digital IT projects, the biggest challenges were not technical, but changing the culture of the IT department and the company as a whole, and getting used to working in a more agile, flexible way.

Before the work began, Brooks and his team held workshops for everyone in the company, including the board of directors, to explain the principles of agile development.

“Everyone had to understand we deliver something quickly, get feedback, and deliver it quicker,” said Brooks. “That has built up trust in the organisation.”

The company’s latest app provides mechanics with a vast archive of maintenance data on motorcycles, and offers instructions on how to diagnose and fix problems.

The motorcycle API gives developers access to data ranging from recommended lubricants, bolt tightening torques, tyre pressures, repair times and diagnostic codes.

Next stop: big data

The company is in talks with garage chains, insurers and roadside assistance companies about developing data analytics tools that will allow them to predict and prevent vehicle breakdowns before they happen.

“We are starting to look at big data – the data we get back from the garages– and how we can use that,” said Lienard.

The system, which is likely to be built around the Hadoop programming framework, will remind drivers when their next service is due, or recommend the optimum time for owners to trade their car in.

It will be able to access historical service data to predict, for example, when a particular part for a car is likely to fail, based on the type of engine and the mileage.

There is huge potential to develop a new range of data services by extending the internet of things to automotive data, said Lienard.

Owners make an average of 54 transactions in the lifecycle of their vehicle, from the moment they buy the car, to when they buy insurance or have their tyre pressure checked. Each one has the potential to be supported by APIs.

“The possibilities are endless if we start data mining our content and put a little more intelligence on what we have,” said Lienard.