Data-driven: How Mercedes-AMG Petronas F1 uses tech
Statistical Data Analysis

Data-driven: How Mercedes-AMG Petronas F1 uses tech

At technology conferences around the world there’s often plenty of talk of ‘velocity’ and ‘the need for speed’ but in the case of Formula 1’s Mercedes-AMG Petronas Motorsport team, going faster really is of the essence. The circus that is F1 today sees cars travelling at up to more than 200 miles per hour on twisting, turning circuits that place enormous demand on precision engineering and driver concentration and control.

It’s a sport that attracts many techies, understandably, and ICT plays an important part in how teams fare around the world. This is a data-driven world where being able to identify causes of failure or sub-optimal performance is hugely important to the performance and safety of these amazing aerodynamic machines, honed through perpetual design changes and constant testing. Less cosmetically exciting, it’s also ICT that underpins the complex logistics of this sporting moveable feast.

Of course, F1 teams are early adopters of technology and Mercedes-AMG Petronas can generate 45 terabytes of data per week in race driving, simulations and testing; CFD alone can account for 30TB of that. A car that’s far fully loaded for testing might have 200 to 300 sensor units, some with multiple sensors in one unit, generating millions of data points per race weekend. One of the team’s SQL databases might have a couple of hundred million rows in a season.

I spoke recently to Matt Harris, head of IT at the team, about some of the work the company is doing, often alongside tech partners including Tibco and Pure Storage.


Been there, done that

“As a team we’ve traditionally been very good at data analytics on the track,” Harris says. “People go on about the Internet of Things and Big Data but we’ve had that on a car for 25 years.”

What is being instrumented and monitored? Pretty well everything from car performance timing to GPS location, weather effects, wind tunnel testing, to smaller components and dynamometers that watch over the power output of an engine, and these are all sources of data.

“We’re always trying to improve our understanding of data,” Harris says. “A driver will give you verbal feedback such as ‘I felt a bump’ but you have to come back to the data to see what happened and we want, and need, to be data-driven. Every decision we can make on data rather than gut feel should improve our performance. The business is beginning to get more data-driven, not just the car.”

In modern sporting parlance much of the car-side endeavour is focused on “incremental gains”: those slight improvements and tweaks that can accrete to make a significant overall advantage. The team lost a grid position by 0.001 of a second, the equivalent of 6cm, so any tiny improvement makes a difference.

Mercedes Canada team shot

Much of this work is taking place where from the pit lane from circuit driving simulators that are “like an Xbox or PlayStation game on steroids, with proper motion and photorealistic imagery” to tracking the information gleaned from those ubiquitous sensors. But there are trade-offs.

“More sensors is more weight; more weight is slower,” Harris says. “On Friday the car is heavier because we run more sensors.” But on race day getting the weight down is as important as a racehorse jockey making his weight for the off.


Exception handling

But more data is not necessarily always helpful and in a business where things literally move very fast – not just cars but schedules and development – it’s critical to be able to distinguish signal from noise and not get bogged down in a data deluge.

“All the time we’re trying to understand what’s happened and what’s changed but we’re time- and resource-constrained,” Harris says. “We want to not have to look at so much data but you have to know how not to miss something important. We’re constrained by having [a maximum of] 60 people at the track and those people can only look at a certain amount of data. If you have systems that understand ‘normal’ then you can make decisions based on exceptions [that’s perfect]. We want to get rid of the ‘noise’. Otherwise it’s like looking for a needle in a stack of needles, not a needle in a haystack.”

Artificial intelligence might be here to help in that quest to only see what is key and even to providing solutions and data science is critical. Harris’s team has the box-office appeal and good fortune to attract tech sponsors and partners. Tibco, for example, supplies analytics tools and expertise that help Harris and co. figure out the knotty problems of engineering, testing and validation.

“A vital part of the Tibco relationship is rather than employ another 20 data scientists you have a whole workforce of data scientists,” he says. “They’re a virtual part of the team for us.”

This is ongoing, iterative work and for once the cliché of saying an IT team is driving at 100 miles per hour making significant changes as they go is not far from the truth.

“It’s a prototype car every time and it’s always being re-engineered whether that’s for altitude, circuit conditions et cetera,” Harris says.


Changing gear

One current effort is to understand more about gearbox wear so Mercedes-AMG Petronas can better project the life of these key subsystems:

“Going around Monaco, a driver changes gear about 100 times per lap and in a race you’re into hundreds of thousands or millions of data points: what happens under traction, in the wet, where there’s a bump…”

Developing a model based on historic knowledge of gearbox wear metrics and establishing what cause wear and physical damage will help the team plan and adjust. But almost everything in F1 has a data element from “when would you do a pit stop, when would you do it if there’s a safety car or virtual safety car, what are other teams likely to do” and beyond.

However, Harris says for now a better question might be not how much lap time does analytics save drivers and their cars per lap, but reliability and how it can help avoid the dreaded Did Not Finish conclusion and no points in the driver and constructor championships.

“A game-changer in F1 could be one-tenth of a second but how soon [could IT be the difference between winning and losing], who knows. Being able to spot and stop a DNF is probably more important. [Tech can help make cars faster] but avoiding a DNF is number-one now.”


Driving forces

What about drivers: do they get intimately involved or do they just want to be left alone?

“Probably the most analytical person I worked with was Nico [Rosberg]. He didn’t just want to see it but to understand it, but also he wants to see the relationships between the decisions and the data. Lewis [Hamilton] is interested but he’s very much in the now. [Valterri] Bottas I haven’t been on the track with, but Nico was an engineer who turned into a driver.”

Processes are also critical of course and Harris says the team works on a democratic basis. “We’re removing layers of superiority and [encourage] being able to speak out on a ‘see it, say it, fix it’ basis. It doesn’t matter about rank.”

As for competitive information, the 2007 fines dealt out for the Spygate controversy mean teams are very keen not to receive competitive information, even when it has been accidentally leaked, for example though a misaddressed email. And anyway, Harris has plenty of his own data to worry about before the next chequered flag is lowered.


Also read:
How IBM gave Red Bull F1 new wings
1000mph land speed record quest is driven by data
Inside Lotus’s F1 factory
Inside Mercedes' F1 factory: Simulator, data and sensors
Data drives Grand Prix success at Lotus


«The CMO Files: Chris Boorman, Automic


C-suite talk fav tech: Matthew J. Olivieri, AdSemble, Inc.»
Martin Veitch

Martin Veitch is Editorial Consultant for IDG Connect

  • twt
  • twt

Our Case Studies

IDG Connect delivers full creative solutions to meet all your demand generatlon needs. These cover the full scope of options, from customized content and lead delivery through to fully integrated campaigns.


Our Marketing Research

Our in-house analyst and editorial team create a range of insights for the global marketing community. These look at IT buying preferences, the latest soclal media trends and other zeitgeist topics.



Should the government regulate Artificial Intelligence?