One morning in March 2012, GlaxoSmithKline engineer Shaun Glover visited the McLaren Technology Group headquarters in Woking, Surrey. A year earlier, McLaren’s CEO and founder Ron Dennis and Andrew Witty, the CEO of GlaxoSmithKline, had signed a partnership deal. “There was an internal call for proposals, so we put ourselves forward,” says Glover, who is the engineering director of a toothpaste factory in Maidenhead, the biggest of its kind in Europe. It makes products for brands including Sensodyne, Aquafresh and Macleans, 400 million tubes every year. “Some of us connected the dots about how a Formula 1 team would be able to help us. Of course, others questioned what a Formula 1 team could ever know about making toothpaste.”
That day, Glover and his team met Geoff McGrath of McLaren Applied Technologies, a company set up in 2004 by Dennis to apply Formula 1’s high-performance culture and working methods to businesses such as GlaxoSmithKline. McLaren’s headquarters is the 57,000-square-meter Foster + Partners-designed McLaren Technology Centre, which embodies the ethos of one of the oldest and most successful teams in the history of Formula 1 racing. Its glass-fronted main building combines with an artificial lake to form a yin-yang shape. The roof is self-cleaning, using a drainage system to collect rainwater and refill a lake which, in turn, helps regulate the internal temperature. The building is connected to the nearby production center, where McLaren Automotive makes its luxury sports cars. Technicians in lab coats and gloves carefully assemble F1 cars in pristine workshops. All desks are free of clutter, food and drink, according to strict rules imposed by Dennis that also forbid exposed pipes and cables anywhere in the complex.
To introduce his guests to McLaren’s high-performance culture, McGrath played them TV footage from the 2008 Monaco Grand Prix. In heavy rain, team driver Lewis Hamilton hit a barrier on lap six and punctured a tire, forcing him to make a pit stop. Throughout the incident, the TV commentator is animated and loud. Monaco is a notoriously difficult circuit. This was surely game over for Hamilton.
McGrath then played the exact same clip, this time accompanied by the McLaren internal audio feed. Immediately after hitting the wall, Hamilton is given precise instructions by his race engineer: “Lewis, you’re coming into the pit, you make a change to the steering, the launch switch, make sure you’ve done that, you’re going to get new tires and you’re going to get fuel.” Then, in contrast to the TV commentary, there’s silence. Shortly after, McLaren’s chief mechanic says: “Bail out”—fuel the car until the end of the race because there will be no more pit stops—and “Tire set 22,” a pre-determined set of tires. More silence. Hamilton’s pit stop lasts nine seconds. When Hamilton exits the pit lane he is told the identity of the drivers behind and in front of him. Nothing else.
“Before the race starts we know what we will do if the tires degrade more quickly or slowly than we think, if our competitors are slightly faster than we thought, if the safety car comes out,” says Andy Latham, a former race engineer at McLaren who is now chief engineer of analytics at McLaren Applied Technologies. “We will have a pre-determined plan for any possible scenario. We evolve those plans during the race. The last thing I want is to have to make tough decisions during the heat of the race.”
The moment Hamilton hit the barrier, all 13 members of McLaren’s pit crew knew what to do. He went on to win the race—and, by season’s end, the championship.
McGrath and his team had visited the toothpaste factory in Maidenhead in November 2011. After studying the factory’s working methods, they figured out that a production-line bottleneck was occurring during the so-called changeover: transition periods when factory workers needed to switch products on the line from one toothpaste brand to another. This meant changing and cleaning the tubes, rearranging the tools in the line and a number of other procedures that halted production. For McGrath, the similarities with F1 pit stops were obvious.
“If I can change four tires on a car in two seconds,” he questioned, “why does it take me two hours to do a changeover in the toothpaste factory?”
McGrath started discussing production with the factory managers and shop-floor workers. Was the tooling standardized? Were they demonstrating a sense of pride in their work? Was there a belief system in the factory? Was the changeover team specially picked or just whoever is available at the time? Armed with answers, the McLaren team made a computer model of the production line, which allowed them to simulate and visualize the process, much in the way it does for F1 races.
“We didn’t tell the shop-floor workers what to do,” McGrath says. “We put them in McLaren overalls and let them play with the system. They saw for themselves what had to change.” The Maidenhead team developed a seven-step process that began before the changeover, mirroring McLaren’s cycle of simulation, pre-planning, debriefing and continuous improvement. Changeover times fell by 60 percent, dropping from an average of 39 minutes to 15, equating to an extra 20 million tubes by the end of the year. “We used to see changeovers as down time,” Glover says. “McLaren sees pit stops as an opportunity to win the race.”
Eleven years since its launch, McLaren Applied Technologies has become McLaren’s fastest growing and most profitable company: In 2013 McLaren Group group generated 268 million (approximately $415 million) in revenues (of which McGrath says McLaren Applied Technologies contributed “tens of millions”), despite the recent poor performance of its F1 racing team. Among its other projects, McLaren Applied Technologies has designed health-monitoring systems for stroke victims and amyotrophic lateral sclerosis patients based on F1 telemetry; created a scheduling system for Heathrow Airport that reduces flight delays; and worked with some of the world’s biggest oil and gas companies, pharmaceutical conglomerates, data-centre operators and sports brands. McLaren has transitioned into a technology group that happens to have a successful F1 team.
Simulating the Future
McLaren Applied Technologies didn’t make any money in its first five years. In fact, it didn’t make much of anything. When McGrath, a mechanical engineer who previously worked in oil, gas and telecommunications, joined the team in October 2009 to become its vice president, the company consisted of himself, software engineer Steve Rose, and engineer Caroline Hargrove. Hargrove had worked for McLaren’s F1 team since 1997. She played a key role in designing and building F1’s first racing simulators.
Today, McLaren has two simulators at its headquarters: one in the basement, near a 145-meter wind tunnel; the other in the area above, where McLaren’s racing cars are built. They are similar: full-size car chassis mounted on a dynamic rig surrounded by 180-degree curving video screens, with a motion system that reproduces the g-forces generated in an F1 race. A one-way mirror separates them from a control room with five desktop computers and a flatscreen that can display in real time hundreds of parameters extracted from the simulator, such as the angle of the steering wheel, the speed the car wheels are turning, acceleration and engine revs. Team and test drivers spend about 180 days a year inside the simulator—seven times more than in an actual car. They arrive for a session before each race and a debriefing after. The engineers don’t usually give them any data—most sessions are blind tests.
“The simulator represents everything we think we understand about the car,” Hargrove says. “There may be discrepancies between our model and reality—the drivers are our filter. They take in vast amounts of information and can pick out anomalies. We often use the driver for that purpose when we are unsure what’s going on. If the driver says, ‘Yes, that feels exactly like the track,’ you know that your model is right.”
During a typical F1 season, McLaren might change up to 70 percent of a car’s mechanical components. It used to build these parts before testing them on the track, but season testing was banned by the sport’s governing body FIA in 2007 to cut costs. Suddenly, teams with a simulator had a major technological advantage. By then, McLaren already was modeling and testing components virtually, before using the simulator to test the effect each new component had on the driver.
Today, every component is tested, in its virtual form, in the simulator.
“Let’s say you want to test a new anti-roll bar,” says Hargrove. “We could build a prototype model, put it inside your car and test it on the road—or we could build a virtual model and test it on the simulator. We know the size, the specs, how it behaves with physics. The program computes how the new component interacts with every other component in the model. That data is fed to the simulator, where the driver then tests it.”
To McGrath, the data coming off the physical product is worth more than the product itself. He calls it the metaproduct.
In the weeks leading up to a race, McLaren builds a detailed computer model of the circuit and the performance of all the other cars in the race. This computer model allows engineers to simulate any possible scenario and predict its outcome. The team runs millions of simulations, running through all possible permutations and variables, such as timings for pit stops, numbers of pit stops, types of tires and safety cars. McLaren calls this a decision-support system: for every scenario, the computer helps the team to pick a strategy that will result in a positive outcome.
During the race, the McLaren team continues to run simulations from its headquarters, updating the models with live timing data and telemetry from the car, performing tens of thousands of new simulations for every lap. “The racing car is the perfect intelligent product,” McGrath says. “It’s continuously improved upon with extreme time pressure and is custom-built to an individual consumer, the driver. We design it in the simulator and leave the telemetry in the product for remote condition monitoring. That intelligence tells us how the product is being used.” To McGrath, the data coming off the physical product is worth more than the product itself. He calls it the metaproduct.
“When we started we were basically three people sitting at a table, wondering how to build this business,” McGrath says. They were trying to answer the question: What would McLaren do if it didn’t make F1 cars? McGrath had a plan, at least in theory. It was a vision that intersected with Hargrove’s simulators and computer models, the telemetry in the race cars and the data design of the prototypes. This is what they could offer to businesses that were traditionally managed retroactively by studying quarterly financial reports—data from the past. McLaren would get them to work with live data, to compete in real time and to simulate the future.
Making Cycling and Medicine More Like F1
Scott Drawer was head of research and innovation at UK Sport at the time of the GlaxoSmithKline project. He enquired whether McLaren and Team GB could collaborate in preparation for the 2012 Olympics in London. Drawer was an F1 fan and he wanted to bring some of its practices, such as telemetry and predictive algorithms, to other elite sports. The two organizations began to collaborate on disciplines that involved interaction between athlete and machine: cycling, sailing, rowing and canoeing.
“Track cycling was particularly successful,” Hargrove says. “Of all the Olympic sports, it’s probably the most akin to F1 except that, instead of an engine, they have a person.” Drawer wanted to understand the correlations between power, cadence and heart rate and how those parameters translated to results on the track, but the cycling team didn’t have enough data points for the sprinters. So McLaren built the Datarider, a small aerodynamic box which nestles under the rider’s saddle and connects to sensors in the bike to collect data related to power, torque and bike angle. The device itself contained accelerometers, gyroscopes and Bluetooth transmitters. Previous sensors used by the cycling team could transmit information at a frequency of approximately 20Hz. Datarider had 200Hz. (The sensors used in F1 run at 1,000Hz.)
“I wanted to make sure everything was calibrated perfectly so we did all the testing on our site,” Hargrove says. “I saw the data from [track cyclist] Chris Hoy, panicked and rang them to apologize. It seemed obvious that I had the calibration wrong as the numbers were so high. They said it was fine—those were the numbers Hoy produced.”
By 2010, McLaren was seeking a partnership to design a bike. It contacted the third-largest bicycle brand in the world, California-based Specialized Bicycle Components. “We wanted to apply data-driven design to make a bike,” says Duncan Bradley, head of high-performance design at McLaren Applied Technologies. “Like any other bike maker, Specialized would design by eye. Test riders would then ride the bike and give subjective feedback. That’s exactly how we would design F1 cars 50 years ago.”
Specialized’s brief to McLaren was to make a lighter bicycle, under the assumption that it would then be able to travel faster. McLaren started by questioning that assumption. “On a bike, you’ve got a stiff structure the frame with a heavy bag of water, the rider, sitting on it,” Bradley says. “We had to consider the human aspect of the movement. How do you maximise the equipment performance with a human in the loop? In this case, it was more: how do you maximise human performance with equipment in the loop?”
McLaren started by studying the frame, attaching more than 20 sensors that could measure the various forces and vibrations that affect a moving bike and the way people ride it. They adapted the chassis rig of Hargrove’s driving simulator to fit on a bike. Once they understood how it moved, they applied the same forces to a test rider in this case Bradley, a keen cyclist. One element at a time — the frame, the tyres, the human — they developed a formula that crystallized their understanding of how people ride bikes. “With that computer model, we could specify whatever parameters we wanted — shape, weight, stiffness — and design quickly,” Bradley says. “We completely flipped the way bikes were designed.”
It took eight months to design the bicycle, the S-Works + McLaren Venge, which came out in 2011. It was 20 per cent lighter than Specialized’s previous top model, the S-Works, but had the same structural stiffness. The same year, Mark Cavendish won the UCI Road World Championships — the first British rider do so since 1965 — riding a Venge. (The two companies have since worked together on another bike, the S-Works McLaren Tarmac.)
“Specialized later told me that they had learned more in six months about bike design than they had in the previous ten years,” McGrath says. “They also told me that we didn’t charge enough.”
The GlaxoSmithKline project enabled McLaren Applied Technologies to enter the healthcare market. Two years later, McLaren provided monitoring sensors for a GlaxoSmithKline clinical trial with 100 stroke patients in the UK and US. McLaren had previously tested the sensors on rugby players, measuring data during training and developing algorithms that could predict when a player would peak in training and when they’d run the risk of injury.
“Most wearables can only tell you how many steps you’ve taken, not how many hours you spend sitting or lying down,” McGrath says. “They lack contextual awareness, which is what you need for clinical trials. I would even question that the wrist is the best place to get any kind of insight—the upper body is far more useful.”
That sort of insight was exactly what GlaxoSmithKline needed. One of the ways to measure a stroke patient’s health and response to medication is to assess their mobility. “Once every month or so we would ask a patient to walk between two chairs ten meters apart, and see how many steps they take and how long it takes them,” Julian Jenkins, GlaxoSmithKline’s vice-president of project planning and management, says. “I don’t know how long it takes for me to walk ten meters, let alone a stroke patient. It was flawed.”
McLaren placed a device the size of a ten-pence piece on the patients’ necks to measure 20 parameters such as gait, cadence and stride frequency. “When I saw the first patient’s data I was astonished,” Jenkins says. “Within seconds, I could tell that the patient was very sick. There’s no test that we could have done before that would reach the same conclusion. I realized then that this was going to change clinical trials.”
GlaxoSmithKline and McLaren are now conducting clinical trials with amyotrophic lateral sclerosis patients and, in March 2015, McLaren announced a partnership with the University of Oxford, with the aim of using analytics to improve patient care and to build a simulator for surgeons.
“Humans are hard to model,” McGrath says. “But we always thought that if we can measure the health and condition of an engine, why can’t we measure the health and condition of a person?”