AkzoNobel: Bring on the color revolution

AkzoNobel has been honing the art of color matching for two centuries. Scientists have been working painstakingly in the lab to adjust, recalibrate, and tweak a color until it’s just right for the latest cars, buildings, and trends in interior decor. It’s a complex task—and the scientists at the Dutch paint company are among the best.

Proud of its 200 years of heritage, AkzoNobel always has one eye on the future. But with innovation comes change. These world-leading color specialists had to consider that the way they had been working for decades—using mathematical models—was no longer the most efficient or innovative. In fact, there was more than a little skepticism toward color prediction models using artificial intelligence (AI).

“We always thought that our physical models were the best that we could have,” says Rob Reijers, Manager of Global Color Apps Development at AkzoNobel. “So why would we change?”

The need to adapt

Pressures on the paint industry have been immense. New colors emerge every day. Manufacturers in the car and other industries constantly dream up new finishes to give their models an edge on the competition.

For a company like AkzoNobel, the key to success is innovation. “For us, innovation is about going beyond: finding new ways to exceed the expectations of our customers and go beyond the scope of their imaginations,” says Pim Koeckhoven, Director of Color Technology at AkzoNobel.

Modern screens have now reached such high levels of quality that consumers see colors on a daily basis—colors that were unimaginable even five years ago.

Color prediction is not a simple science, or a quick one. AkzoNobel’s complex process of deciphering the multiple physical elements that influence color has been honed for many decades in partnership with institutions and universities—organizations that have established ways of working, steeped in years of scientific testing.

This pace of change is driving innovation throughout the company. “When I talk about how we worked in the ‘old days’, I’m talking about last year,” jokes Koeckhoven.

Chief among the changes is the introduction of Azure Machine Learning. In the past 12 months, the core process of color prediction—that is, calculation—has been extended by new technology.

“Up until last year, all our color prediction calculations were based on physical models, just as they have been for the past 40 years,” explains Koeckhoven. “But with the machine learning technology we’ve implemented in the last year, we can now make calculations based on deep learning models. And this is what we call going beyond imagination—who would have thought this could be possible a few decades ago?”

The power of machine learning

Color prediction is no small operation. AkzoNobel has more than 1,000 employees working across various aspects of the color domain. Recently, for the team working on new colors for cars, the new technology has led to significant time savings. “Approximately 15-20 percent of their time,” says Koeckhoven. This time savings translates into both lower costs and a faster time to market, making the whole operation much more cost-effective.

The main impact is seen in the lab, where teams are now able to create more color recipes, more accurately, in less time.

“In the beginning, we were quite skeptical about machine learning technology and what it could really do for us,” Reijers says. 

“But we were really surprised by the incredible accuracy of the results we could achieve with machine learning technology, and also by how quickly we could get these results,” he adds. “The whole experience completely changed my viewpoint on using machine learning.”

The demands of the car industry

The efficiencies generated by machine learning don’t just save time internally, they also have major benefits for how AkzoNobel serves its customers. The car industry is a powerful example.

“Every year, new cars with new colors are introduced. What we do at AkzoNobel is provide the paint to repair these cars when drivers have an accident. The quicker we can provide the correct recipe of a color that is just introduced in the market, the easier and quicker the car can be repaired,” explains Reijers.

“In the past, it sometimes took us up to two years to get a car color available for the market,” he continues. “Now, with all the new technologies we have in place, we’ve reduced that to just one month.”

Smarter calculations, fewer tests

The introduction of the new technology could have meant internal upheaval—a need to retrain all the lab technicians and scientists. However, according to Eric van Winden, Color Researcher at AkzoNobel, the transition was almost seamless.

“The people conducting the trials needed for color prediction hardly noticed the new technology at all,” says van Winden. “They’re still using exactly the same process and software tools. The only difference is that the answers they get are calculated in a much smarter way. This means they require far fewer rounds of tests to get even more accurate results.”

“Deploying the machine learning models in Azure was very straightforward,” notes Reijers. Once they were trained together with partner, Machine2Learn, the models were easy for developers to engage with.

“They were familiar with the models’ open source foundation, which runs on a Linux stack and utilizes Azure container technology.”

In addition, the platform as a service (PaaS) capabilities of Azure increase operational simplicity so AkzoNobel can scale up or down as needed, developing and operating with ease since all the plumbing is taken care of by the platform. It’s fast and efficient.

In fact, once AkzoNobel had finalized its machine learning model, the time it took to deploy the company’s first end-to-end environment on Azure was less than two months.

“With Azure, we’re able to be much more creative than we were before because new technologies can be picked up without a lot of effort. That's what you notice most in terms of a change in the way you work,” says van Winden.

The start of a color revolution 

Great progress has been made in the 350 years since AkzoNobel’s humble beginnings. The small foundry that would eventually become this innovative company—now active in more than160 countries and employing approximately 34,500 people—was founded just four years after Dutch master Rembrandt van Rijn painted The Night Watch.

However, according to Koeckhoven, this is only the beginning. “We are at the start of a color revolution,” he says.

Advances in image recognition, the digitalization of paint applications, and the development of sensor technology all have major implications for the industry. Meanwhile, paint itself continues to develop in incredible new ways, requiring new digital processes. At the same time, color accuracy on screens continues to develop—impacting not just color companies, but retailers, manufacturers, and many other industries.

“When I look at the technology we’ve recently started using, I think we are still very much at the beginning of what we might be able to achieve as a company,” Koeckhoven says. “The world will look very different in a couple of years’ time.”