By Gwen Moran
Employees of Independent Security Evaluators, a Baltimore-based digital security firm, spend roughly 80% of their time working on the tasks listed in their job descriptions and 20% of their time pondering big ideas. Those ideas may include developing a pet project, finding flaws or bugs in systems, or taking on some other thorny problem. But founder Steve Bono wants them thinking about big-picture issues that can ultimately help them be better at their jobs and, whenever possible, benefit the company.
Bono says that, since the company’s founding in 2005, employees have always tinkered or researched in work-adjacent ways beyond their regular projects. And it’s not in any way unusual for coders and other tech-types to go on bug hunting expeditions or spend time probing IT system architecture after the 9 to 5 work day is done.
“[I]n many of those cases, we were proud of the work they produced and felt we should be officially rewarding them for it,” he says. Eventually, the company’s leaders reasoned that they should make this independent research and development (R&D) part of how they do business.
Harvard University School of Business Professor Stefan Thomke advocates for this kind of controlled experimentation in his new book, Experimentation Works: The Surprising Power of Business Experiments. Thomke argues that even experienced managers can’t rely on intuition to improve products, processes, and business models. Instead, they should turn themselves into “experimentation organizations,” where employees are free to pose and test hypotheses to develop new and better ways of doing things within a rigorous framework that challenges findings.
Some organizations have embraced the approach wholeheartedly, conducting more than 10,000 experiments per year, he says. As a consumer, “if you’re using Amazon, Netflix, Microsoft or products such as Bing, Booking.com, you are part of an experimentation ecosystem as a user,” he says.
Creating an effective framework
Becoming an effective experimentation organization requires a solid framework within which it tests its ideas. Thomke recommends that companies put the scientific method to work, that, of course, is the empirical approach to conducting experiments developed back in 1620 by philosopher and scientist Sir Francis Bacon. Thanks to the online and offline tools available to modern-day companies, conducting experiments and testing hypothesis is more possible than ever, he says. Sir Bacon would be pleased.
Experiments must have a hypothesis that can be tested, Thomke says, and the team has to figure out what it wants to learn and whether it can be measured. Experiments should have controls to test for different influences, too. In his book, Thomke points to an experiment Kohl’s put to work. The retailer was looking to cut operating costs in 2013. One idea was to open an hour later in the morning but some executives feared it would have a negative impact on sale. But the company conducted a controlled experiment and found that there was no meaningful decline in sales from opening stores an hour later.
Thomke adds that companies that go the experimental route must be open to some failure. Not all experiments will work out. In some cases, especially in highly regulated environments, organizations may need guidelines and approval processes for experiments. These should be determined and finalized to streamline the experimentation process, he says. And testing with scientific rigor is typically going to give organizations the best results and best insights.
Making room for experimentation has done well for San Diego-based SmartDrive, a driver safety and transportation intelligence company, that frequently tests new features to meet customer demands. SmartDrive’s platform needs to work with many vehicles and offer distinct advantages over the competition, says Ray Ghanbari SmartDrive’ chief technology officer. Continuous trial-and-error led the company to combine vehicle data and video, including integration to vehicle engine control units, to deploy a variety of safety features.
Because experimentation is so ingrained in company culture, SmartDrive has been able to expand its portfolio to 60 patents. “You’re really always asking, ‘What is the problem that we’re solving for?’ Not, ‘What is the specific problem?’” Ghanbari says.
The ROI on experimentation
At Inspection Security Evaluators, Bono says that the developments cost more than they generate on paper, but they’re well worth it. The company has created a major sales event, IoT Village, that it rolls out at industry events; developed two additional patents and two trainings that they’ve sold; and team members’ vulnerability research has added significant contributions to the field and raised the firm’s profile.
But, says Bono, those payoffs are in no way the most important benefit. Employees, he says, “get a lot better at what they do. They are very happy with the opportunities they have here,” he says. That all adds up to a more cohesive team and a more loyal workforce.
Thomke says some organizations get to the point where they’re testing thousands of hypotheses in a year to get to innovation and deeper insights. And several of the companies he cites in his book, including Booking.com and Intuit, as well as other companies that test with scientific rigor, are finding important changes and breakthroughs as a result.
“Maybe there a few people are still thinking about what the ROI is on this approach,” he says. “I think it’s just going to be a question of how can we do this? How do we build this capability quickly? Because if you don’t learn how to do this in the next five to 10 years, I think you’re going to be left in the dust.”
More must-read stories from Fortune:
—Android 10’s 7 most anticipated new features
—This new app puts deepfake technology in the hands of a mainstream audience
—Google hit with a record fine by the FTC for violating children’s privacy on YouTube
—A U.K court may have made police use of facial recognition easier
—Porsche unveils its first-ever electric car
Catch up with Data Sheet, Fortune’s daily digest on the business of tech.
Read more here:: fortune.com/tech/feed/
An updated release of Baidu’s deep learning framework includes a batch of new features ranging from inference capabilities for Internet of Things (IoT) applications to a natural language processing (NLP) framework for Mandarin.
The latest version of PaddlePaddle released this week includes a streamlined toolkit dubbed Paddle Lite 2.0 aimed at inference for IoT, embedded and mobile devices. It works with PaddlePaddle as well as pre-trained models from other sources, Chinese Internet giant (NASDAQ: BIDU) said.
Along with faster deployment of ResNet-50, used for image classification on convolutional neural networks, Paddle Lite 2.0 also supports edge-based FPGAs and other hardware.
New development kits include ERNIE 2.0, and updated version of Baidu’s natural language processing framework. The pre-training framework is said to outperform Google’s BERT technique for NLP training on English and several Chinese language tasks.
As it seeks to attract more machine learning novices, Baidu also said it is adding an AI platform called Easy DL. The tool enables training and building custom models via a drag-and-drop interface. Billed as a “one-stop AI development platform for algorithm engineers to deploy AI models with fewer lines of code,” Easy DL has so far been used by manufacturing, agriculture and service industry users to build more than 169,000 models, Baidu claimed.
Other PaddlePaddle upgrades include expanded support for operators along with easier-to-use APIs. Also included is a “PaddleSlim” module for compressing models.
“By providing hardware support, cloud-to-edge deployment, development kits and Master mode, we’ve significantly improved PaddlePaddle’s performance and feature-set,” said Tian Wu, executive director of Baidu’s AI Group. “PaddlePaddle will keep advancing large-scale distributed computing and heterogeneous computing.
Currently used by more than 1.5 million developers, PaddlePaddle was released to the open-source community in 2016 to help expand AI application development. Baidu Research notes that its platform was used by Intel in developing its Nervana NNP-T processor. Codenamed “Spring Crest,” the Intel neural network processor is designed to train deep learning models at scale and to accommodate a given power budget.
Read more here:: www.datanami.com/feed/
AT&T and Nokia said they would build an ‘innovation studio’ in Germany to focus on the IoT.
By Paul Rainford Also in today’s EMEA regional roundup: Telefonica COO supports consolidation; Nokia and AT&T collaborate on IoT in Germany; smart speakers in smart homes.