Utilizing End-User Input In Product Development
According to a study by PWC, companies that apply customer engagement strategies that involve user-driven requirements are twice as likely to expect growth of 15 percent or even more over the next five years. One-third of companies in the study considered customers to be their most important innovation partner. Collecting user input, therefore, is vital. Yet a study by McKinsey & Company found that more than 40 percent of companies aren’t communicating with their end users during the development process. More than half of the companies in the survey said that they have no objective way to assess the output of their design teams or to set targets for them.
The user data generated by-products is a powerful and as yet underutilized solution. Many products are already produced with embedded usage analytics tools. The user data generated by-products can help vendors and manufacturers shorten product development lifecycles and quickly improve designs. The integration potential of IoT products can make this process even faster. Combine these tools with machine learning and artificial intelligence, and there’s the potential to reduce the need for physical product testing even further while creating a better product.
Some of the most striking results in this area have come from the automotive industry. In automobile manufacture, machine learning has allowed some manufacturers to reduce physical product testing needs by as much as 90 percent, according to the UK Advanced Propulsion Centre. Dr. Richard Ahlfield, founder of Monolith AI, and his team developed a software platform for training machine learning algorithms and analyzing their results. The technology is applicable to numerous industries. As regards the automotive industry, complex data can be used to predict a new vehicle’s behavior on the track before that vehicle is even built.
General Motors has used machine learning to transform product prototyping. In a test of its “Dreamcatcher” system, the prototyping of a seatbelt bracket part resulted in a design that was 20 percent stronger than the original, 40 percent lighter and consisted of one piece rather than eight. Continental, one of the world’s largest automotive parts suppliers, has created an AI-based virtual simulation program that can generate 5,000 miles of vehicle test data per hour. Physical testing would require more than 20 days to generate the same amount of data.
In addition to time, a great deal of money could potentially be saved. According to the Congressional Research Service, R&D expenditures rose from $676 billion to $2 trillion between 2000 and 2018. Yet the McKinsey study revealed that more than half of new product launches fail to hit their internally-generated targets. Increased virtual testing can reduce costs while generating the same amount of data without the expense of physical testing. In fact, according to a different study by McKinsey, more than 80 percent of IT professionals in marketing and sales in the study believed that AI data insights helped to reduce overall costs by as much as 20 percent. And user data can be an important part of this.
Curion is one company currently leveraging these technologies to provide consumer data to companies. The company provides its services to 65 percent of Global 100 companies, and since its inception, has tested more than 7,000 products with more than half a million consumers. Curion has also collected product data from 105,000 consumers in some of the largest metropolitan areas of the United States. By making the sensory data that it has collected from users available to companies – for example, which type of raisins consumers reacted best to – Curion hopes to help brands to create better products and to get them out faster.
The integration potential of IoT devices means that companies can collect more diverse sorts of data more rapidly. But that’s a lot of data – sensors, location data, automation data, and so forth – and information not only about one specific device but also about its interaction with other IoT devices. This creates a new problem: how to sift through the flood of data to pinpoint the essential information. This is a problem tailor-made for AI.
One of the unquestionable strengths of artificial intelligence systems is the ability to process massive amounts of data fast. AI systems also have the potential to make connections between data that a human analyst – or even a team of analysts – may miss. Unlike human analysts, properly trained AI can make connections between different data without the implicit biases that some human analysts might have.
Automation of this process – often called “automation analytics” or “smart data discovery” – can automatically make connections between different data, transforming the process of data analysis itself. In addition, predictive analytics and machine learning can identify trends and anomalies in data. Natural language generation (NLG) tools can provide automated summaries of these analyses in easy-to-understand natural language.
And all of these insights have the potential to make a huge difference at every stage of a product’s development.
Customer feedback is a vital part of product development. Feedback collected by embedded usage analytics tools is objective and reliable, but there’s a lot of it. A combination of artificial intelligence technologies, including automation analytics, natural language generation, predictive analytics, and machine learning can help companies to make sense of all of this data and leverage it to create better, safer, more reliable products, and to get them to market faster. In terms of R&D, it can save companies time and money, and can make companies more agile with regard to responding to problems and preventing them before they begin.