Amazon’s step into building its Amazon Basics brand. These products are designed around the attributes that customers find most important, and these attributes can be inferred from reviews, purchase behavior, and even usage data on some of their devices. Examples of what you see when buying AmazonBasics products are better packaging for electronics, simplicity of design, and easy-to-digest product descriptions. The results show this process of listening to customers and quickly iterating on products works. Amazon’s private label products will earn hundreds of millions, if not billions, in revenue this year.
In addition, Amazon has begun opening physical stores to take its learnings from the digital world to the world of physical retail. For example, it can use data of what books are most popular with readers in the Seattle area to influence what it displays in its bookstore in the University Village shopping center in Seattle.
By opening up physical bookstores and Amazon Go cashier-less stores, Amazon can now close the loop between their digital shoppers and physical shoppers to create custom retail locations that leverage both sets of data. Like we have seen with companies like Indochino and Warby Parker, combining digital and physical customer experiences is a powerful combination and allows companies to have more touch points with their customers. (Disclosure: Our firm is an investor in Indochino.)
This virtuous cycle of data – driven by consumer trust – is driving the growth behind many of the largest machine learning projects in the world today. Whether it is Google search results, Facebook news feed, or Xbox game recommendations, consumers are willing to share more and more data in exchange for better products.
Nevertheless, as more applications are powered by data and machine learning algorithms, the costs associated with AI making “bad” decisions and the costs associated with losing customers’ trust become higher.
For example, one frequently cited statistic in the world of autonomous vehicles is that nearly 95 percent of automobile accidents today are caused by human error. This is a huge number, and autonomous vehicles have the potential to eliminate many instances of human automobile accidents. However, if an autonomous vehicle gets into an accident, there is a risk that other autonomous vehicles using the same software will get into the same accident. Rather than individual humans making individual mistakes, an AI system has the risk of amplifying the same error over many different encounters of the same situation. This can make people distrustful of a system, and it creates a cycle where it becomes harder for a company to collect more data and improve its products.
We have seen several high-profile breaches of trust tied to AI systems this year, from autonomous vehicle accidents to Russian shell companies influencing our political views to malfunctioning robots that laugh at us and even robots that trick people into believing they are humans.
When consumers feel there has been a breach of trust in an AI system, the reactions can be quick and severe. In the case of the Cambridge Analytica scandal, customers immediately deleted their accounts, celebrities made public announcements about deleting Instagram, and Facebook stock dropped more than 10 percent over two days. Since then, Facebook has taken several steps to regain trust, such as tightening apps’ access to data.
In good times, the virtuous cycle of data builds on itself. When things are going well, people share more data, companies build better products, and people continue sharing even more data. However, when things start going poorly, the negatives pile on quickly as well. If people are less willing to share data, companies find it harder to prioritize product development efforts, they lose customers, and they find it increasingly difficult to build good products.
Over the next few years, we will see more traditional consumer product companies using their data to build better products because all companies are becoming technology companies. For example, Alaska Airlines will be able to better predict what flight times and routes I prefer, Nordstrom will know what sales I am most interested in, and Starbucks will use customer loyalty data to provide better recommendations on what drinks I should try.
In the midst of this transformation, the increased regulatory and consumer scrutiny on what companies are doing with consumer data will lead those companies to take firmer stances on adopting a more systematic approach to transparency around data collection, data warehousing, and adherence to data regulations. There was a time in the recent past where the data business felt like it was owned by Facebook, Google, and Amazon, but growing consumer and corporate awareness about the value of data is creating an important shift that enables many other companies to develop their own virtuous data cycles.