5 Steps To Strategy Tuning Through Machine Learning

Conventional thinking in business has long been that strategy decisions are made by humans, while the focus of automation and machine learning should be on execution. With the speed of change and volume of market feedback today, as well as the advances in machine learning, Amazon, Alibaba, and others have proven the value of software driven strategy decisions.

For example, most e-commerce platforms today offer millions of products, with a changing mix daily and a changing market, such that it’s virtually impossible to manually predict a strategy for mapping customer demographics to products displayed online. Only smart software can plow through the volume of live data, recognizing trends, customers, and match offerings to reality.

Alibaba, today the counterpart in China to Amazon, Ebay, and Google here, has demonstrated leadership in this area, and provides guidance for all of us to learn from in a classic book, “Smart Business,” by Ming Zeng. Ming is the former chief of staff and strategy advisor to co-founder Jack Ma for over a decade, and outlines five key steps for automating decisions today as follows:

1. Log and use consumer behavior and product transaction data. In my work with small businesses and startups, I routinely find owners who rely on guessing at key customer drivers, and let their passion drive product focus, rather than data. They think they are saving costs by not using the latest technology to capture data, and minimizing storage. In China, even tiny bike sharing companies now have to digitally track every bike through GPS, and every mobile interaction between a bike and a rider to compete. No human or paper tracking systems are a competitive alternative. The days of manual reservations and receipts are behind us, whether it be with rides, clothing, meals, or entertainment.
2. Configure every decision step into real-time software. Businesses must capture every business decision activity, including customer relations, in digital software so that decisions affecting the activities can be automated and optimized through machine learning. This is a new class of software that can adapt in real time to market changes. In the bike rental business, all operations and rental decisions are made completely by software, with no human intervention. The efficiency gain is tremendous. The software directs humans and trucks to balance the tide of idle bicycles to other areas of a city where the demand is higher at the moment, rather than humans directing software.
3. Get data flowing, and machines talking to each other. With today’s technology, data flow and coordination are readily achieved through common Internet protocols and application programming interfaces (APIs). These allow applications to communicate automatically and almost instantly, even over long distances, to mobile and IoT devices. Way back in 2002, Jeff Bezos at Amazon issued an ultimatum to completely institute internal APIs within the company, and later to their millions of suppliers. Through this focus, Amazon has become one of the first trillion dollar companies, and continues to expand its reach beyond books, e-commerce, and now into groceries with Whole Foods.
4. Record live data in full for all internal business elements. The opposite of live data is static data that is sampled or profiled for analysis at a later date. Live data also requires metrics and infrastructure that can interpret and evaluate the data, and smart businesses must develop these in the algorithms they use in their data-intelligence engines.
5. Apply machine-learning for real-time software decisions. Intelligent real-time software decisions are quickly replacing after-the-fact analytics. Only with full live data, built-in metrics, and artificial intelligence to iteratively improve the decision process, can your business keep up with the pace of change, and unique markets around the world. Uber’s algorithms match car and driver, minimizing wait times and making mapping calculations in ways that no human dispatcher could match. Google search rankings are updated many times a second, based on new info and your profile changes. If your business is not powered by an algorithm, you don’t have a competitive business today.

These steps lead to what Zeng defines as a customer-to-business (C2B) model, where every direct interaction with customers sets into motion a striking reorientation of all business activities, and builds a feedback loop from customers. This allows businesses to automate all decisions productively, to scale and compete effectively, in tune with trends and new marketplaces.