First a confession: I borrowed this blog’s title directly from Gartner Analyst Noha Tohamy and her recent research note Five Things Supply Chain Strategists Need to Know About Machine Learning. I couldn’t do better myself.
Noha’s main message in her note is that demand signals are everywhere—points of sale, distribution channels, the web, social media, etc. It’s hard to take it all in and make use of it—but machine learning can play a big role in helping. Tohamy makes five key points.
1. Machine learning is most suited to analyze volumes of data in a fast response time.
Historical sales are old news, Tohamy says. To best predict demand, companies must account for multiple drivers, and unravel their complex interdependencies. Some, including ToolsGroup, call it demand modeling. “With machine learning, these connections can be discovered by analyzing unique patterns within the data,” she says. “Interest in machine learning has increased as the machine-learning technology becomes more capable of generating faster, more accurate predictions based on identifying patterns in massive amounts of data.”
2. Both machine learning and human judgment are required, not one or the other.
Daniel Kahneman, who shared the 2002 Nobel Prize in Economic Sciences, wrote in his book Thinking, Fast and Slow that “Statistical algorithms greatly outdo humans in noisy environments for two reasons: they are more likely than human judges to detect weakly valid cues and much more likely to maintain a modest level of accuracy by using such cues consistently.”
Still, people have the final say. “Humans are more capable of understanding the upside and risk associated with a decision, making a judgment based on quantitative analytical findings and qualitative business considerations,” Tohamy says. “Finding the right balance of human-based and machine-based approaches will offer the organization the speed and accuracy of science with the art of supply chain domain expertise.”
3. Machine learning is only one of many advanced techniques in your demand analytics toolbox
Machine learning is only part of the total supply chain planning solution. For example Tohamy says, “Machine learning can come up with an accurate prediction of demand for a new product, yet optimization is required to determine the needed levels of safety stock or needed transportation resources to fulfill that forecast demand.”
4. Is now the right time for the organization to embrace machine learning?
“One of the defining characteristics of machine learning is uncovering new interdependencies previously unobvious to the user,” Tohamy says. “Being open to considering these newly found connections and relying on this knowledge for better decision making requires an analytics-friendly culture that is typically present in more analytically mature organizations.”
5. Carefully choose initial machine-learning use cases that promise the highest ROI.
Tohamy says initial successes “can shape the organization's perception of machine learning and impact users' willingness for wider adoption.” Pilots could be demand sensing for promotional events or product introductions, or identifying market drivers that impact forecasts and replenishment plans, reducing lost sales or minimizing inventory obsolescence.
When done right, the business results reach right to the top, says Henning Anthonsen, strategic consultant at Capgemini Consulting. “Depending on its use, machine intelligence has the power to influence a company’s value offerings, key resources, cost structure, customer interaction, strategic partnerships and key decision making processes—considerations that all belong on a C-level.”
Click below to see the entire image above or download the infographic.