In introductory college microeconomics classes students are exposed to the concept of price elasticity – that is, the predicted response of a buyer in terms of quantity demanded when a seller raises or lowers the price of a certain good or service. In the real world, companies in competitive industries are continually trying to extend insights about elasticity and other behavior-response metrics across thousands of customers and products. The problem with this is that real-world business problems bear very little resemblance to the theoretical examples contained in Microeconomics 101 textbooks. First of all, customer behavior is very hard to pin down. Numerous variables affect every translation. How can we say with a high degree of confidence that a price change was the main cause of a change in demand? Why not something else – perhaps an especially strong and persuasive effort by the salesperson to make the sale? Or something completely outside our control, like the weather that day?
Modern business intelligence systems are rising to meet this challenge by encompassing more explanatory variables into their algorithms. But even so there is still a problem. These models are still confined to looking backwards, to past events, to formulate guidance about what to do in the present and future. Sales & marketing decision makers need to complement their insights from historical data with an approach that can work in the constantly changing environment of their markets in real time. That approach has to draw on the processing capabilities of a system uniquely suited to the ambiguities and constant flux of dynamic markets: the human brain.
Analysis on the run
Humans and machines don’t “think” in the same way. Consider the following scenario. A sales representative for a seller of food products has a particular customer with whom she has a track record of selling condiments like ketchup, mustard and soy sauce. The customer is a fairly frequent buyer and so there is an extensive transaction history going back over a number of years. The sales rep’s company uses a quantitative analysis and optimization tool to predict likely demand, providing its sales reps with guidance for pricing and other marketing decisions.
The sales rep is doing some prep work ahead of a meeting later this week with the customer, including reviewing the price recommendations the optimization system has provided. As she glances through the morning paper, the sales rep notices that one of her company’s major competitors has launched an intensive three-week promotional campaign for customers in the local area. Immediately she knows that her customer also buys from the competitor, and no doubt will be fully aware of the competitor’s promotional campaign. She leans back in her chair, looks again at the price recommendations in the display screen of her PDA, and takes a deep breath. What should she do?
Gut instinct versus science
The traditional answer would be: go with your gut instinct. After all, this is what good salespeople do – call up their years of experience to figure out how to play nuanced competitive angles and make the sale. It is likely that within seconds of reading that newspaper announcement about her competitor’s promotional campaign, the sales rep had already processed in her mind the information, evaluated its importance and formulated the outlines of a plan. As the sales call approaches she will refine her thinking even further, and may in fact defer the final decision about how to make the offer until minutes before the meeting, depending on whatever factors she believes to be in play at that time. It would seem, then, that the prudent thing to do would be to override the system’s price recommendations. After all, the computer’s algorithms are not privy to any of the knowledge and insights that are firing up the neurons in the sales rep’s brain.
Overriding the system is not an optimal solution, however. In opting for human insight, the sales rep is losing the benefit of the scientific insights culled from a highly sophisticated analysis of patterns in past activity. That analysis is highly granular – it applies to the unique characteristics of this particular customer and the set of products that are featured in the sales call. Surely there is enough value in these insights to give our sales rep pause before she hits the override button?
Fortunately, she won’t have to make this choice between human instinct and scientific analysis. One of the most important recent innovations in predictive technology platforms is the means to provide real time input from the field – even in the moments right before a live sales opportunity – into the system. We call this making the system “market aware”.
Market aware systems can confer distinctive advantages. The key is how to essentially “code” qualitative input from decision makers into the algorithms that run the scientific analysis. For example, our sales representative needs to convey to the system that the sale will likely be under more competitive terms than usual. In other words, she needs to take a qualitative statement along the lines of “this is going to be a tougher sale because of our competitor’s aggressive promotion” and translate it into a digitized language the computer will understand. There are different ways this can be done – from a relatively simple algorithm numerically ranking competitive intensity to more complex machine learning analytics. Machine learning is a fast-developing field helping to make commercial technology solutions increasingly adaptable to constantly changing environments.
What happens from a practical standpoint is that our sales rep codes the information into a feedback loop that interacts with the system’s algorithms – effectively “informing” them about something happening in real time that they need to take into account. The models are then able to recalibrate their analysis and provide new recommendations to the salesperson. Now she has the benefit of two important – and distinct – types of analysis: the quantitative knowledge of historical demand patterns with this customer and the basket of goods she is trying to sell; and the qualitative judgment around an ambiguous situation that requires brainpower more than computational power. Combined, these two analytical approaches provide what we call quantitative intuition – a means to combine deep granular insights from your company’s transactional data records with the ability to adapt to dynamic market developments in real time.