Why predictive analytics is not a magic ball
Learn what predictive analytics can and cannot do.
Predictive analytics enables companies to determine forecasts and probabilities of occurrence from customer data. The result is, for example, a sales target that you can probably achieve. However, many people make a mistake when interpreting these results.
In another blog article, I already described that managing directors and sales managers would like to know why certain events occur or not, why specific customers churn. Why one product does not sell or sells more difficult than another. Why a promising lead does not turn into a customer in the end and remains a non-binding contact.
We more or less learned the striving for 100 per cent certainty. People look for rational reasons for the occurrence or non-occurrence of events. Especially in business, this comes into practical use when it comes to hard KPIs and controlling. For example, suppose you have a (simple) explanation why a company did not achieve the targeted sales goal. It is easier to communicate this within the company than if the cause is not so exact.
Predictive analytics asks, “What will (probably) happen?”
Sales executives are, therefore, on the lookout. They are looking for missed sales targets in the past to draw what they believe to be the right conclusions and correct their targets. They are also looking for the most reliable clues possible to help them set future sales targets. They are increasingly trying to find these clues with data analysis methods based on artificial intelligence (AI).
These analytical methods used to forecast (probable) future developments belong to the field of business analytics. On the one hand, this involves backwards-looking data analysis, i.e., data analysis related to the past, and on the other hand, it involves forward-looking forecasts and decision recommendations.
The aim here is to achieve predictions for situations and events that are as reliable as possible.
That is precisely the goal of predictive analytics.
If one wants to describe this field of expertise using a concrete question, it is:
“What will (probably) happen?”
Where B2B Sales makes a mistake.
Imagine a sales manager in wholesale who is supposed to set the annual sales target for product A. He has to find out what the target is. After he has concluded that algorithms are reliable calculation artists, to create forecasts for (probable) sales from a vast amount of customer data, he uses predictive analytics software. It calculates a possible turnover, say 10 million euros.
The sales manager rejoices and believes his sales target. But he is making a thinking error!
Let’s remember the question that paraphrases the Predictive Analytics method: “What’s (likely) to happen?”
Applied to our example, this means: it is probable that the wholesaler will achieve annual sales of 10 million euros with product A “Probable”, however, is not the same as “true”.
True would mean that the forecast result, i.e. the 10 million euros, is equal to the sales target. That the result equals the value that the sales manager is supposed to issue as the sales target.
However, this conclusion is wrong. It is “merely” the most likely amount to be converted at the end of the year. Predictive analytics has nothing in common with a crystal ball, which magically predicts the future and 100 per cent correct.
Forecasts are always only orientation values.
But what should the sales manager do now if the probably expected turnover is 10 million euros? What sum should he issue as an actual target? The answer: That is still at his discretion. In our example case, 10 million euros is merely an orientation value. An orientation value that you can very probably achieve. “Can” means, of course, that in the end, an upward and downward deviation is always possible.
So whether the sales manager sets the sales target at 5 million or 15 million euros is still up to him. Whether you achieve a sales target depends on many factors. Some of them you don’t even known at the time you set the mark.
Let’s compare this with the travel route to a specific place Y. To get to Y we enter it as a destination into the navigation system. The navigation system then provides us with the fastest route to our destination, based on a complex data analysis. The result is the route and the time required to get there.
However, this result is also one that is “likely” to occur. No one can say with 100 per cent certainty that you will need the three hours of driving time calculated by the navigation system to get from X to Y.
The result is a route that is “likely” to occur.
On the way, for example, we may have an accident.
Or we might get caught in a traffic jam we hadn’t foreseen yet. Or the weather changes and we have to adjust our driving speed to the weather conditions. Or …
With a navigation system, it’s like predictive analytics: Here the traffic data, there the customer data. Here the probable travel or arrival time, there the potential sales. In both cases, the result is forecasts and probabilities of occurrence. These are orientation values that are generated from an extensive data pool and support decisions.
However, this has nothing to do with a crystal ball that one hopes will accurately predict an event that will occur.
Predictive analytics creates an edge of information.
Knowledge and information in the form of data have become an undeniable competitive advantage, helping companies and decision-makers prioritize. A predictive analytics solution also creates new information at the end of an analysis process. You can use it to gain a competitive advantage over those not using these processes.
In predictive analytics, knowledge and information are available in the form of data with the character of orientation values, i.e. forecasts and probabilities of occurrence. Such information is precious for supporting decision-making.