Skip to main content

The time machine: Predicting future demand flows across large portfolios with unprecedented precision

Our new approach to transferable demand flows does justice to the complexity of today’s portfolios and consumer decision journeys by leveraging advanced analytics, entropic principles and machine learning.

What if...? Time travel, alternate futures and chains of causation have long fascinated philosophers, writers, and filmmakers. But for executives in charge of large portfolios, “What if ?” is more than an entertaining thought experiment.

For them, real value is at stake. What if we increased the price of a specific product? Would customers keep buying it? And what if we ran a promotion for one of our brands? How would this affect other brands in our portfolio? What if we reduced the size of our assortment? Would the savings outweigh the losses?

In this article, we present a new approach to answer these and similar questions. We haven’t actually built a time machine, but from the perspective of pricing and assortment managers, our new approach to demand-transfer modeling may well be the next best thing. It leverages advanced analytics, entropic principles, and machine learning to predict future flows in demand across large portfolios with unprecedented precision and reliability.

The question of how demand will change subsequent to a price increase, a promotion, or a portfolio restructuring program isn’t new. In the past, companies have often relied on experience and intuition to ‘guesstimate’ such effects. But as portfolios are growing, consumer decision making is evolving, and commercial cycles are getting shorter, the experience of even seasoned experts is insufficient for robust portfolio decision making.

Some players are using discrete choice models and econometric modeling to predict the impact of demand flows. While such methods work well for limited ranges, they are unable to account for the cross-effects of pricing and assortment moves across large portfolios at the unit level. Most existing approaches are based on predicted consumer switching, rather than actual switching behavior. In effect, the output of traditional models is neither as precise nor as reliable as it should be for fact-based, granular portfolio management that captures interactions across all SKUs and drives sustainable revenue growth for the whole portfolio.

Please complete the form below to DOWNLOAD A COMPLIMENTARY COPY OF this content.

For more information, please contact us.
*
*
Please provide a business email address
*
*
*
*
*

McKinsey & Company is committed to protecting your information. Your information will be used in accordance with the applicable data privacy law, our internal policies and our privacy policy. As McKinsey is a global organisation, your information may be stored and processed by McKinsey and its affiliates in countries outside your country of residence, but wherever your information is processed, we will handle it with the same care and respect for your privacy.