The Candid Voice in Retail Technology: Objective Insights, Pragmatic Advice

Demand Forecasting in an Omni-Channel World: Far to Go

Membership: Unknown
Status: Unknown
Private: FALSE

The economic disruption of the last decade and a sharp increase in consumer cross-channel behavior have forced retailers to re-evaluate their demand forecasting capabilities. The good news, according to a new study from RSR entitled Crystal Ball 2.0: The State of Retail Demand Forecasting (to be published this week), is that they have indeed managed to take these shocks mostly in stride. The bad news is that while they understand that forecasting as a discipline is going to have to change to continue to be successful, they have no idea what changes to make or how to make them.

Industry thinkers (including some of us at RSR from time to time) have argued the merits one version of the truth for important information assets, and tangentally, generating demand forecasts from granular, unaggregated consumer demand signals collected from every in-house source, not just from the store. What makes these arguments sound even half-way reasonable is simply that today’s technology can support them. For example, Oracle’s Exadata or SAP’s in-memory database technology have been shown to be able to plough through a stunning amount of raw fact data to aggregate, calculate, and present results at speeds which are truly breathtaking. Likewise, today’s high speed always-on networks are able to transport raw data from transactional systems to central data munching engines effortlessly (compared to sound and video data, numbers and letters are lightweight to a network). Finally of course, there are all kinds of new demand signals being generated by consumer facing technologies such as social media, e-commerce, and mobile commerce, that theoretically can save retailers a lot of time and money by enabling them to catch changes in demand much earlier than before. It all sounds good.

Reality Intervenes

In our recent report on demand fulfillment in the omni-channel age (Omni-Channel Fulfillment and the Future of Retail Supply Chain), one of the big issues retailers are grappling with is that although many want to be able to adopt a sell anywhere from anywhere retail model, their legacy supply chain models are not well suited to an omni-channel world. Likewise, many retailers’ demand forecasting capabilities are built around a store-centric demand model. And response to the demand forecast survey show that retailers still struggle with legacy store issues related to exceptional inventory- extremely fast or slow movers and intermittent items. Going from a relatively simple store-centric demand model to a wildly complex omni-channel one ain’t peanuts.

The study shows that while retailers have applied demand forecasting to an increasing number of operational situations as they try to optimize their businesses, they remain challenged to deliver accurate forecasts in many of the applications (Chart 1).

What’s The One Right Way To Forecast?

After all the years of discussion between retailers, analysts, technology providers, and academicians, one might be forgiven for supposing that there is one right way to forecast, and a bunch of wrong ones. But such is not the case, at least as far as retailers are concerned. Our study shows that there are very differing attitudes about how best to execute demand forecasting for the business (Chart 2).

The most popular view (supported by the largest group of Retail Winners) is that different uses require different forecasts that should then be reconciled across the enterprise. The next most-popular answer fall to the other extreme: one single demand forecast to rule all functions — a true single version of the truth. Non-winners favor this option more so than their peers.

It would appear that retailers face a trade-off dilemma: one forecast for all means that everyone starts at the same point but may well end up in completely different places once the forecast moves from prediction to reality. Multiple forecasts mean everyone may start from different places that more accurately fit their needs, but prevent a retailer from gaining an accurate view of the overall picture of their business. The dilemma simply underscores the fundamental dilemma that retailers face around increased use of demand forecasts: do you let the channels plan their demand and then aggregate it up, losing important nuances in assumptions? Or do you start with one version of demand and let channels adapt that forecast for their specific needs, potentially losing the ability to reconcile changes later on? Unfortunately, most retailers are not yet thinking that holistically about their business — certainly not at the level of demand forecasting. And that may be the biggest business challenge of all.

Technology Is Just Part of the Answer

As we said at the top of this article, a big reason why there is so much discussion about the right way to forecast demand in this high-speed omni-channel world, is that technology enables us to move reality and theory closer together. And the new study does show that regardless of what view retailers have about the best way to deliver on the promise, technology is needed to make it happen. But at the end of the day, it’s really a process issue, and the study shows that for retailers who have already invested in tooling their forecasting processes with technology, the best near term opportunity is to get better at the process.

On the other hand, those retailers who don’t assign a lot of value to forecasting capabilities and haven’t (or don’t yet plan) to invest in the technologies to support them, will inevitably find themselves struggling to keep up with retailers who have used the discipline and technology to increase service levels to consumers and position the right inventory in the right place at the right time, without having resorted to the sin of too much inventory, or conversely risking the loss not only of sales but also of customers — all because of out-of-stocks, the wrong inventory, or the wrong value.



Newsletter Articles May 3, 2011
Related Research