Merchandising is typically based on planning two ways – top down and bottoms up. In the top down scenario, the retailer declares a growth target: “We plan on growing sales by 5% next year.” And then it is up to merchandisers to figure out where and how that growth will come. Typically, this kind of planning stays in dollars (or whatever currency you use), and simply breaks down the numbers into smaller and smaller planning buckets until budget goals are set for locations and for rough categories of goods. Then the bottoms-up process kicks in. Retailers look at unit sales by location and roll those unit sales up to an enterprise total for number of units that are needed, typically by season or some other defined time period.
If the retailer is on the ball, they then reconcile these two – the expected budgets for revenue and the expected number of units sold – to see if they come anywhere near each other in terms of expectations. If the number of units expected to sell doesn’t yield the expected revenue budget, either the number of units to sell needs to increase, or the price needs to be higher to close the gap.
Where is the customer in all of this? Theoretically, her demand is expressed by tagging it to a location. That works great when the only locations you have are stores. But it doesn’t work when consumers can cross channels at will. Consumers are not single-channel creatures, and increasingly, neither is the inventory in those channels.
Merchandising will have to cope, but how? The answer is to plan by customer, and then figure out which channel customers are most likely to use to buy each type of product.
That’s all well and fine in theory, but the reality is messy. Even trying to simply visualize that planning cube is tough – merchants typically plan by product vs. location vs. time. Now we’re talking about adding a fourth?
Even if you substitute customer for location – product vs. customer vs. time – how do you plan by customer? You can’t really forecast by customer and roll it up. That has enough messy variability in it to be nearly impossible when talking about trying to forecast product by location. The SKU/location combinations are just now something that technology can take on without choking on it. Try SKU/customer combinations – it’s a bit mind-boggling. As a consumer, I can’t predict my own spending behavior beyond a general bucket of “clothes” budgeted yearly. I can’t imagine asking retailers to predict my behavior when they have to first guess my budget and then guess their likely share of that budget.
Forecasting by customer segment is just as messy. Granted, it reduces some of the variability mess to consider customer groups rather than individual customers, but which groups? Should merchandising use the same customer segments as marketing? Marketing increasingly, thanks to customer segmentation and targeting software, creates segments on the fly – how can merchandising respond to that? Should merchandising plan based on desired or “strategic” customer segments, or should it plan based on the customer segments it currently has? Or should it use basic demographic segmentations from one of the big consumer data houses?
Planning by customer segment is just as messy as trying to plan by individual customer. Now throw in two new areas to think about: pre-demand signals and the demand “location”. Pre-demand signals are buying intentions expressed by consumers in digital channels. This is still a very new area of retail data, and most retailers don’t really know if this information predicts anything useful, nor if an accurate prediction can be had in enough time to do anything about it. Target and Lily Pulitzer had early warnings that the Lily for Target promotion was going to be way beyond their initial expectations, but was there anything they could really do about it? If the product they’d ordered was already on the water, there was no way to add to it in enough time to do any good.
The demand location is just as tricky, because in order to take the step of moving from customer demand to where that demand is most likely to be expressed, retailers have to have a good enough data set that tells them where consumers have historically purchased from in the past – online, store A, store B, etc. That means the retailer’s customer database has to be clean enough, and they have to have collected enough attributable sales to be able to figure out at the minimum some average shopping patterns. Consumers, for example, may tend to buy bathing suits in stores but jackets online. But maybe for children’s clothing the average demand pattern looks more like mine: I take my kids into the store to try everything on, buy a few things that fit, and then go home and buy all the rest online once I know the sizes. Retailers will only know this if they have enough data to piece together the puzzle of how consumers shop across the retailer’s touchpoints.
The bottom line here is that merchandising is going to need to change. The way that retailers plan today will not be the way they plan in the future, because they will need to do more to take into account how customers cross channels as they shop, rather than pretend that online and stores are completely different demand entities. Most retailers don’t have the ability to inject customer into the planning process, and most vendors have only been thinking about this for a relatively short amount of time. Even for the ones who have been thinking about it a lot, challenges remain – none of this is easy.
But one thing is true about omni-channel: the change it brings is relentless. Retailers will either adapt or wither away when that change hits them. Right now, retailers are being tested on how well they can adapt their supply chains. But merchandising will be next.