Web-enabled systems help retailers better forecast demand

Cooking.com is on a roll. Having already added new sites this year to its stable of online stores operated for big-brand partners including StarbucksStore.com, BettyCrockerStore.com and PillsburyStore.com, last month it launched another, FoodNetworkStore.com.

For Cooking.com, which controls inventory for each of its partner sites, where it sells a mix of its own brands and those of its partners, the multi-site strategy has translated into continued robust growth in revenue, which rose 25% last year to $50 million at its flagship site Cooking.com alone.

But additional site operations mean a huge and complicated expansion in the volume of sales it must forecast to ensure it has the right products on hand to produce the highest sell-through rate. And with its products occasionally featured on NBC’s “Today Show,” the national publicity can cause sales of some products to suddenly surge, explains Bryan Handlen, Cooking.com COO.

“We can get a 100% increase in year-over-year sales of some products,” he says. So Cooking.com can use all the help it can get in planning adequate supplies for future sales, he adds.

The retailer, in the middle of converting its e-commerce platforms to the latest version of Microsoft Corp.’s .Net technology, gets that help from a demand forecasting application from JustEnough, which helps the merchant get a handle on what it can expect to sell across its multiple sites in the weeks and months ahead.

The winds are blowing

Cooking.com is not alone. Facing pressures posed by fragmented markets, increased competition and, of course, fickle consumers, retailers are forced to get a better handle on which way consumer whims are blowing to better forecast what and how much they are likely to sell.

Market research shows they need help. 50% of consumer products retailers and suppliers take more than one month to sense changes in customer demand, research and advisory firm Aberdeen Group Inc. notes in a July 2007 study of more than 140 companies, “The Responsive Supply Chain: Managing Market Events in the Consumer Goods Industry.”

“This is unacceptable in today’s business environment,” says the report’s author, Nari Viswanathan, research director for supply chain and logistics.

More retailers are using some form of demand forecasting, web-enabled systems or otherwise. Web-enabled demand forecasting systems like the one used by Cooking.com are making it more feasible to pull and share information needed to forecast sales. Of the companies in the Aberdeen study, 70% were showing significant improvements in out-of-stock situations by better planning merchandise according to customer demand.

Taking control

Concern over demand forecasting is pushing more retailers to take direct control of their available inventory, experts say. “We see demand for better demand forecasting driven by many retailers taking more responsibility for sourcing products themselves rather than relying on an outside agent,” says Kathryn Cullen, an analyst and principal at retail consultants Kurt Salmon Associates. “Retailers are providing information to their vendors on what they think they’ll need over the next year or two.”

One big trend is retailers and consumer products suppliers getting involved in more fact-based discussions, says Eric Peters, CEO of TrueDemand, a vendor of web-enabled demand forecasting technology. “Wal-Mart started the trend with its RetailLink, and now more retailers and suppliers are sharing point-of-sale data, and retailers are asking suppliers to help them make better business decisions,” he says.

“It’s all about that moment of truth when the consumer is at the store shelf,” Peters adds. “The demand forecast is an important part of that equation, and retailers and suppliers are talking more about how they can own that consumer experience at the store shelf. Instead of just viewing the retailer itself as the customer, more suppliers are viewing the retailer as a vehicle to the consumer. So the retailer is a partner.”

Retailers and suppliers have for years shared information for basic replenishment of commodities, though in many cases such efforts have been hampered by slow exchange of information shared on electronic spreadsheets and paper documents. In contrast, today’s demand forecasting systems, relying heavily on web-enabled technology to share information among a retailer’s internal databases as well as with trading partners, go beyond more basic replenishment procedures, which are more short term and based on more limited data on recent sales.

“Demand forecasting is a cut above replenishment,” Cullen says. “It factors in more data, including prior-season sales, to see what retailers will need in the next season. So it’s more sophisticated and aimed at the long term.”

Developments in Internet technology—providing the ability to integrate information among multiple applications to produce demand forecasts, and universal browser access to information—are at the core of newer demand forecasting systems, Cullen adds. “Web enablement is very important to demand forecasting,” she says. “Most sourcing tools are also web-enabled, which is critical so retailers can share information with suppliers around the world.”

XML programming language, web services and application integration supported by web-enabled service-oriented architecture, or SOA, play a crucial role in enabling demand forecasting systems to gather and analyze disparate information—including point-of-sale data, invoices, promotional activity and inventory records—that can be sorted and analyzed using special algorithms for producing forecasts.

Cooking.com maintains about 7,000 product SKUs in its distribution center, but also manages information on close to 100,000 more SKUs either drop-shipped by its suppliers or kept in facilities of its merchant partners.

Before deploying the JustEnough Unlimited Demand Forecasting and Inventory Planning system three years ago, Cooking.com would take several days to first analyze inventory stocks, enter SKU details for each product into a purchase order, then fax the orders to suppliers. In the meantime, managers responsible for bringing in new products in time to meet expected customer demand had no way to quickly tell whether some SKUs were already out of stock, resulting in lost sales and disappointed customers.

“Before we started using the system, we didn’t get to all SKU items with the frequency we can today,” Handlen says. “Now we get a more thorough review with fewer people involved. And because the system factors in more information and uses algorithms, we get a more sophisticated forecast.” Among the results is a 38% reduction in inventory costs because of the ability to maintain leaner stockpiles without risking out-of-stock situations, he says.

The JustEnough system is designed to automatically pull information from multiple applications, including order management, transportation management, procurement, warehouse management and, for store and multi-channel retailers, store point-of-sale. It then produces graphical reports including recommendations for matching purchase orders to forecasted demand.

IRCE 2008 Conference Multi-Media CD-ROM

Cooking.com’s application focuses on daily updates of orders and inventory levels, and it tracks the time from when purchase orders are created to when goods are received in the retailer’s warehouse. This enables the retailer to compare the performance of different vendors and match promotional schedules with the expected time of delivery while also deciding which and how many products to order, Handlen explains.

Cooking.com has deployed the licensed, on-site version of the JustEnough system integrated with a .Net platform. JustEnough also offers an on-demand system delivered over the web.

Demand forecasting can be most difficult for short-lifecycle items such as apparel, which require additional aspects of forecasting. At high-end apparel merchant Brooks Brothers, the use of multiple forecasting tools from JDA Software Group Inc. has boosted accuracy in forecasting demand over the past year from 35% to 60% in its retail stores and from 27% to close to 100% in its outlet stores, reports Heidi Loncki, system administrator. “Our goal is to have 90% to 100% accuracy for every SKU in every store,” she says.

Brooks Brothers uses the web-enabled JDA Demand application within JDA’s Supply & Demand Optimization suite, incorporating technology from Manugistics, which JDA acquired last year. JDA Demand is designed with five statistical forecasting methods that consider separate data characteristics of individual products to forecast demand based on past sales, a process that can be most effective in forecasting consumer goods like paper towels that remain in the market for a long time, says Paula Natoli, product director of JDA’s demand and collaboration applications.

Seasonal tools

But retailers of products such as consumer electronics and fashion apparel, which frequently change in features and styles, usually need additional means of forecasting demand. “Sometimes there isn’t enough history for a mathematical solution to generate a forecast. It’s like trying to find a pattern, but you can’t with only a few pieces of information,” Natoli says. “Then companies may want to use lifecycle or seasonal forecasting tools.”

Brooks Brothers uses JDA’s Seasonal Profiling application, which generates seasonal profiles for individual products to augment sales forecast data in JDA Demand. These profiles figure seasonal fluctuations in demand based on such factors as weather patterns, holiday periods or school schedules, and also apply these factors against local as well as regional or national markets. Combined with general sales forecasts, the seasonal profiles led to the sharp increases in forecast accuracy at Brooks Brothers, Loncki says. “We substantially increased forecasting accuracy with the new profiles,” she adds.

To prepare the Seasonal Profiling application to produce effective profiles of products, Brooks Brothers worked with JDA analysts to factor in the proper set of data points, such as the expected availability of SKUs, new store locations where particular SKUs are expected to be in demand, and how weather patterns could affect seasonal sales.

End-of-season sales then are compared against forecasts. If forecasts come in outside of expected accuracy parameters, Brooks Brothers will go back and analyze data inputs to see if either individual profiles or the overall forecast were poorly prepared. “If it’s the profile, maybe the SKU doesn’t need a profile,” Loncki says, adding that sometimes it’s more effective to combine data on more than one SKU to develop a new profile.

In addition to drilling down to demand forecasts for short-lifecycle products with particular seasons, more retailers are also trying to focus more on individual stores and local markets as opposed to entire chains and regional or national markets, experts say. While some older demand forecasting systems, particularly those designed for the largest retailers, may focus more on demand across an entire chain, “more retailers today are recognizing a need for more store-level forecast data to address top-down forecasting requirements,” Aberdeen’s Viswanathan says.

Rather than planning just from the bottom up, or starting with companywide forecasts and then fitting them to individual stores, a top-down application starts with forecasts tied more closely to each store’s product mix, factoring in things such as regional consumer interest, weather-related demand, and local promotional activity by stores and local competitors.

Planning ahead

Web-based integration supports this further by providing more seamless data sharing among assortment planning and forecasting tools, which help retailers plan different product line-ups for different stores, JDA’s Natoli says.

Other store-focused demand planning tools include the direct-store-delivery tool HighJump Forecast Advantage, from HighJump Software, a unit of 3M. Launched earlier this year, the Forecast Advantage application is designed to let direct-store-delivery vendors, which manage their own inventory, such as baked goods in convenience stores, view on web-based handheld computers demand forecasts in a corporate database based on prior sales, current inventory levels and additional data such as expected sales from holiday weekends.

By viewing such forecasts while tending to each store, a direct-store-delivery driver can see how many items to allocate to each location, says Chad Collins, vice president of global strategy.

Regardless of what demand forecasting technology can do, retailers still must blend a good mixture of art and science in planning products and sales, many experts say. One of the causes of inaccurate demand forecasts, for example, is failing to factor in prior sales lost because of poorly conducted promotions.

“We can show retailers that at 32% of their stores last year they didn’t have the right price on a product because maybe the store or department manager forgot to put the promotional price up, so they sold 14 units a day instead of the expected 17,” TrueDemand’s Peters says. “So then we tell them in our forecast not to roll back their next plan to 14 units a day, just do the promotion correctly and plan for 17 units.”

TrueDemand, which caters mostly to large retailers, bases the licensing costs of its software—ranging from $200,000 to $2 million a year—on a retailer’s number of SKUs and stores, Peters explains.

Another tactic that helps support demand forecasting systems is simply holding back some inventory. “Some of our clients know how to allocate products across their stores, but they’ll hold 20% at their distribution centers to wait and see where demand is going, then ship that 20% to the strongest locations,” says Namita Dhallan, group vice president of product management at JDA.

At Cooking.com, the best demand forecasting technology can only go so far. “We use our demand forecasting system for managing how our business normally runs,” Handlen says, “but it still takes a guess about how much to buy after a product is featured on the ‘Today Show.’”

paul@verticalwebmedia.com