Walking the forecasting TIGHTROPE

May 01, 2006 9:30 PM  By

Forecasting inventory for seasonal demand takes philosopher George Santayana’s dictum “Those who cannot remember the past are condemned to repeat it” and turns it around to “Those who know a product’s history are able to repeat it.” In short, being able to predict the customer demand for a future season’s product requires knowledge of past product performance along with the intuition of seasoned planners.

The process is neither magic nor science, says Debra Ellis, principal of operations consultancy Wilson & Ellis Consulting in Barnardsville, NC. “Inventory management is real simple:It boils down to balancing customer service with risk. The talent is to find the level that’s profitable for you.”

Forecasting methods vary from business to business and from channel to channel, as well as according to company size. Like contact center management, inventory forecasting can be seen as where the front end and the back end meet.

GETTING STARTED

Generally speaking, buyers go to markets, buy initial sample stocks, and begin the inventory management process by negotiating back-stock and buy-back criteria. Once the merchandising department has decided which items are to go in a catalog or on the Website, forecasting specialists — sometimes known as control buyers or rebuyers — will look at factors such as product history, expected response rates, and the size of a catalog mailing to estimate total demand.

Then merchants allocate money accordingly, first to larger categories of items and then breaking down those allocations by item. “It becomes a system of checks and balances,” says George Mollo, president of Nanuet, NY-based merchandising operations consultancy GJM Associates.

Talented forecasters with several seasons of experience can be invaluable to a company’s bottom line. Ellis describes forecasting as “60% history, 20% intuition, and 20% luck.” Even so, she says, “a really good inventory planner will have a 90% accuracy rate.”

Obviously, experience and training are indispensable for someone who handles forecasting. “It probably takes about three seasons for a forecaster to learn the business,” Mollo says, “if you count each season as being about six months.” In smaller companies the role of creative merchant and analytical rebuyer may be combined in a single person.

In any case, forecasting involves in-depth analysis of numbers, both for historical sales and for the ongoing curves of sales coming in. Software forecasting tools take those numbers as they come in, “once the book is active, and show item relationships,” Mollo explains. “The system will take the actual number of units, follow the [sales] curve, and see that you’re going to sell more or less than the forecast.”

Not that forecasting software is a requirement. Some multichannel merchants have been known to consistently achieve 90% or higher accuracy rates using just a spreadsheet, says Ellis. A case in point is Burlington, VT-based garden products merchant Gardener’s Supply Co. Peak seasons for Gardener’s Supply are fall, Christmas, and the spring planting season. Through the 2005 holiday season, Gardener’s Supply relied on spreadsheets and an experienced team to forecast catalog and Internet sales for its approximately 2,000 SKUs. The company’s accuracy rates have been consistently ahead of those for the industry at large, according to chief technical officer Chris Thompson.

Even so, Gardener’s Supply is in the process of implementing the Forecast 21 product suite from Omaha, NE-based Direct Tech. “We’re putting in Forecast 21 primarily for the merchandising focus rather than the purchasing and inventory control aspects, but we’re thrilled to be able to get that extra bang out of it,” says Thompson.

SORTING THROUGH SOLUTIONS

Companies that sells comprehensive warehouse management solutions (WMSs) usually offer some form of forecasting tool as well. For example, supply chain planning and execution software provider Manhattan Associates, headquartered in Atlanta, offers the Demand Forecasting product, available as a stand-alone tool.

Typically, says Rod Dougherty, Manhattan Associates’ director of product strategy, forecasts developed with the Demand Forecasting tool are exported to some other module, such as a replenishment or merchandise planning program. The tool can be used for catalog, retail, or online forecasting; Manhattan Associates has customers that use it for all three simultaneously.

Costs associated with the Demand Forecasting tool are based on company revenue and number of locations. “The big variable is how complex their structure is,” Dougherty says. As for implementation time, “it’s tough to say, but it would probably be somewhere in the neighborhood of 30-60 days of consulting at the minimum.”

Multichannel solutions provider Ecometry, based in Delray Beach, FL, provides basic forecasting functionalities in its core Ecometry Suite, but it also offers a stand-alone Inventory Forecasting Module that includes more-sophisticated features. For instance, says Brian Dean, vice president of strategy and marketing, “it provides real-time analysis of how products at all levels are running, including product category, style, SKU, and special offers, and it dictates when purchase orders are needed.”

A two-user license is included in Ecometry’s base price of $25,000; additional licenses are $10,000 each, and the annual support costs are 18% of gross sales. Given that forecasting is “as much of a science as an art,” Dean says, “I would estimate our accuracy to be between 85% and 98%.”

CUSTOMIZE ME

No matter how sophisticated the software it may employ, every company develops its own forecasting process to suit its particular needs. “You have to customize any system, dedicated or modular,” says Ellis. “It all boils down to a company’s degree of understanding how inventory management works in their industry.” Some companies use the forecasting module associated with a WMS or an order management system (OMS) in addition to a dedicated forecasting solution.

Cranston, RI-based jewelry and gifts merchant Ross-Simons, for instance, runs Ecometry as its main management system but also uses a separate, dedicated forecasting tool. For Internet and catalog forecasting (Ross-Simons’ 15 stores have a separate planning and buying department) Ross-Simons uses IF/SO, a retail/catalog forecasting solution from Sausalito, CA-based Forerunner Systems, in combination with inventory, purchase, and sales information uploaded nightly from Ecometry.

“IF/SO is the primary planning tool to plan future forecasts and buys,” says Stuart Greengart, Ross-Simons’ director of planning. “We use [database management software] Microsoft Access to make buys based on current sales that include both catalog and Internet sales based on info that is uploaded from Ecometry.”

Internet sales, which account for approximately 45% of Ross-Simons’ total revenue, cannot yet be loaded directly into IF/SO. The IT department, says Greengart, is working to create a flat-file table that can generate matchback reports for Internet sales to IF/SO. Once that happens, he continues, IF/SO will become the primary buying tool as well as a planning tool, and Access will become a supplement.

WEB SALES VS. STORE SALES

For Ross-Simons, the desire to integrating Internet sales information with its forecasting tools coincides with a recent milestone for the company: “We sold more on the Web in total in December than we did in all the [catalogs]. That was the first month that it happened,” Greengart says. Although the company’s 13 annual catalogs are the main driver to its Website, there are some differences, Greengart points out: “Stuff on the Website sells more like retail than it does like a catalog.”

In fact, Greengart says, “the whole concept of buying from the Web is almost like buying from a store. It’s what’s there on the shelf, what you present, what you promote; it’s not like [catalog] sales curves, sending the mailing out and waiting for that response.”

Forerunner Systems president/CEO Lewis Richmond agrees that there are similarities between Web and store sales and inventory patterns. “Catalog planning is strictly bottom up. You look over the items, see how an item sold in the last catalog, and then you build up the category from that,” says Richmond. “Store retailers aren’t so concerned about the individual items,” because they’re not publishing a catalog that says the products are readily available.

Richmond developed IF/SO a decade ago to meet the need he saw for an effective, catalog-oriented forecasting system. Richmond notes that catalog sales retain the structure and the timetables of a publishing business, with merchandise required to be in the warehouse as much as 36 weeks ahead of a catalog drop. “It’s an item business, in a way that store retail has never been,” he says. Stores are assortment driven rather than item driven. Nonetheless, Richmond says, large retailers with smaller catalog divisions may very well use the same forecasting solution for both channels.

Forerunner Systems’ catalog clients generally have sales of $50 million-$500 million. Installation costs for IF/SO vary depending on the size of the company but typically run in the mid- to high six figures.

Peoria, IL-based Direct Logic Solutions offers another forecasting option. Its DirectIntegration Suite is designed to manage, analyze, and facilitate catalog planning, e-mail planning, and online planning, according to CEO Chris Cusack, and produces forecasting accuracy rates of more than 90%. The solution runs on a Microsoft Sequel server and does not require significant integration.

“All we require is that [the client] give us access to their order entry data, their inventory system,” Cusack says. Midsize companies could expect a basic planning module that would target its analytics and forecasting ability to cost about $30,000.

Direct Tech’s Forecast 21, the solution being implemented by Gardener’s Supply, was created primarily for companies generating at least $25 million in annual sales. It handles multichannel forecasting with an emphasis on Web and catalog functions. “The costs to implement Forecast 21 will range from $100,000 to $225,000 for an unlimited user license,” says Direct Tech president Craig Harding, who adds that the company is in the process of developing a Forecast 21 Lite product for smaller companies. Direct Tech also offers a merchandise assortment planning solution in the $65,000-$100,000 range, depending on the number of users.

WHAT THE NUMBERS WON’T TELL YOU

No matter what degree of automation you apply to your forecasting crystal ball, and no matter how talented your buyers and rebuyers, a certain element of merchandise forecasting depends on intuition. And sometimes, no matter how much careful, detailed, experienced analysis has gone into producing a forecast, an unexpected wind can blow your calculations off-balance.

At Gardener’s Supply during a recent holiday, the inventory and merchandising teams were looking at a newly printed catalog when they realized that they’d underestimated demand for one modest gift item. “The creative people did just an amazing job presenting it on the back cover,” Thompson recalls,” and as soon as we saw the catalog in the office, we said, ‘Oh, my God, we’re so in trouble.’” Sure enough, the item in question “blew out,” he says. “We were chasing that like you wouldn’t believe.”

Ross-Simons’ Greengart had the same experience with a pearl necklace featured on the cover of this year’s spring preview catalog after the creative folks worked their magic with the presentation. “We had planned for that to do very well, but it’s going to do about $150,000 — just that one page,” says Greengart.

Ed Foy, CEO of eFashion Solutions, an e-commerce operations management firm that oversees online sales for brands that include Apple Bottoms, Members Only, and Baby Phat, remembers a similar instance from the 2004 holiday season. All of a sudden order numbers for a particular item shot through the roof. “We have an intelligent system that notifies us, ‘Something crazy’s going on here. You’re selling through at 15% of inventory, and you’ve only had it in the warehouse for seven days.’ Hello, there’s an alert. We look at the Website; traffic is converting like crazy. What’s going on? Then we find out that Apple Bottoms is on Oprah Winfrey’s Christmas list.” There was no way to meet the demand, and worse, the return rate for that item shot up because the prepack assortment hadn’t included enough larger sizes.

Even though these sort of circumstances can’t be forecast, they can at least be somewhat mitigated through consistent, thorough oversight. Debra Ellis sums it up this way: “The trick to inventory management is the management. You don’t do it once a month or once a week; it’s day after day, and ideally the same person doing it. It’s just looking at the numbers, reviewed by the human eye.”


Barbara Arnn is a freelance writer based in Port Townsend, WA.

FORECASTING CHANGES IN FORECASTING SOFTWARE

The greatest change in forecasting during the past several years has been the arrival of the Internet as a sales channel. Many merchants keep Web and catalog inventory together and forecast for both channels; multichannel merchants with brick-and-mortar stores are far more likely to handle their store inventory using a different forecasting solution than they one they use for direct sales. “People are still feeling their way,” says George Mollo, president of Nanuet, NY-based consultancy GJM Associates. “No one that I’ve talked to has found a magic bullet for how to forecast on the Web.”

“Multichannel is the biggest issue facing people now,” says Debra Ellis, principal of Wilson Ellis Consulting in Barnardsville, NC, “and the [suppliers] that tackle it are ultimately the ones that are going to control the market.”

For that reason, forecasting technology may be due for a shake-up, if not an outright revolution, says Eric Olafson, CEO of Salt Lake City, UT-based retail software solutions provider Tomax. Olafson sees the newest trend as “multi-echelon” forecasting, planning, and replenishment, which includes the stores as well as the catalog and online channels — forecasting, in other words, that functions at the individual store level rather than forecasting for a distribution center. As Olafson says, “Customers don’t visit DCs; customers visit stores.”

Olafson views the recent history of retail as being driven by a “push” model of forecasting that focuses on DC, supply chain, and logistics issues. Future forecasting and inventory management will likely have a completely different structure, he says: “Forecasting around true customer demand in stores and online as the core truth from which to create your supply chain response, as opposed to pushing products out through DCs to stores in a more arbitrary or ad hoc way.”

Perhaps, although Ernie Schell, president of Marketing Systems Analysis in Ventnor, NJ, sounds a more cautious note. “The Web is a media-rich environment for analyzing lots of things about your sales, but it isn’t always necessarily a rich environment for telling you who your customers are,” he says.

But Schell does point out that the Web can serve as a rather sophisticated testing tool. “You could take any product that you wish and test it at various price points, minute by minute, changing the price points based on where the visitor happens to come from,” he explains. “So if they come from Yahoo!, they could see a different price than if they come from Google.”

Indeed, Ed Foy, CEO of eFashion Solutions, which oversees online sales for numerous apparel brands, considers online forecasting techniques key to future forecasting improvements. He believes that the technical ability to capture significant, copious, precise data from customer visits to Websites is something that retailers should be able to translate into any channel.

It is this intersection of untested technological possibilities and the need for precision in forecasting over more than one channel that represents the cutting edge. Online sales data offer information that brick-and-mortar stores have yet to compile: clickthroughs, traffic, conversion rates. “For the most part it is an unknown, but who has more data than a computer?” asks Foy.
BA