NO PAIN, NO GAIN 2004-03-15 (2)

MICHAEL GARRY

Ever since a package of Wrigley gum became the first product to be scanned at a checkout nearly 30 years ago, food retailers have dreamed about the ordering potential of scan data.

If you could collect data on everything you sold, couldn't you use that data to trigger store replenishment orders? Sounds simple, but it wasn't, and with a few well-known exceptions -- notably Shaw's Supermarkets and Meijer -- computer-based ordering using scan data remained outside the realm of most supermarket operators.

Over the last few years, though, as more third-party software packages have become available and as computing power has grown exponentially, some retailers have been able to enter the brave new world of computer-based ordering. Many more are "kicking the tires," said Barbara Anderson, a store-level replenishment consultant and president of bvac inc., San Carlos, Calif. "There is simply a ton of interest," she said.

At Food Marketing Institute's Marketechnics show workshop earlier this month in San Francisco, the chief information officers of two food retailers reported they had indeed tested computer-based ordering and found that it worked well -- if you were able to maintain the proper operational discipline in the store. The retailers -- Price Chopper Supermarkets, Schenectady, N.Y., and United Supermarkets, Lubbock, Texas -- expect a rollout of their respective store ordering systems.

Computer-based ordering actually falls into two categories: computer-assisted ordering (CAO) and computer-generated (also called computer-automated) ordering (CGO), explained Thomas Nowak, senior vice president of information systems/CIO at Price Chopper, one of the workshop speakers. To add a touch of confusion, the other workshop speaker, Scott Gilmour, vice president and CIO at United Supermarkets, referred to computer-generated ordering as computer-automated ordering. (Since this can also be abbreviated CAO, in this article CAO will only refer to computer-assisted ordering.)

As Nowak described it, CAO is the forerunner of CGO. In CAO, "people still make the final decision," he said. "The computer provides the information in terms of sales forecasts for the next few days, and historical sales, including promotions, but the final ordering decision is made by a human." In CGO, the more advanced scenario, "the computer takes over the ordering process," Nowak said. "The humans become inventory-control people."

Both processes make use of forecasts based on scan data to determine orders, but CGO also relies on a perpetual inventory system that keeps careful track of the inventory status of every item in the store.

Perpetual inventory tracks real-time store inventory based on sales, receivables and adjustments, such as returns, damage and spoilage; CGO takes that inventory calculation and incorporates shelf capacity, forecasted sales (including seasons, events and promotions) and service-level objectives in generating orders.

At Price Chopper, "we feel the promise of CAO and CGO are being realized," said Nowak. Yet, he cautioned that even though retailers now have the computing power and software to make both applications work, "this is not an easy process because of all the operational issues you do have to go through."

Gilmour shared Nowak's caution. "With an initiative like this, people in the organization can get a little nervous -- the chief financial officer, for example -- because you're changing from a manual inventory counting system to depending on an automated system," he said. "So, you need lots of disciplines in place." Among these disciplines are: scanning all items at the front end, including transfers and reclamation items; and accurate receiving and invoicing at the back end.

The Evolutionary Approach

Price Chopper, a promotional operator with 105 stores in the Northeast, began its foray into CAO in 2001 when it decided to test the IMI Retail system, from IMI Americas, Mt. Laurel, N.J., a division of Industri-Matematik International, Stockholm, Sweden.

United is also using a system from a European vendor -- SAF AG, Tagerwilen, Switzerland -- in combination with an application from Softechnics, Garland, Texas. Another vendor offering store ordering software is DCM Solutions, Irving, Texas; Ukrop's, Richmond, Va., is using DCM's system. Other vendors include TCI Solutions, Manugistics and SmartOps. Nowak said retailers in the past were limited by the absence of good third-party software in this area, though that is no longer the case.

As Nowak explained it, Price Chopper's motivation for looking at IMI's software was to address age-old retailing dilemmas: out-of-stocks, inventory carrying costs, and transportation efficiency. The chain also wanted to make store ordering simpler and more accurate, and make it less dependent on the knowledge base of employees who might leave the company or change stores. Anderson observed that less ordering means more time for employees to serve customers and merchandise products.

Retailers interested in using automation to drive store ordering can go right to pure computer-generated ordering -- the "big bang approach" -- or start with CAO and graduate to CGO. Price Chopper took the latter, evolutionary approach. "I'm a strong proponent of evolution, especially if you can get incremental benefits at each stage, which we did," said Nowak.

One of the key benefits of starting with CAO, said Nowak, was that it helped store employees get used to the idea of a computer doing some of the work. "People said the extra information was great, so it helped them to accept the change," he said. Forecasts proved accurate, even for promotional goods.

According to Nowak, it took Price Chopper four months to install CAO in the first store in all departments plus perishables, another four months to put it in another eight stores, and a final four months to roll it out to the rest of the chain.

CAO enabled Price Chopper to achieve "double-digit decreases" in out-of-stocks and in inventory carrying costs, said Nowak. An unexpected benefit was locating "lost items" -- an average of 200 authorized products per store that had somehow fallen out of the ordering cycle.

These results belied the notion that the benefits of CAO are limited, said Nowak. Indeed, Price Chopper did so well with CAO that it "paid for the entire project," he said. "So, the rest is gravy." Having an early payback was beneficial, he noted, because the subsequent phases -- perpetual inventory and CGO -- are not easy. Current prices were unavailable last week, but in 2002 SN reported the cost of the IMI System starting at $300,000.

Nowak disputed the notion that labor costs would increase with the installation of CAO. Instead of using paper reports, employees can find all the information they need in a handheld terminal, he said. "Overall, we didn't find any additional time involved with CAO," he said.

Even at the CAO stage, accurate POS data is important. In this regard, Price Chopper benefited from earlier implementations of software that required accurate data, such as labor scheduling and gross profit calculation. Eliminating the quantity key from the registers can also help improve scan data accuracy.

Also important is accuracy in warehouse shipping and store receiving. Again, in this area, Price Chopper benefited from the installation of a voice-directed selection system from Vocollect at the warehouse level.

CAO requires employees to make timely inventory adjustments, said Nowak. "You need to instill in your folks the need to research things, such as out-of-stocks, on a daily basis." To that end, the system provides employees with alerts for unusual orders or missing items.

The disciplines applied in CAO are even more crucial in CGO, which is far more complex. So, even though it had successfully implemented CAO chainwide, Price Chopper took a year to install perpetual inventory and CGO in one store, Nowak said. "That's not an indictment of the software, but a testament to what you have to go through. It's not a one-size-fits-all process." Among those hurdles: knowing how much space is available on the shelves; incorporating causal data for in-store displays; and addressing security issues in moving from a batch to an RF (radio frequency) system.

After the one-year learning period, Price Chopper was able to install CGO in all center store categories at the test store, including grocery, dairy, frozens and nonfood, in five weeks, said Nowak. The chain found CGO produced additional reductions in out-of-stocks, as well as improvements in back-room inventory. CGO is now mandatory at that store. CGO also enabled Price Chopper to pursue "fact-based shrink initiatives," determining the root cause of shrink in different categories, he noted. The "visibility in demand" also helps the chain's buyers and will help suppliers eventually, he said.

According to Nowak, Price Chopper now plans to roll out CGO to the rest of the chain in center store categories, though he didn't offer a timeline. He added that the system would also be applied to perishables and direct-store-delivery products. He predicted that CGO will become even more effective when RFID (radio frequency identification) starts to be used to improve inventory and receiving accuracy.

First in the U.S.

United Supermarkets' move to CGO began in 2002 after the chain had already used a CAO application from Softechnics for three years in all of its stores, which now number 47, said Gilmour. (He referred to CGO as computer-automated ordering.)

United thus became the first U.S. retailer to test a combined SAF/Softechnics CGO solution, with SAF providing the forecasting engine and Softechnics the perpetual inventory piece. Two stores were used in the test.

Like Price Chopper, United started out with an advantage in receiving; in United's case, it was that the chain was already receiving detailed invoices from its warehouse through NEX (network exchange) transmissions. United had two years of historical sales to use in the test, though SAF had to write a program to simulate the impact of promotions during that period.

Gilmour noted that the night before the test went live, store associates in both stores got an accurate count of targeted inventory in the back room, on shelves and displays. The inventory in the test included paper goods, canned fruits and vegetables, cereal, pet food and frozen foods.

After a three-month test in late 2002, United found the results to be "encouraging," said Gilmour. Out-of-stocks were cut by 58% in one store and 62% in the other; inventory volume was reduced by 17% and 22%, respectively. The results tracked closely to projections, he added.

Gilmour said the CGO system was set up so that inventory counts were only done in response to alerts such as zero inventory, too many zero sales, or more sales than forecast. About 6% of the items in the pilot triggered such alerts, he said.

United now plans to hire a "perpetual inventory coordinator" in a few months to handle training and prepare for a rollout of the system to all its stores, said Gilmour. Last week, he told SN the next step is the installation of the Retail.net new-item management system from Tomax that will support hosting at the stockkeeping level, which will help with inventory.

Gilmour expects the Tomax system to be installed by September; in the fourth quarter, United will install computer-automated ordering in all dry grocery, frozen, dairy and nonfood categories in the two test stores. That comprehensive test will be followed by a "management appraisal," though he said he expects the system to be rolled out chainwide.

United also plans to incorporate the SAF software with Spaceman space planning software from ACNielsen in order to improve sets and assortment, and take advantage of the "extra 10% to 15% of space this is giving us," he said. The chain also wants to convert all of its suppliers to the NEX system that provides detailed invoices.

United will also select a software provider that will enable the chain to extend the program to perishables.

Barbara's Caveats

Not many people can be described as having written the book on computer-based ordering, but Barbara Anderson can. Anderson, president of bvac inc, San Carlos, Calif., published a book on computer-assisted ordering a decade ago, and currently consults, writes and offers seminars on the topic. Among her clients is SAF AG, supplier of the forecasting engine used by United Supermarkets (see main story).

Anderson holds many strong beliefs about how computer-based store replenishment should be approached, and offered the following caveats to would-be practitioners:

Beware of perfectionism. "There have been some spectacular failures of systems that work in pilot, but fail in rollout," observed Anderson. That, she said, can be attributed to grocers seeking "a perfect, accurate inventory" rather than concentrating on what can be maintained. "The inventory to support CAO should be good," she said, "but does it have to be accurate to the last can of cream-of-mushroom soup? No!"

Avoid major maintenance. Large retailers who group stores into clusters for forecasting purposes, especially seasonal forecasts, should realize that they may be "signing up to maintain both profiles and clusters for hundreds of stores," she said. "That is a very big deal, and they need to ask the provider how is that going to work -- not for a month-long pilot and a few stores, but for hundreds of stores over years."

Don't lose the uniqueness of individual stores. "Every store may start 'spring' differently because of different weather patterns," she said. "Some stores may have a different ethnic mix, and so a different response to religious holidays, etc."

More frequent forecasting may not be better. "The retailer must look category by category at the best forecast time horizon and not assume that anything you do more often is better," she said.

Don't forget store operations. "I cannot tell you how many CAO meetings I have gone to run strictly by headquarters staff who have not even considered inviting store ops to the meeting," she said.

Keep an open mind. In the RFP (request for proposal) sent to vendors, retailers should not "assume the math technique" about an application -- that is, assume that it works in a particular way. "There are so many exciting and new and interesting methods of managing inventory, promotions and seasons, right now," she said. "This is a great time in the industry for fascinating areas like pattern recognition, artificial intelligence and neural nets."