TECHNOLOGY: Just enough
While "too much is never enough" might have worked for David Bowie when he promoted MTV back in the 1980s, when it comes to back-room inventory, too much is simply too much.
Inventory glut was a big problem for Market Basket, a 38-store grocery chain based in Nederland, Texas, with units in southeast Texas and southwest Louisiana. The grocer was in serious need of a better system for counting and, more important, ordering products.
"We kept seeing our inventory increase," says Skylar Thompson, Market Basket's president. "Every time we would change a manager or change a schedule, we would have problems with the inventory increasing. There's a cost of holding that inventory, and we could use that money for something else."
Too much inventory meant too many damaged products, high levels of shrink, and inordinate amounts of labor devoted to maintaining all that stuff. Too much was definitely too much for Market Basket.
"To get rid of some of the extra inventory, we would have to put it on discount," recalls Thompson. "Also, we had to reclaim quite a bit at a reduced amount. Some of the stuff that went out of date, like dairy, we weren't getting full credit for."
Sometimes, not enough
On the other hand, too little wasn't good, either. "For the high-velocity items, there were out-of-stock problems, as well," says Thompson. "It was the worst of each situation."
In response, Thompson took the offensive, with a computer-assisted ordering system from Plano, Texas-based Retalix USA. The system, which manages approximately 12,000 to 15,000 SKUs, depending on the size of the store, helped Market Basket reduce store inventory by 10 percent, out-of-stocks by 2.5 percent, and shrink by 25 percent.
"We have maintained that 10 percent reduction in inventory for all of the items managed by the system," says Thompson.
The system, DemandAnalytX (DAX) is a Web-based replenishment tool for retailers, manufacturers, and wholesale distributors that uses point-of-sale data as its main driver for demand forecasting. It relies on complex algorithms to predict demand based on product movement history.
"The store manager inputs the minimum amount that they want to see on the shelf at any time," explains Michelle Flocke, scanning supervisor for Market Basket. "The system then displays the forecasted order that the software runs for them, showing the history on this item, or the history on that item when it's on promotion. It forecasts how much you will sell until your next delivery."
The store manager evaluates the forecast and has 12 hours to decide what adjustments are needed for the order. Otherwise the forecasted order is sent as a finalized order to the company's supplier.
Getting to this point isn't easy. "The system needs about three months of daily data to get a good trend," says Thompson. "We gave them a year's worth. The software adjusts for seasonality; it starts seeing weekly and monthly trends as you go in and out of summer and winter on certain products. It adjusts quite well, and quickly. The more you use it and the more data is entered into the system, the more it optimizes."
Since the system is Web-based, Market Basket runs the application from a server housed at Market Basket's main supplier, Grocery Supply in Houston, which has a data center that's staffed 24 hours a day. Every night, scan data from the stores is transmitted to the server, which recalculates the data and then produces orders that are sent back to the stores. Once approved, the orders are shipped based on the supplier's delivery cycles.
Strict discipline
Since the DAX system makes its predictions based on the POS data and counts of inventory on hand, it's essential that an accurate physical inventory is taken the evening before going live with a particular category. Otherwise it's a true "garbage in, garbage out" scenario.
Because of this, when an error occurs, it's usually a human one, such as a bad count or a mis-pick on the supplier's end, which means it can easily be corrected by fixing procedures in handling inventory.
"Some of the problems we have run into occur when there are changes to the orders at the store," says Thompson. "The usual process of reviewing the order is to look at the holes in the shelf and check the on-hand inventory. In some cases the inventory isn't accurate, because we got a mis-ship from our supplier, or at some point in time we made a bad inventory count and failed to count something properly."
Once the system is rolling and all counts are correct, not many changes are needed to the system, other than those for seasonal adjustments. "In dry grocery, I would say an average of 5 percent of the items per order are adjusted, most of it because of promotions," says Flocke. "The store managers like to display promotional items at the front of the store at the beginning of the ad, and have it on hand for the duration of the promotion."
Since people determine the timing and duration of promotions, there's no way for the system to predict the demand, so the store manager will adjust the order manually at the beginning of the promotion. When the promotion is finished, the manager must adjust the system once more -- otherwise it will create an order based on the promotional sales of the item, which will then be higher than what's needed.
Like promotions, other nonroutine events that may skew sales figures can be earmarked so they aren't applied to regular forecasts. "We're in a hurricane zone," says Thompson. "In the case of a hurricane, we can actually go in and earmark some particular week and what was unusual about it, so it won't generate an order based on those figures the same period the next year. Within a couple of orders, the system corrects itself based on the new item movement."
As long as the store manager is disciplined in doing inventory counts, the system can be trusted to do the right thing, according to Flocke. But turning over control to the system is not easy for everyone, especially for those who began their careers before the advent of computers.
"We've had a few stores that thought they were smarter than the software, and they didn't trust it like they should and didn't do the inventory counts exactly right," says Thompson. "But we knew who those guys were and got them back on the program. Some people are just not as disciplined as others."
Spotting noncompliance is generally an easy task, according to Flocke. Typical users of the DAX system change only 5 percent of each order generated by the system. Higher percentages of adjustments -- outside of promotions -- are usually indicative of a manager making up for mistakes or a lack of discipline in inventory counts.
Conversely, if a manager is taking twice as long as the average user to review the orders in the store, it's a pretty good indication that he or she is trying to micromanage it instead of just looking at the exceptions.
Labor tradeoff
This might sound like a lot of labor—and it is. But it pays off in spades, says Thompson. "I won't kid you about the labor," he says. "I guess there's a tradeoff in labor. If you take the total time it takes to order with computer-assisted ordering, doing all of the cycle counts, all of the maintenance, it's about the same amount of labor as before, when we were doing the manual ordering system and entering it into a Telxon. But the system wasn't deployed for a savings of labor. We needed it for more accurate orders."
Still, labor is saved in other areas, such as handling excess inventory. "If we factor in that labor -- having to reorganize the back room from time to time because of having excess inventory -- then yes, there are some labor savings," says Thompson.
Most store managers have become comfortable with the system, adds Thompson. "Eighty percent of our stores have bought into it and are following the program, and we're seeing the results in the form of improved store conditions and reduced store inventory," he says. "There are fewer out-of-stocks, less spoilage, and better dates on perishable items like milk and dairy, which our customers like.
"And the store managers love the back-room conditions," Thompson continues. "There's one-third to one-half of the inventory in the back room, compared to what was in there before we implemented the system. They don't have to hunt for things as hard; they know that all that are in the back are high-velocity items and deal items, and not back stock that they've been trying to get out from under for several months."
As of July 2004, all products from Market Basket's main supplier—roughly 50 percent of the products ordered—were online with the system.
Thompson is now planning to get produce and more of his meat items into the system, as well as his nonfood DSD items, because in addition to reducing out-of-stocks and shrink, the system also protects against greedy salespeople working on commission. "Any time a vendor has a salesperson on commission, the tendency is to sell you more inventory than you need," says Thompson. "The system is a good check against that."
Sorry, Bowie: It looks like Market Basket isn't the place for you anymore.
Inventory glut was a big problem for Market Basket, a 38-store grocery chain based in Nederland, Texas, with units in southeast Texas and southwest Louisiana. The grocer was in serious need of a better system for counting and, more important, ordering products.
"We kept seeing our inventory increase," says Skylar Thompson, Market Basket's president. "Every time we would change a manager or change a schedule, we would have problems with the inventory increasing. There's a cost of holding that inventory, and we could use that money for something else."
Too much inventory meant too many damaged products, high levels of shrink, and inordinate amounts of labor devoted to maintaining all that stuff. Too much was definitely too much for Market Basket.
"To get rid of some of the extra inventory, we would have to put it on discount," recalls Thompson. "Also, we had to reclaim quite a bit at a reduced amount. Some of the stuff that went out of date, like dairy, we weren't getting full credit for."
Sometimes, not enough
On the other hand, too little wasn't good, either. "For the high-velocity items, there were out-of-stock problems, as well," says Thompson. "It was the worst of each situation."
In response, Thompson took the offensive, with a computer-assisted ordering system from Plano, Texas-based Retalix USA. The system, which manages approximately 12,000 to 15,000 SKUs, depending on the size of the store, helped Market Basket reduce store inventory by 10 percent, out-of-stocks by 2.5 percent, and shrink by 25 percent.
"We have maintained that 10 percent reduction in inventory for all of the items managed by the system," says Thompson.
The system, DemandAnalytX (DAX) is a Web-based replenishment tool for retailers, manufacturers, and wholesale distributors that uses point-of-sale data as its main driver for demand forecasting. It relies on complex algorithms to predict demand based on product movement history.
"The store manager inputs the minimum amount that they want to see on the shelf at any time," explains Michelle Flocke, scanning supervisor for Market Basket. "The system then displays the forecasted order that the software runs for them, showing the history on this item, or the history on that item when it's on promotion. It forecasts how much you will sell until your next delivery."
The store manager evaluates the forecast and has 12 hours to decide what adjustments are needed for the order. Otherwise the forecasted order is sent as a finalized order to the company's supplier.
Getting to this point isn't easy. "The system needs about three months of daily data to get a good trend," says Thompson. "We gave them a year's worth. The software adjusts for seasonality; it starts seeing weekly and monthly trends as you go in and out of summer and winter on certain products. It adjusts quite well, and quickly. The more you use it and the more data is entered into the system, the more it optimizes."
Since the system is Web-based, Market Basket runs the application from a server housed at Market Basket's main supplier, Grocery Supply in Houston, which has a data center that's staffed 24 hours a day. Every night, scan data from the stores is transmitted to the server, which recalculates the data and then produces orders that are sent back to the stores. Once approved, the orders are shipped based on the supplier's delivery cycles.
Strict discipline
Since the DAX system makes its predictions based on the POS data and counts of inventory on hand, it's essential that an accurate physical inventory is taken the evening before going live with a particular category. Otherwise it's a true "garbage in, garbage out" scenario.
Because of this, when an error occurs, it's usually a human one, such as a bad count or a mis-pick on the supplier's end, which means it can easily be corrected by fixing procedures in handling inventory.
"Some of the problems we have run into occur when there are changes to the orders at the store," says Thompson. "The usual process of reviewing the order is to look at the holes in the shelf and check the on-hand inventory. In some cases the inventory isn't accurate, because we got a mis-ship from our supplier, or at some point in time we made a bad inventory count and failed to count something properly."
Once the system is rolling and all counts are correct, not many changes are needed to the system, other than those for seasonal adjustments. "In dry grocery, I would say an average of 5 percent of the items per order are adjusted, most of it because of promotions," says Flocke. "The store managers like to display promotional items at the front of the store at the beginning of the ad, and have it on hand for the duration of the promotion."
Since people determine the timing and duration of promotions, there's no way for the system to predict the demand, so the store manager will adjust the order manually at the beginning of the promotion. When the promotion is finished, the manager must adjust the system once more -- otherwise it will create an order based on the promotional sales of the item, which will then be higher than what's needed.
Like promotions, other nonroutine events that may skew sales figures can be earmarked so they aren't applied to regular forecasts. "We're in a hurricane zone," says Thompson. "In the case of a hurricane, we can actually go in and earmark some particular week and what was unusual about it, so it won't generate an order based on those figures the same period the next year. Within a couple of orders, the system corrects itself based on the new item movement."
As long as the store manager is disciplined in doing inventory counts, the system can be trusted to do the right thing, according to Flocke. But turning over control to the system is not easy for everyone, especially for those who began their careers before the advent of computers.
"We've had a few stores that thought they were smarter than the software, and they didn't trust it like they should and didn't do the inventory counts exactly right," says Thompson. "But we knew who those guys were and got them back on the program. Some people are just not as disciplined as others."
Spotting noncompliance is generally an easy task, according to Flocke. Typical users of the DAX system change only 5 percent of each order generated by the system. Higher percentages of adjustments -- outside of promotions -- are usually indicative of a manager making up for mistakes or a lack of discipline in inventory counts.
Conversely, if a manager is taking twice as long as the average user to review the orders in the store, it's a pretty good indication that he or she is trying to micromanage it instead of just looking at the exceptions.
Labor tradeoff
This might sound like a lot of labor—and it is. But it pays off in spades, says Thompson. "I won't kid you about the labor," he says. "I guess there's a tradeoff in labor. If you take the total time it takes to order with computer-assisted ordering, doing all of the cycle counts, all of the maintenance, it's about the same amount of labor as before, when we were doing the manual ordering system and entering it into a Telxon. But the system wasn't deployed for a savings of labor. We needed it for more accurate orders."
Still, labor is saved in other areas, such as handling excess inventory. "If we factor in that labor -- having to reorganize the back room from time to time because of having excess inventory -- then yes, there are some labor savings," says Thompson.
Most store managers have become comfortable with the system, adds Thompson. "Eighty percent of our stores have bought into it and are following the program, and we're seeing the results in the form of improved store conditions and reduced store inventory," he says. "There are fewer out-of-stocks, less spoilage, and better dates on perishable items like milk and dairy, which our customers like.
"And the store managers love the back-room conditions," Thompson continues. "There's one-third to one-half of the inventory in the back room, compared to what was in there before we implemented the system. They don't have to hunt for things as hard; they know that all that are in the back are high-velocity items and deal items, and not back stock that they've been trying to get out from under for several months."
As of July 2004, all products from Market Basket's main supplier—roughly 50 percent of the products ordered—were online with the system.
Thompson is now planning to get produce and more of his meat items into the system, as well as his nonfood DSD items, because in addition to reducing out-of-stocks and shrink, the system also protects against greedy salespeople working on commission. "Any time a vendor has a salesperson on commission, the tendency is to sell you more inventory than you need," says Thompson. "The system is a good check against that."
Sorry, Bowie: It looks like Market Basket isn't the place for you anymore.