Associated Grocers of Florida believes in learning from experience.
As a cooperative wholesaler, AG of Florida, Pompano Beach, collects reams of data on its past sales to member stores — data that can be leveraged to help its buyers make prudent buying decisions.
And as a company operating in Florida, AG of Florida lived through two horrific hurricane seasons in 2004 and 2005 — an experience that has reinforced the need to help its stores be well prepared for future hurricanes.
All well and good, but how does the company make best use of all this raw data and experience to plan and act intelligently in the future?
That’s where Mitch Herman and Atlanta-based Logility come in. Herman, a former grocery buyer for AG of Florida, is now its Logility Forecaster. He oversees Logility’s demand-driven Voyager forecasting engine that analyzes two years of sales history for each of 50,000 items being shipped out of the wholesaler’s two warehouses, in Pompano Beach and Ocala.
Based on that analysis, the system advises AG of Florida’s 10 buyers on how much they should order to meet demand at their retailers’ stores. The wholesaler’s buying process was previously a largely manual process.
AG of Florida has been using the Voyager system, including its demand-planning, inventory-planning and replenishment-planning applications, for about a year. What Herman soon realized after the initial implementation was that, given AG of Florida’s location in the hurricane belt, the technology could also help the company plan for the extreme shock to the supply chain that can be delivered by Mother Nature.
“We had two record hurricane seasons in a row and we had the Logility tool,” he told SN. “So I tried to find ways to help our buyers by figuring out what they would need, and I started doing an analysis. I ended up with a list of what made sense to deal with [a hurricane].”
AG of Florida, which serves stores in Florida as well as 45 countries, is not the only food distributor to develop a hurricane plan. Another grocery cooperative, Associated Grocers, Baton Rouge, La., has a hurricane preparation and response strategy that it used last year when the Gulf Coast was devastated by Hurricane Katrina (see “First Responders,” SN, Oct. 24, 2005). Wal-Mart Stores, Bentonville, Ark., is well-known for its Emergency Operations Center, which operates 24/7 to monitor and respond to events that disrupt its business.
Other retailers are beginning to use forecasting technology from vendors such as Teradata and SAP to generate orders for stores and warehouses. But by tapping the predictive power of the Logility system, AG of Florida is one of the few food distributors — Wal-Mart is another — to use forecasting technology to deal with an unexpected event, such as a natural disaster, and its aftermath. “For competing against companies of our size, we definitely have a leg up,” Herman said.
Herman’s first step was to feed copious amounts of data into the forecasting engine on what actually happens to consumer demand when a hurricane hits a community. The system came back with reports that enabled him to compile lists of what products — and quantities — stores will demand in anticipation of a hurricane’s arrival and what they will need after it has struck. The latter lists differ by whether an area has lost electrical power or lacks clean running water.
“When a hurricane comes everyone gets energized, but working hard is not always working smart,” he said. “By exploring this long before the storm, decisions are made calmly by the best people in your company. It gives you a big jump over running around and figuring out what to do.”
Some products are stocked in greater quantity throughout the hurricane season, including certain bottled water and shelf-stable milk. “Shelf-stable milk is a really good example because it doesn’t take up a lot of space in the warehouse and has a decent shelf life, so we can keep extra on hand,” Herman said.
Although AG of Florida stocks reserves of bottled water throughout the season, “there’s no way we can keep enough water on hand to service all of our customers in a hurricane, especially with power outages,” he noted. But the wholesaler keeps enough water in inventory so that when a hurricane comes, buyers will have sufficient time to bring in the rest.
Once a hurricane appears on the radar, Herman downloads a spreadsheet with his hurricane recommendations into the Logility system. The system applies a feature called “inventory build” that instantly changes the purchasing pattern for select items. “All of sudden, buyers are buying more water, insect repellent, canned meat, SPAM and bleach,” he said. “This allows us to very quickly put out purchase orders. That’s important because [in hurricane situations] often the people who get product are the ones who place orders fastest.” Once the crisis has passed, the inventory build feature is switched off and normal buying resumes.
By basing its decisions on data derived from actual hurricanes, AG of Florida has been able to discover some “curious” and unexpected trends, Herman noted. For example, matches turn out to be a coveted item prior to a hurricane because in a power outage people need them for cooking with charcoal.
Another twist is the demand for SPAM. While it is in great demand leading up to a hurricane, once the power is out, nobody wants it. “It’s only a good panic item,” Herman said.
In addition to driving orders for some products, hurricanes cause retailers to cancel orders for other items, such as dairy and frozens that can’t be sold if the power goes out. However, when power is restored, a mad scramble ensues to become restocked. Herman found that it was extremely helpful to have a track record of what retailers will need to recover in that scenario. “We usually get a huge spike the first one or two weeks after the power’s up and then it’s normal,” he observed. “They want enough quantity but not too much. So balancing that can be challenging.”
AG of Florida’s hurricane planning sometimes calls for switching from rail to truck deliveries. “When a storm comes to Florida, trains start backing up in Georgia” and deliveries may be delayed a month or more, Herman said. Though it may cause freight costs to change, using truck deliveries is “a bit of an insurance policy.”
Response to Ernesto
In late August, AG of Florida was able to put its hurricane strategy to the test in response to Hurricane Ernesto. Fortunately, Ernesto changed to a tropical storm before entering southern Florida and did not cause major damage. Still, AG of Florida, which serves that area through its Pompano warehouse, was in a good position to support its retailers.
“We were able to focus our energy on getting what we needed into the warehouse and not on figuring out what we needed in the first place,” Herman said. “We were able to get hold of brokers, make sure carriers were coming and appointments were set. Those are the things that determine whether we are going to service customers.”
Since each storm is a unique event, Herman does not want it to set a precedent for buying decisions the same time the following year. So the Logility system will “smooth out” buying decisions ramped up for a hurricane, adjusting the company records to appear as though nothing unusual happened. “That way, next year when we look at this year, all they will see is what they should see,” Herman said.
At the same time, the hurricane records are kept in a separate file, helping AG of Florida to further refine its hurricane strategy.
Despite its reliance on technology, AG of Florida also leaves room for human judgment in shaping its plans, allowing buyers to factor in new trends or products. “I may have bought eight trucks of bleach because a storm was coming through, and I just didn’t sell it,” he noted. “That’s because people don’t use it like they used to [for sanitation]. They’re buying bottled water instead.”
“I don’t think we’ll ever get to the point in the grocery business where a computer can spit something out and no human judgment is applied,” Herman added.
Of course, sometimes a storm shifts course and negates the orders that have been made. “We may have to cancel some POs, but that’s the price you pay,” Herman said. In any event, he has found that the program is highly adaptive, allowing AG of Florida to “expand on our successes and learn from our mistakes.”
Herman said that the biggest challenge in implementing the forecasting system was getting the trust of the buyers. “You’ve got people doing things a certain way for 30 years and it’s hard to accept change,” he said. “You’ve got to prove it to them.”
One year into the implementation, “everyone uses it as a tool,” he said. “I’d say it’s proven itself.”