In 2008, Keith Nash was a director in the supply-chain practice of Deloitte & Touche LLP. In the trough of the Great Recession, he was brought in to help Lennox International expand the North American distribution network of its Residential Heating and Cooling business unit. It might have looked like a straightforward consulting engagement from the start, worries about the economy aside, but Nash couldn't have predicted how things would turn out.

For one thing, he ended up working for Lennox. Today, Nash is vice president of supply chain logistics in the company's residential division. But the project itself ended up entailing much more than he expected.

Perhaps any undertaking of ambitious scope will yield surprises along the way. And this one promised to be huge. Lennox didn't just want to broaden its distribution footprint; it wanted to reconfigure the whole network, adding a tier of regional distribution centers and up to 300 "store-like" locations for ensuring product availability.

The company's wish list included an increase in the U.S. and Canadian distribution network from 80 to 130 retail sites, along with a boost in SKU locations for goods and spare-parts from 450,000 to more than 700,000. All of this, of course, was to take place simultaneously with optimized inventories, balanced allocations and a marked improvement in service levels.

Lennox is a leading manufacturer and distributor of climate-control systems for heating, ventilation, air conditioning (HVAC) and refrigeration markets. The Residential Heating and Cooling unit accounts for roughly 43 percent of the parent company's $5bn in U.S. sales.

One might assume that a maker of such items would be subject to a fairly predictable demand curve: air conditioners in the summer, heaters in the winter. And that’s true to an extent. But Lennox’s business is complicated by a certain amount of intermittent demand - products that sell in low volumes, at a pace that's difficult to predict.

Lennox has a significant investment in inventory, with 98 percent of SKUs generating 62 percent of revenues, in addition to those slow movers. And product can't afford to sit for long; in a given year, nearly half of a line of finished goods might be replaced by new models.

The customer-service bar had already been set at a high level. Around three-quarters of orders are delivered to buyers the next day, and 20 percent require same-day pickup. In addition, Lennox guarantees that finished products will continue to be serviceable for at least 15 years.

Need for Accuracy

All of which meant that Lennox needed a system for forecasting and fulfillment that could calculate accurate stocking levels at SKU locations, while accounting for demand variability, inventory expense, service levels and other considerations. With those requirements in mind, Deloitte recommended a hub-and-spoke network that included eight regional distribution centers.

To run the operation with efficiency and accuracy, Lennox needed an application that understood the company’s high degree of demand variability. After all, it was dealing with more than 110,000 part and location combinations throughout the supply chain. A reliable forecast rested on its ability to address “the wide variety of those demand behaviors,” says Pat Smith, general manager of ToolsGroup, the software vendor that Lennox chose for the task.

Lennox would end up implementing ToolsGroup's SO99+ application for supply-chain planning and demand modeling. But it was more than a matter of plug and play. First, says Nash, the company had to clean up its data.

Lennox was a longtime user of the SAP APO (Advanced Planning and Optimization) tool, which would serve as the platform for SO99+. Nash says the company hadn't bothered to keep dates correct in the system.

So 10 years of past-due sales orders needed to be cleaned up before it could begin addressing master-data challenges.

For products with intermittent demand, it was especially important to get the numbers right. Applying a normal standard deviation to the forecast could yield a figure for safety stock that was around 25 percent too high. Lennox might find itself saddled with months of unneeded inventory.

The ToolsGroup application employs a modified version of the Poisson distribution, a method of calculation that expresses the probability of a certain number of events affecting the forecast. "It’s been proven to be the best for intermittent-demand modeling," says Nash.

Recognizing Patterns

With the help of SO99+, Lennox is now able to model demand patterns related both to seasonality and the variable consumption that was so difficult to forecast previously. It can determine the optimal mix of inventory and service levels down to the store level, accounting for such factors as delivery frequency, order size and distance to be traveled.

The company can set global service policies by group or category, with SO99+ automatically calculating the right service levels for each individual SKU location. The system sets up reorder points within SAP APO, and specifies minimum order quantities.

Lennox initiates pre-deployment planning with the aid of demand simulations that permit it to see how various scenarios might play out prior to implementation. And it’s able to pre-build inventories in anticipation of seasonal peaks.

The system allows weekly reordering and updating of safety stock and order quantities, weekly requirements forecasts for purchasing and manufacturing, weekly demand sensing and exception management, monthly inventory planning for the sales and operations planning (S&OP) process, quarterly service-policy adjustments tied to sales, and the creation of build plans for new stores and inventory locations.

Nash says Lennox has seen an increase in forecast accuracy for parts and supplies from around 55 percent to just over 70 percent by region. Each percentage point of improvement correlates to a percent of safety-stock reduction, so a 10-percent rise means a $1m cut in stock. In all, he says, the company has taken nearly $10m out of safety stock since implementing the software and reconfiguring the distribution network.

At the same time, the company’s fill rate has gone from between 70 and 80 percent in 2010, when Nash joined the company, to more than 93 percent today. By the end of this year, he expects the figure to hit 98 percent.

Lennox has seen a corresponding rise in service levels, with stockouts dropping from 9 percent to 4 percent, even as the company has grown its store network by 30 percent over two years.

Nash says Lennox is continuing to explore ways to use the system to boost service and take more working capital out of the mix. “Some of my team is still learning how powerful it is,” he says. “I’ve known it for a while.”

Resource Links:
Lennox International
ToolsGroup