Tech Check: Getting Demand Planning Right

Demand planning systems go beyond what simple forecasting tools offer. "Forecasting tends to be single-number. With demand planning, we actually have different components of what that number is. Salespeople may be looking at a particular region by revenue, while operations folks are looking at it by quantity," says Gartner analyst Karen Peterson.

The systems produce baseline forecasts from shipment or order history data, so reliable data is a critical requirement. "In 50% to 70% of implementations, [users pull] data from disparate legacy systems," and that data is often inaccurate. Cleaning it is the biggest hurdle to implementation, says i2's product marketing manager, Karen Laucka.

Once a baseline forecast is established, other data can be layered onto the model. Leiner Health feeds data about historical shipments into its i2 system and then adds order history information. In that way, back orders that never shipped are accounted for in the forecast, says Rajesh Varma, director of demand planning. Adding customer point-of-sale (POS) data improves accuracy, but not all customers will provide it, and integrating it isn't easy. "If we could get POS information, it would reduce inventory lead time by as much as five days," Varma says.

"The problem is [customer data] comes in all different formats, and building those interfaces is a daunting task," says Noman Waheed, a senior manager at consulting firm Accenture Ltd. in New York. Most suppliers today concentrate on receiving POS data from a few large customers.


Gross sales:


Reduction in inventory after three years:


To this mix, suppliers may also add data for events such as trade promotions, inserts, seasonality and more. "All of those have an impact," says Waheed, but "most organizations have trouble getting historical sales data, never mind this data." Finally, suppliers can add in nonstatistical data, such as sales and marketing forecasts, or market data from research companies.

At this point, "I have four or five data streams and can find variances between forecasts," says Waheed. Automated tools can spot variances, but management still makes the final call in a "consensus process where all the representatives sit down" and discuss specific customers and specific sales periods, he explains."This is cumbersome, but it's the most effective method we have come across," Waheed says.

Initially, companies should focus on specific goals such as reducing inventory and improving customer service, says Phil Robers, global director at Paris-based consulting firm Cap Gemini Ernst & Young. But they should then go beyond that. "If you understand demand patterns, you should [use] things like promotions to influence demand in your favor."


Drop in back orders in the first year after implementing demand planning software:


Reduction in peak inventory:


The next step is supply chain collaboration, where retailers and suppliers share forecasts, promotion plans and other data to determine the final forecast.

"True collaboration means we look at the same data, manipulate the data and come to the same conclusion—all online," typically by giving customers Web access to forecast data, says Mike Maguire, SAP AG's vice president of supply chain field operations. Though many suppliers collaborate with a few key customers on forecasts, few do it online due to customers' reluctance to share data and the complexity of maintaining such relationships.

Greg Harris, Bandag's manager for production planning and logistics, says that after three years, he's still optimizing the use of internal data. He welcomes customer data, but he'd be cautious about letting customers manipulate forecasts. "You're relying on the customer to have the same expertise that you do, and that's not the case," he says.

Ultimately, says Peterson, "we'll see application-to-application integration" between suppliers and customers using technologies such as XML, and collaboration will be part of that process. But for now, integration is a big headache. "If two trading partners want to integrate information differently—integrating products into different geographies, for example—today, it's done on a customized basis," she says. "Configuration rather than coding is where we want to evolve. Today, it takes a lot of pain and effort to make these work."

Special Report

Missing Links

Stories in this report:



Copyright © 2001 IDG Communications, Inc.

Bing’s AI chatbot came to work for me. I had to fire it.
Shop Tech Products at Amazon