MTN is the foremost cellular network in Africa. Its network coverage in South Africa is one of the most extensive in the world, embracing almost 880,000 square kilometers and providing telecommunications access to 92 percent of the country's population. MTN also has operations in many other African countries, including Uganda, Swaziland, Rwanda, Cameroon and recently Nigeria.

The company's total revenue grew 39 percent in financial year 2000 to reach R8.34 billion ($1.132 billion) on the back of strong growth in the number of MTN subscribers, both in South Africa and the rest of the African continent. In large part this growth can be attributed to MTN's record of innovation and leadership within the cellular industry, and a focus on maximizing its existing investments.

"The global telecoms community has been seduced by the allure of the technological revolution of 'wireless broadband' and they have all mortgaged themselves to the hilt in pursuit of market dominance," says Yvonne Muthien, MTN group executive corporate relations. "MTN's focus is on leveraging what is in place. We balance major investments in technologies with better delivery of customer services."

One of the greatest challenges for a cellular phone company trying to run a tight ship like MTN is to monitor and manage the procurement of handsets, which are ordered from preferred manufacturers. Too few, and the danger is that you disappoint and lose potential customers. Too many, and you tie up precious resources and you lose customers anyway because they don't want to buy old products. "It's a potential lose-lose situation," says Margie van Straten, statistical specialist at MTN.

The problem is that fluctuations in demand are difficult to predict. In 1994 a subscriber base forecast predicted that there would be 341,000 subscribers in 2004. In fact, the current active subscriber base has already passed the two million mark. Expecting this growth to continue, in 2000 MTN found that historically, sales staff were responsible for predicting the level of demand for handsets. The Procurement department would then add or subtract a margin to ensure that "enough" handsets were bought in. "We got different perspectives from different people, and therefore different results," says van Straten.

Customer MTN South Africa
"We therefore tried moving to another approach which involved manually extracting historical data from a variety of operational systems: data on stocks in stores, in warehouses, orders and current sales per channel." But this was a very time-consuming manual process. It would take more than ten hours to compile the report, by which time the information was outdated.

"Also, when we compared the report against the number of new handsets on the network, the tables did not tally. There was a lot of duplication in the data, which took a lot of manual intervention to unravel, and even then we could not be sure that the information was accurate."

"It was apparent that we'd need to update this report on a daily basis if it was going to provide reliable information, and this would have placed huge demands on our time," says van Straten. "So it was clear that we had to find an alternative solution, and fast."

That solution had to be what van Straten calls a SPOT (Single Point of Truth) solution, providing a clear and unambiguous picture of what is actually happening. It had to be capable of supporting a more structured approach and it had to be updated daily in a production environment, placing minimum demands on the hardware used for transaction processing. "In other words, it had to be a local data warehouse," says van Straten.

MTN chose SAS because the solution "had to provide easy access to massive volumes of data from a variety of engines and sources, a key SAS strength," says van Straten. "In addition we were very comfortable with the product, we already had extensive in-house SAS experience and we were used to excellent support from SAS South Africa," she adds. "Our most critical business requirement was to understand what quantity of which type of handset must be delivered when and to whom. We wanted to keep the project 'bite-sized' and manageable to ensure that the critical requirement was met, a difficult challenge in a large organization with a wide variety of needs."

Following a data modeling project undertaken in association with SAS Quality Partner Pulse Management Services, MTN populated the local data warehouse and developed the analytical tools to forecast handset sales. In fact, MTN has to understand and predict trends in two very different marketplaces: pre-paid handsets and contract handsets.

With pre-paid handsets, the customer buys recharge vouchers allowing a certain amount of airtime, which can be used when the customer needs it. "Sales fluctuations reflect a fairly typical retail model with seasonal and trend components, though there was not enough historic data to reveal cycles," says van Straten. "We trialled a SAS forecasting tool that had originally been developed for the US Census Bureau, and got excellent results. We have achieved forecasts that are always within about five to seven percent of what actually happened, compared with 20 percent or more using the old sales forecast.

"In contract sales, where the customer pays at the end of a billing month, it is far more difficult to predict trends. Sales are up and down all the time, and there is no discernible pattern, seasonality or cycle," says van Straten. "Instead, fluctuations reflect what is going on in the market. As prices fall and promotional activity kicks in, sales pick up. But at any time, competitor activity can cause market churn or the economic environment might cause people to temporarily leave the market altogether."

The forecasting model based on the Census Bureau's tool simply didn't fit this situation. MTN therefore used a Moving Average model. "This has given significant improvements on past results, and though it has understated sales this has been by a consistent amount over the first three months," says van Straten. "In contract sales we additionally need to look at the data in the light of external factors."

One of the hurdles that van Straten's team had to overcome was the skepticism of the sales team. "But when they saw the quality of our forecasts they were astounded. They simply could not believe that SAS could manipulate the astronomical amounts of data involved, and they wanted even more information, even faster. Now for the first time they can get reports that detail what customers have bought, activations and stock levels.

To start with, the salespeople wanted to analyze their forecasts in Excel, using their own personal layout. "It was far from ideal since nothing was standardized, but with SAS software you could import the data and standardize it. But now we will be doing away with Excel altogether and displaying the data graphically on the web, with trends and seasonality identified, and enabling flexible forecast analysis."

A SAS/IntrNet application and WIGS (Wireless Internet Gateway Service) will now enable field sales staff to interact with the SAS Data Warehouse any place, anywhere. "What better way for a cellular telecommunications company to deliver information to its sales force than by phone?" says van Straten.

"As a result of this project we have a better idea of which handset models sell better, where and why. Overall we know our organization, our suppliers and our customers better through having this information. We have detected very definite shifts in buying patterns among our customers. For example, we previously had virtually no high-end handset sales in the pre-paid market, but these customers have become more fashion and quality conscious, a trend that has been revealed through the SAS application. Such information is useful to us and it is also useful to our suppliers, helping to improve the benefits we get from the relationship," concludes van Straten.