|
Thomas Tileston is in the business of making Hollywood blockbusters. He doesn’t do it from in front of or behind the cameras, however. He does it by using business intelligence software to predict the right number of DVDs that need to be manufactured and shipped to retailers around the globe.
As the vice-president of global forecasting at Warner Home Video, there is big money on the line for making sure he gets the number right.
In the first week a DVD goes on sale, 45% of the total sales of that release will take place. By the end of the first month, 75% of the sales will happen. Ship too few DVDs out to retailers and millions could be lost in sales. Ship too many and there are inventory costs to be swallowed. “It’s a short window of opportunity, so there’s a lot on the line in trying to get as close to the mark as possible,” he says.
About five years ago Warner began using software from SAS Institute in combination with its Teradata data warehouse installation to more accurately predict demand. The results were immediate and dramatic, he says. In one stellar case, the company used its business intelligence software to get the numbers right for the Harry Potter release in 2004, resulting in then industry-leading sales of $4.75 billion.
Since that time, however, Warner has been able to greatly increase the range of attributes being used to make its forecasting, and take the technology in new directions.
A big new innovation, says Tileston, is that it has begun using the BI software to help identify errors in the data used to do the forecasting. Data is pulled into the Teradata warehouse on a nightly basis from Warner’s SAP enterprise system. Because the underlying data comes from a wide range of sources, there is often a problem with the data quality – not an uncommon problem in organizations big and small. In Warner’s case, a lot of the data associated with the DVD titles is typed in manually and in some cases the error rate was as high as 30%.
Tileston’s team has been able to use the SAS software to look for errors in the data, and in many cases can fix the data using business logic and return it fixed to the SAP system. If the data cannot be fixed automatically, it is returned to the business unit for further inspection. Using the new system, Tileston says the error rate in the data has fallen to less than 1%.
With more accurate data, Warner can have more confidence in the forecasting being made. And over time, Tileston has been able to increase the number of attributes being analyzed as well as the sources of the data, from the amount earned by each film per theatre, to the star power of the actors. Even though the analysis is a lot more complex, Tileston says the company is still able to produce a forecasting model on some 300 titles each week in about three hours.
“It’s been a complicated piece of work,” he says of the data quality project, “but because the business has seen such a dramatic increase in the quality, they’ve really gotten behind it.”
Only registered users can write comments.
Please login or register. |
