Customer intelligence improves B2B sales efficiency

Customer intelligence solutions can prove highly successful in optimizing marketing activities. Overtoom International, market leader in office and warehouse supplies, used SAS solutions to fine-tune their marketing campaigns. SAS partner Python Predictions encouraged Overtoom to review their selection of targets for every marketing action. The result was proof that predictive analytics lead to an increased turnover and/or a smaller marketing budget.

OVERTOOM PROJECT PROVES WORT H OF SAS-BASED PREDICTIVE MODEL

The efficiency of B2B marketing campaigns is difficult to measure and optimize, in part because of the complexity of the typical B2B sales context. “B2B customers have a complicated purchase behavior,” observes Olivier Serruys, Director Benelux of Overtoom International. “Purchasing decisions are taken by various individuals or even groups of people and we don’t always know the real decision makers. Usually the purchase process is rather long and runs through a variety of channels, ranging from account managers to e-commerce sites. That’s why we need to carefully target our marketing campaigns.”

Go further by adding more intelligence

Over the years, Overtoom International has made significant efforts to improve their marketing campaigns. “Ten years ago we managed to reduce our mailing expenses substantially without any turnover loss,” says Serruys. “We did it by filtering our CRM database using an RFM segmentation technique.” RFM stands for Recency (did the customer purchase recently), Frequency (how frequently do they purchase), and Monetary Value (what is the monetary value of their purchases). This technique segmented the customer database into nine customer categories. The assumption is that marketing campaigns may work well for some of the categories — depending on the type of campaign — but certainly not for all of them.

Serruys notes that Overtoom used this technique to reduce monthly mailings from 200,000 to 95,000. “But we felt that this segmentation was not yet ‘smart’ enough. We wanted to go further by adding more intelligence.”

Qualifying each customer individually

That additional intelligence was found in the SAS CI solutions used by Python Predictions. They proposed a shift away from segmentation towards prediction. Dr Wouter Buckinx of Python Predictions explains the difference: “Segmentation implies putting customers in a limited set of predefined categories. As a result you treat every customer in a given category in the same way. You’re still likely to send promotional offers that they are not interested in, or worse, not send promotional offers they would be interested in. Prediction, on the other hand, qualifies each customer individually, based on a much broader set of variables.” SAS CI solutions enabled Python Predictions to use about 3 million data records from the Overtoom CRM database and additional sources to analyze each customer’s purchase behavior. They used 850 variables to qualify customers, including information about the company and the contact history, purchasing details, sales channels used, complaint history, and seasonality information. Dr Buckinx confirms that this SAS-based method enables a much more precise assessment of customer behavior.

Predicted customer reactions confirmed

Based on this in-depth analysis of past customer behavior, Python Predictions constructed a compact but powerful list of indicators of a customer’s response potential. “It turned out that many important indicators of customer quality were not even present in the previous segmentation scheme,” says Dr Buckinx. “With our current list of indicators we are able to quite accurately predict how each customer will react to a given marketing campaign.” Furthermore Python Predictions calculates the optimal size of the target group for each marketing campaign in order to maximize turnover and minimize marketing expenses.

The predictions have proved to be virtually spot-on rather than illusory or mere wishful thinking. The effect was measured by running the two models side by side in real life. The customer base was cut in two and targets were selected for each marketing campaign based on either the segmentation model or on the prediction model. The results of this test coincided extremely closely with the predictions. “We were confident they would,” says Dr Buckinx. “That’s why we agreed that our remuneration for the project was defined variable, dependant on a proven revenue increase.”

Balance turnover and marketing budget

The SAS-based predictive model boosted sales by 10% with an equal marketing budget. Further analyses enabled a further increase in turnover and at the same time economized on marketing expenses. “We can use the resulting efficiencies in different ways,” says Serruys. “We can maintain our marketing budget for certain campaigns and increase revenues, or we can save on marketing expenses while maintaining our revenue. In either case, we gain on profitability.”

Python Predictions further optimized the system with automated feedback procedures following each marketing campaign. The system monitors data and model quality before and after the action. As a result, Overtoom now has an automated customer scoring system to keep customer quality indicators always up-to-date.

Copyright © SAS Institute Inc. All Rights Reserved.