Advanced Predictive Modeling Made Simple (Sort Of)

(Multichannel Merchant) Predictive modeling has been widely adopted in the direct marketing industry. So much so, it has become the direct marketer’s best friend.

Although complex statistical algorithms are used in developing predictive modeling applications, the overall concept and primary goal — to find customers or prospects who are most likely to exhibit the desired behavior — is simple. In the context of customer acquisition, the goal is to find prospects who will respond to a direct marketing offer and subsequently become a long-term customer.

Direct marketers are maximizing the return on investment for their programs by using model-driven targeting strategies. In one case, a media industry marketer was able to generate annual savings more than $4 million by targeting the right people.

Here are a few tips to get you started:

Use segmented modeling techniques. You can improve the predictive powers of models by using segmented modeling approaches. Quantitative methods such as tree-based algorithms can aid in creating the optimal segments. Often different demographic and economic factors drive response and conversion behaviors for unique prospect segments. Segmentation approaches can differ greatly by industry and by the specific client situation.

For example, in the financial services industry, marketers can typically boost the power of their targeting models by developing separate modeling algorithms for different credit segments. It’s common to find that the response drivers for high-credit quality segments are different from the response drivers for high-risk segments. By the same token, separate models by geography have been proven to generate better results in the auto insurance industry

Make continual improvements. Predictive modeling is never a one-shot process. Direct marketers can improve the predictive performance of models by continually incorporating the results from their most recent direct marketing campaign and response activity.

There are several reasons model updates lead to better marketing results. First, consumer response dynamics change constantly, especially within highly competitive and fast-moving markets, such as credit cards, auto insurance, and retail. Second, richer response data become available as direct marketers launch more campaigns, and more data typically lead to better targeting models. And finally, marketers can over time perform tests to randomly selected audiences, which could create an “unbiased” pool of responders and converters that they can use to develop better models.

Continual modeling improvements can boost results for direct marketers across many industries. In some cases, they can improve response rates in the 15%-25% range over baseline models.

Maximize available data for modeling. Richer data typically translate into better predictive models. One data category that tends to make significant industry- and client-specific contributions to predictive power is derived data. For example, the distance to a store location is a strong driver for companies with retail locations and can be calculated by taking the longitude and latitude data into account.

Additionally, summarized promotion history data, such as the number of prior contacts during the past year, can be a strong contributor to the model’s predictive power. Geo-targeting indexes, such as customer or responder penetrations, can also lead to significant improvements.

Client-specific derived data can be combined with third-party sources to create a rich pool of valuable customer data. There is a wealth of third-party demographic, psychographic, financial, and industry-specific data available.

Direct marketers can use advanced predictive modeling techniques to gain a competitive advantage and boost their marketing ROI. Advanced modeling is a big part of audience selection and can drive greater levels of revenue and profits — especially in competitive environments. Through continued improvements in predictive modeling, you can achieve even aggressive goals and deliver the financial returns expected.

Ozgur Dogan is senior director of database marketing solutions for Lanham, MD-based database marketing company Merkle (www.merkleinc.com).

If you’d like to receive articles on database marketing delivered to your desktop every week, subscribe to “List & Data Strategies,” a free weekly e-newsletter from the editors of “Multichannel Merchant,” a sister publication of CHIEF MARKETER. For details, click here.