VOLUME 1, ISSUE 2
IS SOMETHING MISSIONG FROM YOUR CRM DATABASE?
Click here for PDF download.

Many companies are frustrated that the massive investments they have made in customer databases (CRM) have never fully achieved their promise. CRM databases typically record transaction histories and some simple demographics - dry facts that often don’t easily translate into marketing actions. Data Fusion can help bring the database to life, filling gaps with rich information about customers’ attitudes and needs, allowing companies to precisely target their direct marketing efforts.

For example, a financial services firm may want to understand each customer’s attitude toward risk, so they can better target investment offers. A pharmaceutical company may want to know each doctor’s attitudes towards potential side effects, so they can decide which products to detail. An online retailer may want to know what media their customers watch, so they can more precisely target ad campaigns. An airline may want to identify which attitudinal segments their frequent flyers fall into, so that they can tailor perks to each flyer’s needs. But these types of information are impossible to get by simply tracking transaction histories. And in most cases it isn’t feasible to go out and survey all customers to get this attitudinal information.

The solution? Data Fusion. Data Fusion is an advanced analytical method designed to help fill in the gaps in a customer database. The core idea is that by using advanced statistical models and data collected from a sample of consumers, missing information for each customer in the database can be predicted reasonably well. So, for example, we can field a survey with a sample of financial services consumers asking them how they feel about financial risk, then link these responses back to the CRM database to make a statistical prediction about how each customer in the database feels towards risk. This process is sometimes referred to as “scoring” the data base. Once the predictions are merged into the database, they can then be used to tailor direct marketing campaigns.

While Data Fusion can pay huge dividends, it can be tricky to do well. The Modellers, however, have been pioneering work in Data Fusion for over a decade, and can help ensure it’s done right. For example, traditional Data Fusion methods such as Discriminant Function Analysis (DFA) often give predictions that are no better than random guesses. We have found that the key to a truly accurate and useful solution is not only good “linking variables” (variables that overlap between the database and the survey), but a sophisticated methodology, such as expectation maximization (EM) or a Bayesian missing data approach, that uses information from the database and the survey to settle on a stable solution.

Getting Data Fusion right requires an experienced analyst, one who can help to identify the best set of linking variables and choose the right methodology. Bottom-line:  You can improve your CRM-based marketing efforts using Data Fusion and The Modellers.