Profiling
There are many reasons to segment a customer file, ranging from: differential treatments based on corporate policy, identifying affinity groups, and defining custom communication groups for CRM.
Segmentation
Segmentation utilizes customer behavior data and appended demographic attributes. Segment definitions are defined via multivariate analysis techniques: cluster, factor, or discriminant analysis. Once defined segments represent a homogeneous customer population. Models developed for homogeneous segments have less variation due to the differences between customers and are therefore more predictive.
Response Modeling
Thorough knowledge of customer data is the most important element in the creation of powerful models. This in depth understanding of the relationships in your data can be leveraged in predictive models. Our unique research techniques require seeing your business through your customer data. We employ sound statistical methods to ensure all models will discriminate in roll-out applications. In addition, we develop models that will remain predictive over time.
Acquisition Response Models
Response models require information on responders and non-responders to your direct mail or telephone offer in addition to a "snap shot" of their attributes at the point-in-time the offer was extended. These models provide a means of discriminating between prospects that will and will not accept your offer. The benefit is better targeting of your best prospects. The drawback is a longer lead time to read results and apply in a roll-out; however, a response model will always outperform a clone.
Clone or Look-a-Like Models
These models work on large compiled files. By supplying your ideal customers, a clone model defines their attributes and identifies replicates in the compiled file. A key benefit of a clone model is the speed with which you can roll-out and gauge its success. One drawback is the possibility of not identifying new markets for your product since you are basing the analysis on current customers. In addition, care must be taken when defining the ideal customers for cloning. If proper names are not selected, the strength of the model may weaken in a roll-out.
Performance Models
Performance models are utilized to predict a customer’s payment behavior. Performance models may be employed to target payers or to identify non-payers depending on the need. For direct marketers utilizing a soft offer, performance models can keep bad debt to a minimum. Telematch’s database of demographic attributes can be used alone in acquisition efforts, or in concert with customer behavior data in cross sell promotions to target offers and limit bad debt.
Assessment of compiled list and enhancement data
Knowing which enhancement data to bring on file is a complex proposition. We have the experience to evaluate outside demographic/psychographic data based on its ROI. We use proven techniques to assess how such data can improve your customer segmentation or models by running analyses with and without such data. The difference in these two analyses, projected over the life of the contract, forms the basis of the ROI calculation. To evaluate the true strength of such data, we employ various unique multivariate statistical techniques. |