|
PhD analyst, Dr. Kate Webster shares 15+ years of experience. Read on for her recommendations, priorities and advice on how leverage your data to meet your business objectives.
I would start by asking myself three sets of questions
1) How can I utilize my house files? What do my overall customers look like compared to a U.S. sample (what is my market penetration)?
2) Who are my house file stars--my best customers and how do I segment the “very good” (e.g. recently purchased, frequently purchase) from the “not-so-good” (e.g. lapsed customer) so I can identify “look-a-like” prospects?
3) Which of those look-a-likes has the highest statistical probability of becoming a best customer for my company?
Be sure to clearly define “pain points”. Is my business concerned about maximizing existing data/housefiles, acquiring new customers, retaining existing customers, creating cross-sell opportunities, or reactivating my lapsed customers? It’s important to define goals at the onset of a modeling project to ensure the success of the model and to meet customer expectations.
Focus on what matters. It’s critical to focus on the quality and breadth of the data. A house file of customer data that is standardized, employs good data hygiene, is highly mailable, and is enhanced with rich demographical and behavioral elements provides analytic power and accuracy to any model.
Find a good analytics partner who offers these qualities.
Houses Multi-Source Data: has data resources providing breadth, depth and good match rates to enhance the modeling process. Collaborative: without collaboration your analytic partner cannot incorporate your business rules and areas of pain into the model. A Good Track Record: ask for demos, references and case studies. Partners who guide the “where do we go from here” question: always provides recommendations associated with specific outcomes that are actionable and solution oriented. Comprehensive: Offers additional tracking or back-end analysis to help you understand the quality of the model and capture trends.
Avoid the one great misconception about analytics and modeling. Many business leaders feel that predictive modeling is just too complex or expensive for small to mid-size businesses. They may also feel that they do not have enough data (customers/responders) to be able to complete a model. The idea of complexity, sample size, and unreasonable expense may have been true a decade ago, but predictive modeling and database mining are now a natural component of doing business. As a result, analysts are now forced to define the process, the deliverables, and the methods associated with predictive modeling in ways that are understandable, meaningful, and most importantly, actionable.
Above all, understand that predictive modeling is all about continuous improvement. Predictive modeling is not a static deliverable or one time event. Be sure to build a model that addresses a desired outcome, define your lists carefully, and create messaging around the desired results. Deploy your modeled data via a marketing campaign to the consumers and/or businesses who have the highest propensity to become a customer...then...
Then...wait! This is where the fun and real learning begins. Through response analysis you may test the efficacy of the model by customer groups, response rates by offer/channel, responses associated with customer segments: all of which can be very revealing–especially when you include ROI into the mix. Follow-up analysis can reveal changes in demographics and behaviors over baseline. Capturing these changes drive refined targeting efforts for the next campaign.
Then the fun begins all over again...the modeling process is an interactive feedback loop that keeps us informed of changes in our customer groups and provides us with the added value to address those changes in a proactive fashion. The marketplace is dynamic and our methods must be proactive if we are to stay ahead of the curve and remain competitive in a changing market place. Modeling can give your business that competitive edge! Try it, you will like it!
Dr. Kate Webster has been Chief Statistician for AccuData Integrated Marketing’s Performance Modeling Group for over 4 years. She has a Ph.D. in Experimental Psychology from the University of Rhode Island and has held many leadership positions over the past 15 years including teaching multivariate statistics at the graduate and undergraduate level, statistical consulting for mid-size businesses, analyst for private and non-profit financial and health industries, as well as a statistical analyst for the Alaska State Legislature. She can be reached by email at kate.webster@accudata.com.
|