RAUXA HOME LOGIN REGISTER

1.25.2012 | Jesse Roberts

• 
• 
• 
• 
• 
• 
• 
  • BOOKMARK & SHARE
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 
    • 

Five ways to use predictive analytics to identify cross-sell and up-sell opportunities.

Companies have access to more information about their customers than ever before. Yet many marketers struggle to capitalize on this wealth of knowledge in meaningful ways that allow them to cost-effectively enhance the customer experience, deepen customer loyalty, and secure future revenues. To do this, Predictive Modeling, an approach focused on helping companies glean actionable intelligence based on historical data, remains direct marketing’s strongest tool. Here are five ways to use predictive analytics to identify opportunities:

1. Identify and organize what you have to work with.
Begin by assessing your available resources. What kinds of data and analytic tools do you have to work with? Try breaking everything down into three goals: short-term, leveraging what is readily available, medium-term, solving for gaps and other areas of improvement, and, long-term, which has three components. Improving the value of your data quality in terms of record deliverability and desirability, improving your data quantity, in terms of deepening your market penetration to reach as many new prospects as possible, and, improving your data warehouse infrastructure, for robust campaign targeting, greater relevance and cost-efficiency.

2. Be careful not to get lost amid success metrics by ensuring your models ask the right questions.
It’s true to say that a response model improves DM performance –but what about those responders’ conversion propensity and potential value? Perhaps “product-specific net conversion models” would be better approaches? Asking your data the wrong question is probably the most common mistake in dataset selection for the model build.

3. Are your prospect and customer databases sizable enough to warrant the investment of custom segmentation?

While modeling predicts certain Key Performance Indicators (KPI)–say, Response %–it does not necessarily explain why. The majority of such predictive tools examine how data elements combined “move together” and to what mathematical extent, but without significant description. A model’s variables may prove out some intuitive insight regarding your target population’s wants/needs, but they are not intended to do so – they should not be considered a profile.

Segmentation allows you to group prospects by their demography, psychographic, or attitudinal natures, as well as their purchasing patterns (for example: their migration among equipment, products, and services). Consumers vary significantly in their purchasing patterns by cause, value and timeframe. Typically, 6–8 relatively “unique” customer personalities reside within larger prospect/customer universes.

Each is unique in its performance levels and demographic/psychographic composition. Understanding such differentiations can be critical in building relevance within creative, copy, and messaging, in addition to testing offers, calls to action, and response channels. Comparative profiling is always recommended across all segments.

4. Leverage multiple analytic tools simultaneously.
Predictive Modeling helps to identify targeting, and Segmentation allows for greater relevance between the recipient and your outreach. When using both analytic tools, be sure to assess performance by both model decile/segment. When combined, you’ll find variances in performance by decile per segment. This informs model depth selection on any KPI basis, such as response, cost per response, conversion, cost per acquisition, and cumulative values.

5. Reporting
Understand your campaigns’ lifecycles and identify appropriate tracking periods. It’s always a good idea to validate your analytic tools used and compare this against prior campaigns’ performance to assess potential toolkit degradation. In addition, you should be sure to measure incremental performance levels. Maintain a “no touch” perpetual holdout group to assess organic market behavior, and compare that versus the “treat” population to assess each KPI’s incremental difference–this is the true gain of your marketing dollars.