Predictive Models

Predictive Models for Enhancement and Model Development

Predictive models improve your ability to reach the right audience.  We like to think of these predictive models (aka propensity models) as “orthogonal” to traditional demographic models because they give you a whole new dimension for targeting new prospects or gaining insight about existing customers.

For example, think about how difficult it would be to use traditional demographics to reach people who are in-market for a new car or a new home.  Only a multi-dimensional model based on millions of data points will work in this case.  That’s how these models are created.  They start with known buyers, users or other propensities and use predictive analytics to create and test a model that can then be applied to all U.S. households, which are scored on a scale (typically 1 to 20).  These propensity models can then be used, along with other demographics as needed, to either select a precise target audience or to append to existing customer records.  Then you can you can use the models in cross-sell/up-sell campaigns or to personalize interactions.

Furthermore, thousands of models exist with several dozen for just about every industry, ready pre-built and off-the-shelf (and updated monthly) to help improve your data-driven marketing and analytics.

Categories of Predictive / Propensity Models

The primary categories for which pre-built predictive models are available follow.  These models predict industry-specific brand affinity and product propensity and provide actionable insight into consumers and the marketplace.

Within each category, models exist for a wide range of attitudes and propensities related to:

• In-Market Propensity
• Attitude and Behavior Propensity
• Brand Propensity (thousands of specific brands across all product categories)
• Media Usage Propensity
• Purchase Channel Propensity
• Product Propensity
• Spending Propensity

In-Market Propensity Models

We especially like the “in-market” models since they are almost impossible to create using traditional third-party demographics.  It is also very difficult to create models based on your first-party data because of your limited scope of interaction and overall knowledge of behaviors and activities both online and offline.

Here are just some of the thousands of examples of in-market predictive / propensity models:

• Furniture Propensity: In Market for Furniture – likelihood to spend $2,500 plus on furniture in next 90 days
• Remodel Propensity: In Market to Remodel – likelihood to spend $1,000 plus on renovation in next 90 days
• Cosmetic Procedure Propensity: In Market for Cosmetic Procedure – likelihood to have a procedure in next 12 months
• Get Engaged Propensity: In Market to Get Engaged – likelihood to get engaged in next 12 months
• Have a Baby Propensity: In Market to Have a Baby – likelihood to have a baby in next 12 months
• Retire Propensity: In Market to Retire – likelihood to retire in next 12 months
• Sell Business Propensity: In Market to Sell a Business – likelihood to sell a business in next 12 months
• Buy Vehicle Propensity: In Market to Buy a New Vehicle – likelihood to be in market for a new vehicle
• Buy Home Propensity: In Market to Build or Buy a Home – likelihood to build/buy a home in next 12 months
• Buy Watch Propensity: In Market to Buy a Watch/Jewelry – likelihood to buy for $2,500 plus in next 12 months

Attitude and Behavior Models

Attitudes and Behavior models also go far beyond what one would typically expect from third-party demographic targeting/enhancement data elements, for example:

• Highly Likely Investor – top 10% of households likely to be investors
• Price Sensitive Penny Pinchers – likelihood to be price sensitive
• Large Contributor – likelihood to have Contributed More than $500 in past 12 months
• Zoo or Aquarium Visitors – likelihood to Visit Zoo or Aquarium While Traveling
• Luxury Fashionista – interested in high-end designer clothing and couture
• Likely to Switch Insurance Provider – likelihood to switch insurance provider
• Internet Investors – likelihood to have used internet for investments past 12 months
• Sleep Quality – predicts the likelihood of sleep quality
• Diet Lifestyle – predicts the likelihood of diet being followed
• Affluent Tech Early Adopter – likely to be an affluent tech early adopter

How Were these Predictive Models Developed and Why Should I Trust Them?

Some of the largest, most highly-respected consumer data companies in the world built these models.  This list of companies includes Acxiom, American Express, Claritas, and Semcasting.   Each has access to known purchases, demographics or other behaviors and attitudes on millions of consumers.   Their data scientists analyze the correlations between the known information about households that have been observed to have a certain behavior with those that do not.  Then they use advanced statistical methods to create multi-variate propensity models.

Almost every U.S. household can then be scored using a scale that depends on data granularity but is typically 1 to 20 (some models use a 1 to 10 or “decile” scale and others use a scale of 1 to 100).   The analysts back test and prove each model before it is released.  Therefore, if a model does not prove to be highly predictive it is not published for use.

Also, the companies that generate these predictive models are the leaders in privacy and ethical use of data and relevant regulations.  Some models are restricted: we will need to review the specific use-case to determine if the model is allowed for that case.   For example, different versions of many models comply with FCRA regulations so they can be used by financial institutions.  The models themselves are updated periodically and the underlying demographic database is updated monthly.

In addition to propensity models, you will find many other data options within Demographics and Behavior Data and Segmentation Models.

Data Marketing Strategies team can help you determine which of the thousands of available pre-built predictive models are best for your customer database, CRM system or Customer Data Platform, or specific analytic project.