Power your Marketing.

How Marketing Models Minimize Guesswork

How Marketing Models Minimize Guesswork and Maximize Predictability

Wes Sparling

In today’s direct marketing universe, we have options for how we target prospective customers. We can target based on personas, on one or more criteria generally referred to as “criteria selection,” or we can target based on models. In my last article, I expanded on personas and how they are used in targeting. In this article, I will briefly touch on marketing models and how they are used in targeting.

Personas and Marketing Models Don’t Always Mix

First off, it’s important to understand that personas and models are not easily combined in most cases, as they are really two different approaches to targeting. Personas are perhaps more interesting to marketers in mass media tactics when communicating large benefit propositions that apply to large groups of targets. A persona might help you think about your prospective audience as a group with common characteristics. You might then execute a TV campaign with messaging that appeals to these common characteristics.

For many of our customers who are new to direct mail, we work with them to refine their approach from a targeting strategy that is broad and mass media focused toward one that is individually based and molded. In all cases, we are typically working from a customer or response file and creating a picture of the respondents or customers based on data attributes. Then we move out into the broader universe to find targets and rank them in accordance with how much they look like the data picture we created. These marketing models can be created to accept real time data feeds and can be updated automatically or manually using database integration and appropriate data feeds.

Models Look at Data on a Personal Level

A model is truly individually based on a number of data factors; each ranked by relative importance to define and isolate a specific group of individuals from a broad population. There are a number of model types, methods of model creation, and analytics that help us create models. These models can be relatively simple, such as a demographic profile lookalike model (which uses only demographic attributes in the analysis and rank orders them based on a customer file), or we can go far deeper.

A simplistic way to think about modeling by personal characteristics might be as follows:

Suppose the model indicates that those who are most likely to respond have 10 common attributes. If the most important attribute of the 10 was found to be a demographic factor such as “female,” and the second factor was a particular “age range” (let’s say between 35 and 40 years old), then the model would select only females between the ages of 35 and 40. A different and more likely scenario is a model rank ordering a series of selection such as sex, age, FICO score, home ownership, catalog shopper, political persuasion, technology interest, etc.

Models Work Together to Find the Best Fit

If our example model were far more complex, and in fact had multiple models as part of a single aggregated model score (which we would call “meta models”), then perhaps these demographic attributes could intersect with financial attributes or other record level attributes to create a very detailed, statistically-based targeting profile. The result would be a selection from a population who “fit” the model. This target would be scored and ranked.

Allowing for a best fit and highest probability of response lets a marketing model developer rank prospects into groups where fit is better, with the upper deciles usually being best fit and the lower deciles the worst fit. The result is a rank order and probability of model performance from prospects selected from the broad population based on the model. In short, if the model is a response model, there is a way to predict the response expected from decile 1 versus deciles 2-10. This allows marketers to make priority decisions.

Models Help Keep Cost Per Acquisition (CPA) Low

The result of an effective modeling effort is a lot less guessing and a whole lot more predictability in our clients’ marketing results. Models provide the opportunity to narrow the focus of your marketing toward those customers most likely to respond or buy. This drives down CPA by increasing the efficiency of your efforts. For most clients who spend a significant amount of money to market products or services, predictability is important. That’s why this is the primary use of a model for most marketers. It is, however, far from its only use: You can model against predictable revenue, claims, churn, or other factors from a host of measurable key performance indicators (KPIs).

In a modern direct marketing environment, careful thought must be given toward customer journey, attribution, profiles, marketing models, and the KPIs needed to effectively measure success and failure. To develop an effective and powerful campaign, your marketing strategy will need to carefully consider what type of segmentation approach you use and how deep you go into modeling or other targeting approaches. At IWCO Direct, we can help you reach your goals—so get in touch with us today. We look forward to starting the discussion.

link https://www.iwco.com/blog/2017/10/27/marketing-models-audience-segmentation/
Wes Sparling

Author

Wes Sparling

Vice President of Marketing Strategy at IWCO Direct. Graduate of Grand Canyon University. Joining IWCO Direct in 2014, he brings the "plan diligently and act definitively" philosophy to a variety of client campaigns. This father and former Arizona High School Track and Field Regionals discus champion loves stream fishing for trout, rooting for the Los Angeles Dodgers and his three dogs.

More Posts by Wes

Subscribe

Subscribe to SpeakingDIRECT to have new articles delivered to your inbox as they post. We promise to keep it fresh and interesting.

Steve Myrvold, EVP of Operations and Customer Experience, Elected to the APTech Board of DirectorsRead More
keyboard_arrow_upBack to top