Show Me The Data: Model What Doesn’t Exist

Can you build a model to predict the potential for a product that doesn’t yet exist? Sure—if you can define your idea clearly enough for consumers to understand the product features and you can estimate the potential market for that product or service.

If you combine research with attitudinal modeling you will be able to define your target audience and see if there will be enough to justify developing the product and the marketing campaign.

Clearly defining your idea is the key to modeling a prototype product. Defining a context for your idea will let your respondents visualize what they might get. The concepts of a self-cleaning oven or a handheld device to read electronic books on a small screen or last-minute overnight package delivery during the holidays are all items that can be clearly communicated to consumers. With this succinct definition you can conduct a nationally-representative random telephone survey.

What you want to end up with is a yes/no response indicating their projection of their likely use of the product described. Finish your survey with questions that will let you create a demographic profile of the respondents by age, household size, income, education level, length of residence and gender.

You will need to determine how representative your sample is by comparing the overall demographic profile of both potential users and non-users to census demographics for their areas of residence. If your sample is representative, you can turn your research results over to your internal or external modeling analysts.

The analysts will go through a multistage process. They will recode and transform the data to prepare it for data exploration and modeling. They will perform frequency analyses and decision tree and variable interaction analyses as well as logistic or other interaction and regression modeling. They will be able to produce a statistical model that really does distinguish potential users from potential non-users even if the differences between the groups are subtle.

The factors of this model can then be applied to a prospect database. Your list broker can provide you several options for excellent consumer cooperative prospect databases that include previous mail-order buying history as well as full demographics.

Applying the model to the database will allow the list provider to give a likely-to-buy score to hundreds of thousands of potential buyers of your non-existent product. You can receive a table listing the deciles of the scored prospects to use to estimate the size of your potential market.

Once you create your product you can use the same table to select the prospects to which you would like to mail your offer.

If you are speculating about producing a new product, you don’t need to go blindly into significant investments in research and development, production and marketing. By spending a fraction of your development costs on marketing research and sound modeling techniques you can have a very good idea of the likely market and the number of buyers for your product before it has ever been produced.

The largest consumer product companies perform this type of experimental research and modeling all the time. They do it because it is cheaper to do than to market a product likely to fail. You can learn from their example and model something that doesn’t yet exist and actually get something out of nothing to benefit your company.

Bill Singleton ( is a manager of analytics and consulting services at The Allant Group.