Wiland Blog

Better response rates. Higher average order or gift sizes. Greater long-term value. Broader reach. These are all success metrics that brands and organizations might emphasize in a given marketing campaign. It is what we at Wiland work to help them achieve every day with our predictive marketing audiences.

But these goals are usually not isolated nor exclusive of one another. A brand or organization might be seeking to balance high response rates with generous purchase or gift amounts, or to find a larger universe of prospects while still prioritizing long-term value.

While an individual model is predictive as a standalone tool to meet a specific campaign goal, we find that it’s through multi-model marketing audiences that organizations benefit from the most predictive power—translated tangibly into the largest audience volumes and the best campaign results across multiple KPIs.

So what are multi-model marketing audiences? By understanding how they work and why they are so powerful, data-driven marketers can be better equipped to run campaigns that perform consistently well.

Back to the Basics

Predictive modeling can accurately identify the people who are most likely to engage with, spend with, or give to a particular organization. It creates strong prospect audience segments for acquisition campaigns that outperform typical list services segmentation. Predictive modeling finds variables in the data that are unique to an organization’s ideal customers or donors that might not be intuitive, but that may have strong resonance when analyzed in relationship to one another. Predictive modeling that is driven by artificial intelligence (AI) and other types of machine learning is particularly powerful in this regard due to its efficiency and effectiveness when analyzing massive volumes of data. It enables the creation of ranked universes of prospects, customers, or donors that perform strongly by interpreting information that human analysis simply could not approach.

The Strength of Individual Models

Individual models each have a specific strength and typically focus on a single success metric. For instance, Wiland’s CLEAR™ Advantage model uses a proprietary decision tree algorithm to identify prospects with the likelihood of being high long-term value customers or donors, but with less of a focus on high initial response rates. Meanwhile, Wiland’s Enhanced Acquisition Response model uses a machine learning regression combination to select prospects who are the most likely to respond to a specific offer, but with less of a focus on the performance of those customers or donors after the first response.

Using an individual model can be an excellent choice for newer brands and organizations, or if a company’s success metric is perfectly in line with a single model’s unique strength. But it’s when models are combined as the building blocks of a multi-model marketing audience that they can achieve more complex, robust campaign goals.

The Multi-Model Difference

A multi-model solution is a final modeling algorithm that uses individual models as inputs rather than the thousands of variables that the individual models themselves access. This means that the number of inputs into a multi-model solution is considerably less than in an individual model. However, the multi-model solution benefits from the many different variables, modeling techniques, and strengths of each of the individual models used within it.

The best multi-model solutions include a blend of models built using different methods. For instance, Wiland’s Universal Performance (UP)—an integrated multi-model audience ranking solution—incorporates the wide variety of algorithms that our individual models bring to the table. A single Wiland UP can utilize the intelligence of machine learning regression and decision trees, AI gradient boost logic, and other proprietary model techniques developed specifically for prospecting. We also often find that the combination of Profile models (those that indicate the characteristics of ideal buyers or donors) with Response models to be highly effective. The Response models bring all of the intelligence from a client’s past promotions while the Profile models help control for existing client bias. This ensures that the modeled universe can expand to include both people with known client criteria and the new, potentially undiscovered prospects likely to provide high value for the client.