Skip to Content Skip to Footer
5 Steps to More Effective Audience Targeting

5 Steps to More Effective Audience Targeting

Kate Cullen and Lauren Littlejohn

Follow these steps to better harness your customer data and remove the guesswork from targeting

Marketers have mountains of customer data and a continually evolving tech stack at their disposal. This has made using primary research and contextual signals alone to identify target audiences artifacts of a bygone era. Marketers are in a unique position to use data to retroactively evaluate the performance of a campaign and determine its audience. These predictions will take the guesswork out of targeting and offer a greater sales lift than traditional targeting practices ever could.

But if you are sitting on heaps of customer data with no idea where to start, do not be disheartened. Follow these five steps to target customers more effectively using data.

1. Build a Data Repository

Track who your customers are, what communications they are receiving on which channels, and which products/campaigns resonate with them. It’s never too late to start.

2. Analyze Your Customer Base and Learn Everything Possible About Them

Feed all the segmentation work and customer attributes you’ve collected over time into this repository. These customer dimensions will make the model smarter over time, the more it learns about your customers. Identify gaps in knowledge about your customers and prioritize which data points are most useful to pursue.

3. Choose a Machine-Learning Model and Feed It Data

The model-building process is the most important, as it’s what weaves together all data points to return a recommendation. You will need enough data to form three different modeling groups for training, model validation and one holdout. The training observations inform the model which attributes about the household or product are predictive. The validation observations help to identify which model or version of a model is best, by giving you a set of unseen data on which to evaluate the accuracy of each model.

Once you finalize your model, use it to predict the response of your holdout group. How accurately the model predicts this last group of data informs you how well the model will generalize to the unseen data you are trying to predict.


4. Optimize and Fine-Tune Based on Your Model Observations

Your model will continue improving over time with the more data you have available to feed it. Ensure you have the infrastructure in place to seamlessly feed data back into your models.

5. Withhold a Control Group and Measure

Focus on the importance of upfront data used for targeting as well as closing the data loop by returning how the campaign performed for each customer.

Our work should not stop at understanding customer segments within market research. Rather, it should continue by partnering with data scientists to determine how we connect the mass of data at our fingertips and allow it to show us how to make our marketing more effective. We invite you to dispel the way targeting has always been done and explore a new way of harnessing data to remove the guesswork in targeting.

Kate Cullen is lead consultant on product, strategy and innovation for KPM at 84.51˚.

Lauren Littlejohn is director of data science for KPM at 84.51˚.