Skip to content

Market Modeling for Optimal Targeting

Case Study: A Car Parts Company wanted to expand into New York in 2023 and wanted to make sure they were targeting the highest opportunity areas, both with stores and online marketing.

Step 1: Define business objective then gather and validate appropriate data.

Step 2: Find insights, model and predict future outcomes, and propose actions.

Step 3: Follow through on actions, measure the effects, adjust as needed.

Industry Success

  • I’ve applied similar market modeling at Pfizer for targeted marketing, field force sizing, and new product launches.

Outcomes of Value from this Case Study

  • Targeted marketing to counties with 1.5x more spending per person, resulting in higher conversion rate, customer value, and lower cost to acquire customer
  • Gaining deeper understanding of the customer base and their drivers
  • New product, location, and service launch guidance
Step 1: Find the target outcome data. I started by gathering data on car part sales over the previous few years. I found sales at the county level and normalized it to taxable $ spent per working age adult in the county that year. We see most counties cluster between 200 and 500 $ per working age adult with some outliers above that range.
Step 1 Continued: Visualize the target outcome data. We can visualize the normalized sales (Score) by county from 2022. We see that there are a few top counties spread throughout the state, with a few high sales counties clustering together.
Step 2: Join the target data with demographic data to find predictive relationships. With hundreds of demographic variables available from the American Community Survey, we are able to find drivers of a given product purchasing behavior. We saw armed forces population and working from home was actually associated with increased normalized parts sales, and the count of people with taxable earnings (a proxy for population density and public transportation) was negatively correlated with car parts sales.
Step 2 Continued: Validating the predictive accuracy of our sales model. Using the data available at the start of the project (2022 and prior) to predict the 2023 results, we see strong directional accuracy which helped target the highest opportunity areas. A further quantification of the results will be seen below.

This image has an empty alt attribute; its file name is CarPartsResults-1.png

Step 3: Acting on the predictive model and measuring impact. We selected the counties we predicted to have sales per working age person in the top 20% to market to. Our selected counties had 1.5x higher value per person.
, , ,