The Rise of Propensity Modeling and Value Based Marketing in 2023
As the cookie continues to crumble, marketers have turned their attention to first-party data, grabbing up as much as possible from analytics tools, site logins, CRM systems, and more. First-party data can be sophisticated with useful information about potential or current customers, even helping to gauge the value of these customers.
However, first-party data often lacks the scale needed for marketing efforts. So, how do brands scale their efforts, while finding the best customers? For many of these brands, propensity modeling and value-based bidding is the answer.
Propensity modeling uses machine learning and AI to analyze millions of data sets and predict whether someone will take a particular action, such as making a purchase or churning. This allows marketers to identify the highest value consumer (those most likely to take the desired action) and focus their efforts on reaching them with the right message at the right time (value-based marketing).
Still, there are several pitfalls that marketers should avoid when using propensity modeling and value-based bidding, such as unrealistic goals, not properly aligning goals or data paralysis. Let's break down how marketing based on value works and how to make it successful for your business.
Since not all customers are created equally, it's important to capture enough data to help define their worth. While the ultimate goal is lifetime value (LTV), this is often hard to calculate, so don't be deterred if you don't have LTV metrics yet. Instead, use other conversion events to help prove value.
For example, imagine a large financial institution trying to sell a high yield savings account. A standard conversion typically answers a yes or no question.
- "Did the user start filling out an application?
- "Did the user open an account?"
- "Did the user fund their account?"
While a funded account is typically worth more than an incomplete application, what if the value can be defined?
- "How much money did they put in their account?"
- "What is the customer's LTV?"
Once the data is collected, propensity models can be created for each of these conversion events. For example:
- "I want to find users who are likely to fill out an application."
- "I want to find users who are likely to open an account."
- "I want to find users who are likely to fund their account."
- "I want to find users who are likely to fund $5,000+."
- "I want to find users who are likely to have an LTV of $50,000+."
When we moved a large online bank from using account opens to instead use funded account data in their display bidding strategy, the average cost per account opened increased by nearly 50 percent.
Sound amazing? Ready to get started with marketing based on customer value? Here are some pitfalls to avoid making sure your campaign is a success.
Potential Pitfall #1: Unrealistic Goals for Customers
While it would be nice if every customer was a unicorn, it's just not realistic. There needs to be a combination of quality with quantity to gain enough scale. Test various segments and values to find that balance. So, while targeting users likely to fund at $1 million might not work, targeting users that fund at $100,000 might. Also make sure to pair propensity targeting with value bidding to ensure you don't overpay for customers.
Potential Pitfall #2: Quality Might Be Less Quantity
When moving from quantity to quality, realize that quantity will most likely decrease. Yes, propensity models will help to scale, but you might pass up some less valuable conversions that normally would be in marketing reports making CPA metrics look higher.
Potential Pitfall #3: Data Paralysis
No data scientist team? No LTV? This doesn't mean marketing based on value can't be accomplished. Get to the data that is possible, and then work on up to more value. While working with a live ticketing company, we started by creating a propensity model based on the likelihood of purchasing a ticket within 90 days. The model was highly successful, with a double-digit percentage drop in cost per acquisition through search and room to move toward ticket value in the next model.
Potential Pitfall #4: Misaligned Goals for Marketers
If targeting and bidding is based on value, goals should also be based on value. Metrics need to change from a typical CPA to more robust metrics like return on ad spend (ROAS), or cost per net new revenue.
Using propensity modeling and bidding to value will be an essential part of future marketing plans, helping to scale first-party data and use AI to pay for customers based on their value.
The views and opinions expressed are solely those of the contributor and do not necessarily reflect the official position of the ANA or imply endorsement from the ANA.
Lindsey Levich is senior director of business development and Google marketing platforms and services at Bounteous.