Every B2B marketer wants to do a better job of identifying which prospects are in-market and ready to buy, and which are just kicking tires. While there are a lot of vendors who say they’ve got behavioral scoring figured out, the degree of accuracy and coverage can vary widely. If you can avoid scores that sales doesn’t trust and instead find an approach that predicts winners at a very high rate, you’ll reach hero status in your organization. Below are four questions to help you evaluate behavioral modeling techniques:
1. Does the model tell you when prospects will buy?
Many behavioral models simply add points for different activities, but in order to know if someone is likely convert in a fixed timeframe, say “the next three weeks,” you need to take timing into account. A good behavioral model will look at the concentration of activity, the breath of engagement, and it will decay activities at the appropriate rate.
2. Can your behavioral scoring stand on its own without demographic data?
The most robust behavioral models can accurately predict winners without demographic-based information. While your fit and behavioral score complement each other, you want each to stand on it’s own merit. If a vendor munges the two into a single score it is hard to tell if you have a good account with a low behavior score or a bad account with a good behavior score.
3. Is your model looking across every interaction or a small subset?
If it’s not considering all the data at your disposal, you may be missing out on important insight. Many marketing automation systems assign point values to a few key pages . This makes sense in a world where you’re manually defining rules, but with machine learning you want to look at every email click, website visit, and social engagement.
4. Does it look at the combination of different activity types to identify patterns?
Instead of adding points for a given action, your model should discern which unique combinations are most predictive of purchase intent. For example, if you’re building a lead scoring model in your marketing automation system, you might say, “IF this happens add 5 points, OR IF that happens add 10 points.” However, when your model leverages machine learning, it can build its own complex formulas with AND statements, such as “IF a lead does this, this AND this in 24 hours, they’re likely to engage with you this week.”
Once you’ve honed in on an approach that passes this behavioral scoring test, and the model has proven it can deliver statistically accurate scores, you’ll be able to pinpoint the prospects who are likely to buy this quarter or even this week. Click here to learn more about Infer’s predictive behavior scoring, or contact us and we’d be happy to schedule a demo.