Demand Generation: AI Lead Segmentation in 3 Steps

B2B lead segmentation is manual and complex. But it doesn’t need to be. Here’s a way to use AI to do intelligent lead segmentation. 

The Challenge of Segmenting B2B Leads

Let’s say you sell martech software to B2B companies. You’d think segmenting based on would be easy. Just look at a lead’s title. A CMO, well that would be a decision maker. A demand generation specialist? They’d be a user. A product marketer? They’d be an influencer.

Most people create a rule in their marketing automation platform (MAP) to do keyword matching (e.g., look for a title that contains ‘Chief Marketing Officer’ or ‘CMO’) on the title. The match then puts the lead in the appropriate persona campaign. Here’s a sample of doing this in Marketo via a smart list

Creative titles break these rules. Head of Global Marketing. Chief Marketing Digital Officer. And my favorite, VP of Making it Rain. If you don’t have a rule for it, it never gets assigned to a campaign.

The Power of AI in Classifying Job Titles

You can use AI to classify job titles into 3 categories: decision maker, user, and influencer. And then build your MAP automation around these titles instead. That should greatly simplify your rules.

‍

1. To start, I asked ChatGPT to give me an exhaustive list of roles within the marketing team. 

2. I then gave it some context about what this hypothetical company team does, and then gave it this list of generated roles and asked it to categorize the roles into decision maker, user, and influencer. 

An alternate approach would be a zero-shot training approach where I give the system sample titles for each of these 3 roles and then ask it to classify the list of roles that it had previously created.

‍

3. You can see from my last prompt that I told the system that it is a NER (named entity recognition) system. I gave it context by feeding it all of the marketing roles and their classification into decision maker, user, and influencer. And then gave it a list of titles of leads found inside a marketing system. And then had it classify them.

Were the results perfect? No. But you can definitely see the potential. And getting this to work is much better than having to code complex rules in Hubspot or Marketo for every title under the sun. 

‍

‍

‍