Looking for a way to prioritize leads without relying on gut instinct or whatever’s at the top of your CRM? If you’re in sales ops, enablement, or just the one who has to explain why your pipeline is full of “maybes,” you’ve probably wondered if there’s a better way. That’s where building a custom scoring model in Propensity comes in.
This guide is for sales teams who want more than a black-box “AI score,” but don’t have time to get a PhD in data science. No fluff, no buzzwords—just the nuts and bolts of making scoring work for your deals, your way.
Why Custom Scoring Models Matter
Let’s be real: Most sales “scores” are either too generic or so complex nobody trusts them. Out-of-the-box models might work for someone else’s pipeline, but your team’s deals are different. Maybe you care about company size, maybe you don’t. Maybe last email open means everything, or almost nothing.
A custom scoring model lets you:
- Focus reps on the right leads, not just the loudest ones
- Spot deals moving off-track, fast
- Make reporting actually useful (not “activity theater”)
But don’t get suckered by hype: No scoring model is magic. It’s just a tool to help you make better bets. The real work is deciding which signals matter for your sales process—and actually using the insights.
Step 1: Get Clear on What You Want to Score
Before you even open Propensity, get specific. What are you trying to predict? Most sales teams care about things like:
- Likelihood to close (classic)
- Likelihood to respond to outreach
- Likelihood to book a meeting
Pick one. If you try to score everything, you’ll end up with a mess. Write down what “success” looks like for this model—e.g., “We want to predict which leads will close in the next 60 days.”
Pro tip: Talk to your reps. They usually know which signals actually mean something (and which are just noise).
Step 2: Gather Your Data (Don’t Overthink It)
Propensity can pull in a lot of data: CRM fields, activity logs, product usage, website visits, you name it. But more data isn’t always better. Start simple:
- Contact/company info: Industry, size, location
- Engagement: Email opens, replies, meetings booked
- Deal history: Past wins/losses, deal size
- Product behavior (if you have it): Sign-ups, logins, key actions
You don’t need a complete data warehouse. Just make sure the data you use is:
- Accurate: If your CRM is full of junk, clean it up first
- Consistent: Fields should mean the same thing everywhere
- Relevant: If nobody cares about Twitter followers, skip it
What to ignore: Vanity metrics (e.g., number of CRM fields filled in), incomplete fields, or anything you can’t explain to a new hire in one sentence.
Step 3: Define Your Scoring Criteria
Now comes the “custom” part. Decide which factors should go into your score. Here’s a no-nonsense approach:
- List your variables: Rank them by what actually seems to matter (not what’s easy to pull).
- Assign weights: Not all signals are created equal. Maybe “Booked a demo” is worth 10 points, but “Opened an email” is only 2.
- Set thresholds: What score means “hot lead”? What’s a “cold prospect”?
Example:
| Signal | Points | |-----------------------|--------| | Booked a demo | 10 | | Replied to email | 5 | | Company size > 100 | 4 | | Visited pricing page | 3 |
Don’t get too fancy—your first version should fit on a napkin.
Step 4: Build Your Model in Propensity
Here’s where Propensity does the heavy lifting, but you still drive. The basic workflow looks like this:
- Connect your data sources: Plug in your CRM, marketing automation, product analytics, etc. Most integrations are point-and-click, but check that fields match up.
- Create a new scoring model: In Propensity, you’ll usually choose “Custom Score” or similar. Pick the object you want to score (Leads, Contacts, Accounts).
- Add your criteria and weights: Map each signal to your fields. For points-based models, set the score per action. For advanced users, you can try regression or machine learning—but don’t do this unless you understand what’s happening under the hood.
- Set your scoring formula: Propensity lets you choose simple additive (points-based) or more complex logic. Start with additive; you can always get fancier later.
- Test with sample data: Run your model on a handful of known deals. Do the “good” ones actually score higher? If not, tweak weights or swap out signals.
Pro tip: Don’t try to automate everything on day one. Manual review beats auto-pilot, especially early on.
Step 5: Roll Out Your Score (and Actually Use It)
A scoring model nobody trusts is worse than no model at all. Make the rollout simple:
- Show the score in context: Put it front-and-center in your CRM or wherever reps work.
- Explain the “why”: Share how the score is calculated. If reps think it’s a black box, they’ll ignore it.
- Use it to prioritize: For example, “Anyone above 15 points gets a call this week.”
- Get feedback: Ask reps what’s working and what isn’t. Tweak as needed.
What doesn’t work: Forcing reps to follow the score if it doesn’t match reality. If the model is wrong, fix the model—not your team’s instincts.
Step 6: Iterate and Improve
No model is perfect out of the gate. Some signals will turn out to be useless. Others you thought didn’t matter might be gold. Here’s how to keep improving:
- Review closed deals: Did your high scorers actually close? If not, what’s missing?
- Cut dead weight: Remove signals that don’t move the needle.
- Add new signals: As your process changes, update the model.
- Keep it explainable: If you can’t explain the score to a new rep in 60 seconds, you’re overcomplicating it.
Pro tip: Review your scoring model every quarter. It’s not “set and forget.”
Common Traps (and How to Dodge Them)
Here’s where most teams slip up:
- Chasing perfection: A “good enough” model you use beats a “perfect” model you never finish.
- Overfitting: Don’t cram in every possible signal. More isn’t always better.
- Ignoring rep feedback: If your team hates the score, something’s off.
- Letting the model go stale: Your sales process changes—so should your score.
Keep It Simple, Ship It, and Keep Tweaking
Custom scoring in Propensity isn’t rocket science, but it’s easy to make it harder than it needs to be. Start with what you know, keep it simple, and don’t be afraid to adjust as you go. The best models are the ones your team actually uses—and trusts. Forget the hype. Build the tool that helps you win more deals, and let the results do the talking.