If you’re using Madkudu to score leads, you’ve probably realized the out-of-the-box setup is just a starting point. Default scoring models are nice, but your own funnel, data, and definitions of “good lead” are what really matter. This guide is for marketers and ops folks who want to actually trust their lead scores—and not just take them at face value.
Whether you’re frustrated with weird scoring results, or you just want more control, I’ll walk you through customizing predictive scoring fields in Madkudu so they actually fit your business. We’ll cover real steps, what’s worth tweaking, and common mistakes to skip.
1. Understand What Madkudu Predictive Scoring Really Does
First, a reality check. Madkudu’s predictive scoring uses your lead data to guess how likely someone is to convert, using things like company info, job title, website activity, and whatever you feed in from your CRM or MAP. It’s not magic. It’s a mix of data science, some black-box modeling, and whatever signals you set up.
Before you start customizing, know: - You can tweak which fields and signals Madkudu uses to score. - You can’t see every detail of their proprietary model logic (unless you’re on the “Data Science Studio” plan, and even then it’s not all open). - Your results are only as good as the data you send in. Garbage in, garbage out.
If you’re hoping Madkudu will fix a messy funnel or spot "perfect" leads just by toggling settings, slow down. The goal here is to align scoring with how your sales and marketing actually work.
2. Map Out Your Funnel and Define What a “Good Lead” Means (For Real)
Don’t jump into Madkudu yet. First, get your house in order:
- Talk to sales. What do they actually want? Which leads are a waste of time?
- Audit your CRM/MAP fields. Which fields are reliable? Which are junk or empty?
- List your “must-haves.” Is company size a dealbreaker? Industry? Job title? Website behavior?
Pro tip: Write down your “ideal customer profile” and the red flags for bad leads. If you can’t put this on paper, you’re not ready to customize scoring.
3. Review Your Data Inputs in Madkudu
Madkudu can only use what you send it. Double-check:
- Are all the fields you care about (like industry, revenue, title, etc.) syncing cleanly?
- Are fields named consistently? (“Company size” vs “Number of Employees”)
- Are there gaps or bad values (like “N/A”, “Unknown”, “Test”)? Clean those up first.
If you’re missing critical data, fix your integrations before you build fancy scoring logic. Otherwise, you’re just building a castle on sand.
4. Dive into the Madkudu Predictive Scoring Model Settings
Now, let’s get hands-on. Madkudu usually gives you a couple of ways to customize scoring fields, depending on your plan:
- Business Plan: Some field mapping and weighting, but limited model editing.
- Data Science Studio: Full access to model configuration, custom signals, and more (if you’re not on this plan, you’ll hit some walls).
Where to Go:
- Log in to Madkudu.
- Go to the “Predictive Scoring” section.
- Find “Model” or “Signals” (the exact spot depends on your plan).
- Look for “Fields Used in Scoring” or “Attributes”—this is where you can map and prioritize fields.
5. Customize Scoring Fields: What to Change (and What to Ignore)
Here’s the meat of it. You want Madkudu to pay attention to what matters for your funnel—not what some generic SaaS company cares about.
a. Add or Remove Fields
- Add fields that actually predict conversion for you (e.g., “Tech Stack,” “Annual Revenue,” “Product Usage”).
- Remove fields that are empty, unreliable, or just don’t matter (e.g., “Middle Name,” “Fax Number”—seriously).
b. Map Custom Fields
- If you use custom CRM fields (like “Account Tier” or “Lead Source”), map them so Madkudu sees them.
- Double-check field types (string, number, etc.) and values—Madkudu can misinterpret these if they’re messy.
c. Adjust Weighting (If Available)
- Some plans let you tweak how much certain fields influence the score.
- Don’t go wild here. Small changes can have big ripple effects—test before you trust the numbers.
d. Create Custom Signals (Advanced)
- On higher plans, you can define your own “signals” (like “Requested a demo AND is in the target industry”).
- Use these for real buying signals, not just vanity metrics.
What to ignore: Don’t get sucked into using every field “just in case.” More fields don’t mean better scores. Every extra field is another way to get garbage results.
6. Test, Validate, and Don’t Trust Blindly
Once you’ve customized fields, don’t just turn it on and walk away.
- Run historical leads through the new model. Do your “A leads” actually get scored as “A”? Are there surprises?
- Spot-check examples. Pick a handful of recent closed-won and closed-lost leads. Do their scores make sense?
- Ask sales for feedback. Seriously—do this. They’ll spot the weird stuff you miss.
Got bad matches? - Revisit your field mapping and weights. - Check your data quality again. - Sometimes, less is more—try removing noisy fields and see if scores improve.
7. Roll Out Gradually—And Watch What Happens
Don’t push your new scoring to every team or tool right away. Instead:
- Pilot with a small group. Pick a few reps or campaigns and watch how the new scores work.
- Monitor conversion rates, handoff quality, and feedback.
- Keep an eye on exceptions. Are valuable leads slipping through? Are too many junk leads getting “A” scores?
Iterate. Predictive scoring is never “set it and forget it.” Your business will change, your data will change, and your model needs to keep up.
8. Common Pitfalls and Real-World Advice
You’ll hear a lot of hype about AI-driven lead scoring “transforming” pipelines. Here’s the honest truth:
- It’s only as good as your data. No amount of algorithm magic will fix bad or missing info.
- Don’t trust the default model. It’s generic. Your business isn’t.
- Don’t overcomplicate it. Every extra field or rule is another way to break things.
- Review it quarterly. What worked last quarter may not work now.
- Score inflation happens. If “A” leads are everywhere, scoring means nothing.
Pro tip: Keep a “bad leads” list and regularly spot-check why they scored the way they did. That’s the fastest way to spot issues.
Keep It Simple—And Keep Tweaking
Customizing predictive scoring fields in Madkudu isn’t rocket science, but it does take some real work. Focus on your best data, only use fields that matter, and don’t expect overnight magic. The key is to start simple, test everything, and keep iterating as your business evolves.
You’ll get better results—and a lot fewer headaches—by resisting the urge to tinker endlessly or chase every shiny feature. Get your basics right, and your lead scores will actually mean something.