If you’re in B2B sales and tired of chasing dead-end leads, you’ve probably heard the hype about lead scoring. Done right, it helps your reps spend time on folks who might actually buy. Done wrong, it’s just another dashboard nobody trusts. This guide is for sales and ops teams who want a straight-shooting walkthrough on setting up a lead scoring model in Madkudu—no fluff, just what actually works.
Why Bother With Lead Scoring?
Let’s get this out of the way: most sales teams waste a ton of time on leads that will never convert. Lead scoring is just a way to stack-rank your inbound leads so you can stop guessing who’s worth a call. Madkudu automates a lot of the heavy lifting—but only if you set it up right.
But don’t expect magic. Even the best models are only as good as your inputs and your process. If your CRM is a mess, or your sales team ignores the scores, no tool will save you.
Ready? Here’s how to actually build and launch a lead scoring model in Madkudu that your sales team will use (and trust).
Step 1: Get Your Data House in Order
Before you touch Madkudu, check your data. Lead scoring is garbage-in, garbage-out.
- Audit your CRM and marketing data. Are fields like company name, email, job title, and website filled out and accurate?
- Standardize where you can. Pick one source of truth for company info (like Clearbit or LinkedIn) and stick to it.
- Fix obvious junk. Delete or merge obvious duplicates. If you have tons of Gmail/yahoo leads, decide upfront if you care about those.
- Map fields. Know which fields in your CRM map to the concepts Madkudu uses (like “Company Size” or “Industry”).
Pro tip: If your Salesforce or HubSpot instance is a rat’s nest, spend the extra day cleaning now. You’ll save way more time later.
Step 2: Connect Madkudu to Your Systems
Madkudu needs access to your CRM and (optionally) your marketing automation platform.
- Connect your CRM. Madkudu has built-in integrations for Salesforce, HubSpot, Marketo, and others. Usually it’s OAuth-style—takes 10-20 minutes.
- Pick what to sync. Don’t just sync every field. Start with the basics: name, email, company, job title, website, lead source.
- Set up enrichment. Madkudu can enrich leads with firmographic and technographic data, but only if you turn it on. Decide if you want to pay for enrichment, or if your CRM is already doing this via another tool.
- Test the connection. Import a few records and check that fields line up. Fix any mapping issues now, not later.
Step 3: Decide What “Good” Looks Like
If you don’t know what a “good lead” is, no model will help you. Get sales and marketing in a room (yes, an actual meeting) and answer:
- Which leads do we want more of? Look at your last 6-12 months of closed/won deals. What do they have in common? Think company size, industry, tech stack, job function, country.
- Which leads waste our time? Be honest. Are there patterns with duds? (Example: “We never close deals from agencies under 10 people.”)
- What signals actually matter? Ignore vanity signals (like “opened an email”). Focus on things that predict purchase: company revenue, job seniority, tech used, etc.
Write these down. You’ll refer back to them constantly.
Step 4: Build Your First Madkudu Model
Madkudu gives you two main ways to build models: out-of-the-box (Fastlane) and custom (Pro). Here’s what works for most B2B teams:
Start with Fastlane (Out-of-the-Box)
- Let Madkudu analyze your historical data. It’ll spit out a first-pass model based on who converted in the past.
- Review the signals. Madkudu will highlight which factors (like company size, industry, location) seem to matter most. Don’t blindly trust this—gut check with your sales team.
- Set initial scoring thresholds. Madkudu usually uses a 3-tier system: “Low,” “Medium,” and “High” fit. You can rename these, but don’t get cute.
Or, Go Custom (Pro)
If your process is mature, or you want more control:
- Choose your signals. You can handpick which attributes go into the model (like company revenue, employee count, tech stack).
- Assign weights. Decide which attributes matter most. Don’t overweight tiny details—keep it simple.
- Exclude junk. Filter out lead sources or industries you never convert.
Honest take: Unless you have a ton of historical data and a data scientist on hand, the out-of-the-box model is good enough to start. You can always tweak later.
Step 5: Test and Tune (Don’t Skip This)
Here’s where most teams mess up: launching a model and never checking if it works.
- Run a backtest. Madkudu can show you how your model would have scored past leads. Look for obvious misses (like great customers rated “Low” or junk leads marked “High”).
- Spot-check with reps. Pull recent leads and ask your sales team if the scores feel right. If not, dig into why.
- Tweak thresholds and weights. If your model is flagging too many “High” fits, tighten it. If nothing is ever “High,” loosen up.
- Rinse and repeat. Revisit every couple of weeks at first, then monthly.
Don’t overfit. Trying to make your model perfect will just make it brittle. It’s better to have a “good enough” model you trust than a “perfect” one nobody uses.
Step 6: Push Scores to Your CRM and Use Them
A lead score only matters if sales can see and use it.
- Map Madkudu scores to CRM fields. Usually, you’ll create a custom field in Salesforce or HubSpot like “Madkudu Fit Score.”
- Add the field to lead/contact views. Make sure reps see it in their main list views and reports.
- Build simple workflows. For example:
- Assign “High Fit” leads directly to reps for follow-up.
- Route “Low Fit” leads to nurture sequences or disqualify.
- Train your team. Spend 30 minutes walking reps through what the score means—and more importantly, what it doesn't mean. This is a filter, not gospel.
What to ignore: Don’t build a Rube Goldberg machine of automations right away. Keep it simple. If reps ignore the score, don’t blame them—ask why.
Step 7: Keep It Honest—Monitor and Improve
Lead scoring isn’t “set and forget.”
- Watch conversion rates by score. If “High Fit” leads aren’t actually converting, something’s off.
- Ask for feedback. Set up a Slack channel or monthly check-in for reps to call out weird results.
- Adjust as you learn. If your ICP changes, or you launch in a new vertical, revisit your model.
Warning: Don’t chase every false negative or positive. Look for patterns, not one-offs.
Pro Tips and Common Pitfalls
- Don’t overcomplicate. Every extra signal or rule adds noise. Start small.
- Ignore the “black box” hype. Madkudu uses machine learning, but you should always sanity-check its logic against your real-world experience.
- Don’t expect 100% accuracy. A score is a guide, not a guarantee. Some great leads will slip through, and that’s normal.
- Document everything. When you change something, write down why. You’ll thank yourself later.
Wrapping Up: Start Simple, Iterate Often
Lead scoring in Madkudu can save your team a ton of time—but only if you keep it simple and tune as you go. Don’t wait for perfect data or a flawless model. Launch, listen to your reps, and tweak as you learn. It’s better to have a basic, working filter than another fancy tool nobody trusts. Good luck—and remember, your sales team’s feedback is as important as the algorithm.