Key takeaways
- A B2B lead scoring model should rank leads on fit and qualify them on a hand-raise. Activity alone should never add up to an MQL.
- At Pipe17 the old model made anyone over 100 points an MQL and gave a page view the same 20 points as a demo request. About 70% of MQLs had never asked to talk to anyone.
- I made hand-raisers the only hot MQLs, agreed 3 tiers with Sales and closed the routing gaps. MQL to SQL went from about 20% to 40%.
- Test the model against the leads Sales rejected, and check response speed before you blame the score.
01What a B2B lead scoring model should do
A B2B lead scoring model decides which leads go to Sales, how fast and who works them. A good one uses fit to decide whether a lead deserves a rep's time and a direct request, like a demo form, to decide when. Page views and email opens can order the nurture queue, and that's all they should do.
This note is about designing the model. The metric it moves, and how to measure it, is in my note on MQL to SQL conversion rate.
02Why the 100-point model failed
At Pipe17, more than 60% of inbound leads were small businesses that would never buy, and the score explained why. An MQL was anyone who crossed 100 points. A page view was worth 20 points, the same as a demo request.
So a busy visitor who would never buy crossed the line before a good-fit buyer who had visited once. About 70% of MQLs got there on engagement alone and had never asked to talk to anyone. When most of the queue is people browsing, Sales has no reason to trust the label.
The flaw is in the adding. When every signal goes into one total, enough weak signals equal one strong one. The fix is to give each kind of signal its own job.
| Signal | Examples | What it should decide | The usual mistake |
|---|---|---|---|
| Fit | Company size, company type, platform | Whether a rep should spend time on the lead at all | The fields are blank, so fit can't count |
| Hand-raise | Demo request, contact form | Whether the lead is a hot MQL today | Scored the same as a page view |
| Engagement | Page views, email opens, downloads | The order of the nurture queue and what to send next | Added up until it crosses the MQL line |
| Account intent | Intent scores on the account | Which accounts move up a tier for outreach | Kept in a tool reps never open |
| Disqualifiers | Too small, existing customer | Who never reaches a rep | Left for reps to reject one at a time |
03How I design the model
- Read the rejected leads first.The reasons Sales gives for disqualifying leads tell you what fit means. At Pipe17 the top reason was company size.
- Make hand-raisers the only hot MQLs.A hot MQL became a demo request or a contact form and nothing else. Everyone else went to nurture.
- Agree tiers with Sales, one owner each.We settled on 3. Tier 1 is hand-raisers and accounts showing intent, worked by account executives. Tier 2 gets SDR outreach. Tier 3 gets marketing nurture.
- Match outreach to the ads.Four outbound segments (3PLs, brand aggregators, Shopify Plus brands and NetSuite users) each got messaging that matched the ads those accounts were seeing.
- Close the routing gaps.Every new contact without an owner went to the account's owner if there was one, and round-robin across the sales team if there wasn't. A tier only works if every lead in it has an owner.
- Put intent where reps work.I synced account intent scores into HubSpot as 12 contact and company properties, so reps could see account intent without leaving the CRM.
- Qualify before the form.Since company size was the top rejection reason, every search ad said who the product is for: high-volume brands. A company that's too small can rule itself out before it costs a click and a rep's time.
MQL to SQL conversion went from about 20% to 40%.
04How to test the model
Audit the rejected pile
A scoring model can be wrong in both directions. It can send junk to Sales, and it can let good leads get rejected. To check the second, I joined 820 recorded sales calls to 999 disqualified leads. 94 of those leads had a call on record, and 35 records across 30 companies turned out to be worth another look.
Check response speed before you blame the score
31 of 95 disqualified paid leads were marked no-response. Nobody reached them, so nobody knows whether they fit. That's a speed-to-lead problem as much as a scoring or media problem, and tightening the score won't fix it.
The same check applies to whole channels. When I audited LinkedIn, 42% of its leads had never had an owner assigned. The details are in are LinkedIn Ads worth it for B2B SaaS.
05Mistakes I'd avoid
- Letting activity add up to an MQL. Engagement is for nurture order. A lead who hasn't asked to talk isn't ready for a rep, however many pages they've read.
- Scoring fit on empty fields. At Pipe17, 13,316 companies, half the database, had no company type, which blocked intent scoring and tiering. I wrote a HubSpot API script that classified them by industry. A full pull also showed 1,974 good-fit brands sitting in the lowest tier, only 13 of them with any intent score.
- Counting a list upload as a hand-raise. A scanned badge at a trade show is not an MQL. I keep event contacts on their own track, described in trade show lead generation.
- Building it without Sales. The tiers worked because Sales agreed who owned each one. I wouldn't ship a model Sales hasn't signed off on.
- Never looking at the rejects. Without an audit you only ever see the model's wins.
Once the model holds, send the qualified stages back to your ad platforms so bidding learns from them. That's covered in value-based bidding. Rebuilding scoring and routing with a sales team is a core part of what I do as a B2B demand generation consultant.
FAQQuestions people ask
- What is a B2B lead scoring model?
- A B2B lead scoring model is the set of rules that decides which leads go to Sales, how fast and who works them. It usually combines fit, such as company size and type, with behavior, such as a demo request. The useful ones are agreed with Sales and tested against the leads Sales rejected.
- Why do point-based lead scoring models fail?
- They let activity add up to a qualification. At Pipe17, an MQL was anyone who crossed 100 points, and a page view was worth the same 20 points as a demo request. Farhan Munshi found that about 70% of MQLs got there on engagement alone and had never asked to talk to anyone, so the label told Sales very little.
- What should count as an MQL?
- A lead that fits and has asked to talk. When Farhan Munshi rebuilt scoring at Pipe17, a hot MQL became a demo request or a contact form and nothing else. Everyone else went to nurture. MQL to SQL conversion went from about 20% to 40%, because reps were only handed people who had raised a hand.
- How many lead tiers should a scoring model have?
- Few enough that each tier has one owner. At Pipe17, Farhan Munshi agreed three tiers with Sales. Tier 1 was hand-raisers and accounts showing intent, worked by account executives. Tier 2 got SDR outreach. Tier 3 got marketing nurture. If two tiers get the same follow-up, they're one tier.
- How do you know if a lead scoring model is working?
- Audit the rejected pile and check response speed. At Pipe17, Farhan Munshi joined 820 recorded sales calls to 999 disqualified leads and found 35 records across 30 companies worth another look. He also found 31 of 95 disqualified paid leads were marked no-response, which points to follow-up speed as much as the score.
