Update: this article includes historical references or examples of customer-defined rules. The current ByeBouncer service checks each field and returns independent results, without scoring or a contact acceptance decision. We do not establish identity or ownership, or detect bots or fraud. See /en/docs#validation-v1 for the current contract.
B2B lead scoring means giving every lead a score so you can decide who sales calls today, who marketing nurtures and who gets dropped. Most models we see have two layers: how well the lead fits your ideal customer (fit) and how interested they look (intent). They miss a third one, duller and cheaper to measure: whether that contact exists and can actually be reached.
Without that layer, the model rewards leads you cannot reach. A disposable email with a 'Director' title and three pricing page visits scores high and will never answer you. Here is a simple scheme, limits included, that you can build in a spreadsheet or your CRM.
The three layers of B2B lead scoring
- Fit (profile): job title, company size, industry, country. It is what the form says or what you enrich later. It changes slowly.
- Intent (behavior): pricing visits, demo requests, downloads, email replies. It changes fast and decays within days.
- Contact quality: does the email accept mail, is the phone a real line, did the signup come from a risky IP? It tells you whether the lead is reachable, not whether it is good.
How to build the model, step by step
- Define a 'good' lead with your own data: look at your last 20 to 50 won deals and what they had in common before buying.
- Pick 5 to 8 fit attributes and weight them by how much they show up in those wins. No more: every extra rule is one nobody maintains.
- Pick 4 to 6 intent actions and weight them by closeness to purchase. A demo request is worth far more than opening a newsletter.
- Add disqualifier rules with negative points or a hard block: no buying authority, competitor, hard bounce, unsubscribe.
- Set thresholds and test them against your history before going live.
The thresholds you see in industry guides are illustrative, not standards. A common example is 80 or more for 'hot' (call today), 40 to 80 for 'warm' (nurture by email) and under 40 for 'cold'. Use them as a starting point and recalibrate against your real conversion rate: with a high ticket and few leads, your 'hot' line may start much higher.
The missing layer: contact quality in your score
This is where the score gets honest. Instead of treating an invalid email as one more data point, treat it as a condition: if the lead is unreachable, the rest of the score matters little. A baseline scheme that works for us:
- Email with valid format, a domain with MX records and not disposable: no penalty.
- Disposable email or domain without MX: block, it does not enter the sales flow.
- Catch-all or uncertain email: reduced score, not zero. No verifier can promise you that mailbox exists.
- Phone with correct format, identified carrier and an active line per HLR: adds points. Disconnected line or impossible format: strong deduction.
- Signup from an IP linked to a proxy, anonymous VPN or poor reputation: deduction, ideally combined with other signals before blocking.
Notice what this scheme does not claim. A verified email is not an interested lead, and an active phone is not a phone that picks up. Contact quality answers one question only: is it worth spending sales time trying to reach this person? And a VPN IP is not always fraud: plenty of legitimate people browse that way, especially behind corporate networks.
A worked example
Take two leads at a software company. Lead A: operations manager at a mid-sized company, requested a demo, valid corporate email, active mobile number. Lead B: 'Director' at a large company, visited pricing four times, but signed up with a disposable email from a proxy IP.
With a fit-and-intent-only model, B can score the same as A or higher. With the contact layer, B ends up blocked or far down the list: the disposable email takes it out of the sales flow and that rep's time goes to A. It is not magic, it is just not spending calls on contacts you already know you cannot reach.
Common mistakes when qualifying B2B leads
- Rewarding intent without checking the lead is reachable (the example above).
- Adding points with no ceiling: a lead who opens 30 emails is not 30 times more ready.
- Not decaying the score: intent from three months ago is not worth the same. Subtract points for inactivity.
- Building the model in marketing alone. Sales has to validate what a worthwhile lead is, or nobody trusts the number.
- Jumping to a predictive model without enough history. With few closed deals, a machine learning model mostly memorizes noise; start with manual points and migrate when you have the data.
Where ByeBouncer fits
In production we verify email, phone and IP and return the signals per lead so you can use them as one more layer in your score, in your CRM or via API. We do not decide for you what a good lead is: your won deals define that. What we give you is the part that does not depend on opinion, whether the contact exists and is reachable, with the limits each signal has (a catch-all stays uncertain, and HLR confirms a line, not that someone answers).
See how we score lead quality →
If you want to test the contact layer on your own data before redesigning anything, take a small batch and compare it against your won deals. If your wins barely contain disposable contacts, you already know how much of your current pipeline is noise.