Most B2B lead scoring models die the same death: marketing builds one in isolation, sets arbitrary point values, and hands sales a queue of 'hot' leads that turn out to be a junior analyst from a company with 8 employees. Sales ignores the queue within a month, and the whole exercise becomes a reporting vanity metric. The fix is not a better algorithm - it is a shared definition of what a qualified lead actually looks like, built with sales input before a single point value is assigned.
Why Most Scoring Models Break at the Sales Handoff
The core problem is that marketing scores on engagement - email opens, page views, content downloads - while sales qualifies on fit and intent. A prospect who reads three blog posts and downloads a whitepaper scores 85 points in most CRMs, but if they work in a 10-person company outside your ICP, they will never buy. Engagement signals measure curiosity, not purchase readiness. Conflating the two is what fills your pipeline with leads that look warm on a dashboard but go cold the moment a rep dials them.
According to Gartner's B2B buying research, the average enterprise purchase now involves 6 to 10 decision-makers, meaning a single engaged contact rarely signals real buying intent on its own. A scoring model that ignores account-level fit and buying group composition will consistently send sales after the wrong person, even when that person is genuinely interested.
The output is a trust gap. Once sales reps learn that a 'marketing qualified lead' regularly converts to a closed deal at under 5%, they stop working the queue and revert to self-sourced pipeline. At that point, your entire demand generation operation is running on faith rather than evidence.
The Two Dimensions Every B2B Score Needs
A functional model scores on two separate axes: fit and engagement. Fit is firmographic and technographic - company size, industry, geography, revenue band, tech stack, and job title of the contact. Engagement is behavioral - pages visited, time on site, pricing page views, demo requests, email click-throughs. Neither axis alone is sufficient. A perfect-fit prospect who has never engaged with your brand needs nurture, not a sales call. A highly engaged prospect who is outside your ICP needs to be recycled or disqualified, not handed to a senior rep.
Assign separate scores to each dimension and combine them into a two-by-two matrix rather than a single composite number. A lead that scores high on both axes is a true MQL. A lead that scores high on fit but low on engagement goes into a nurture track. A lead that scores high on engagement but low on fit gets a shorter nurture cycle with explicit disqualification criteria. This structure gives sales a legible signal rather than an opaque number, which is what makes them trust it.
Building the Fit Score: Firmographics First
Start by pulling the last 24 months of closed-won deals from your CRM and looking for the attributes they share. Do not guess at your ICP - mine the actual data. You will typically find that 60-70% of closed revenue clusters around 3-5 specific industry verticals, a narrow employee count range, and a handful of job titles in the buying group. Those clusters become your tier-one fit criteria, worth the most points in your model.
A simple starting framework for fit scoring might look like this:
- Company in a tier-one industry vertical: 20 points
- Employee count within ICP range (e.g. 200-2,000): 15 points
- Contact is a director-level or above decision-maker: 15 points
- Company uses a technology in your integration ecosystem: 10 points
- Geography matches a territory you actively sell into: 10 points
- Company in a tier-two adjacent vertical: 8 points
- Contact is an individual contributor or unconfirmed role: 2 points
Cap the fit score at 70 points and make it the dominant input into your combined model. If a lead cannot reach at least 40 fit points, it should never be routed to sales regardless of how much content they consume. This single rule eliminates most of the junk that erodes sales confidence.
Building the Engagement Score: Weight by Intent Proximity
Engagement points should scale with how close a behavior is to a purchase decision. Visiting the pricing page once is worth far more than downloading a top-of-funnel ebook, even though the ebook download looks more impressive in a weekly marketing report. Assign the highest engagement points to actions that indicate active vendor evaluation: pricing page views, demo requests, comparison page visits, contact form views, and return visits within a short window.
A mid-tier engagement tier covers behaviors that suggest real interest but not yet buying mode: watching a product webinar past the 50% mark, visiting a case study page, or clicking a feature-specific email. Bottom-tier behaviors - blog reads, social clicks, generic newsletter opens - should receive minimal points, often 1-2 each, and should have a decay rate applied so that a contact who engaged six months ago and has since gone dark does not retain a high score indefinitely. Point decay is one of the most commonly skipped steps in scoring model builds, and its absence is a major reason queues get polluted over time.
If you are running intent signals alongside your behavioral data, third-party intent scores from providers like Bombora or G2 can be incorporated as a multiplier on your engagement score rather than as a standalone input. A contact who scores 30 engagement points but whose company is showing active category intent on a third-party platform is meaningfully more valuable than a contact with 30 engagement points and no external signal.
Getting Sales to Co-Own the Model
Run a working session with 3-4 sales reps and your sales manager before you finalize any point values. Show them the fit criteria you have drafted, ask them to rate each attribute by how strongly it predicts a deal closing, and adjust your weightings accordingly. This step takes two hours and produces two outcomes: a more accurate model because reps know things about buyers that are not in your CRM, and genuine buy-in because the team helped build it rather than receiving it as a directive from marketing.
Set a formal 60-day review cadence after launch. Pull every lead that was handed to sales, track what happened to it, and run a win/loss analysis by score band. You will almost always find that your initial thresholds need adjustment - the MQL cutoff is often 10-15 points too low in early iterations. Revisiting the model on a fixed schedule, rather than only when someone complains, keeps it calibrated and keeps sales engaged with the process. This kind of systematic review is the same principle that drives a strong growth audit for stalled pipeline, applied at the lead qualification layer.
Connecting Scoring to Your Lead Generation Channels
A scoring model is only as useful as the data feeding it. If your paid channels are pulling in contacts who cannot reach a minimum fit score, you have a targeting problem upstream, not a scoring problem. Review your channel mix and check whether the firmographic profile of inbound leads matches your ICP at the account level. If LinkedIn campaigns are consistently delivering contacts from out-of-ICP companies, tighten your company size filters and job function targeting before blaming the scoring model for low MQL rates.
Our B2B growth audit service regularly surfaces misalignment between targeting parameters and ICP fit as the primary driver of low-quality pipeline - not channel performance in isolation. If your lead volume looks healthy but your MQL-to-SQL rate sits below 20%, the issue is almost always that acquisition and qualification were never designed around the same definition of a good lead.
One practical fix: add a hidden field to your lead forms that auto-populates with the UTM source and campaign name, then map that field into your CRM. Within 90 days you will have enough data to calculate a fit-weighted lead quality score by channel, which tells you exactly which campaigns to scale and which to cut. This approach is especially useful if you are already investigating why lead form clicks are not producing qualified leads despite reasonable volume.