Lead scoring is the practice of assigning each lead a number that predicts how likely it is to become a customer, from fit attributes such as company size and industry, intent signals such as pricing page visits and replies, and product usage where a free tier exists. The score sets the order sales works leads in. A score that ignores whether the contact channel works ranks dead records alongside live ones; applying reachability as a multiplier, zero for invalid and half for catch-all or unknown, fixes that.
Two families of model share the same inputs. Points models assign fixed weights by hand: a senior title is worth 10, a pricing-page visit 8, an invited colleague 20. They are easy to explain and easy to change, and they are the right start for any team with fewer than a few thousand outcomes. Predictive models fit the weights to historical conversions with logistic regression or a tree model and tend to beat hand-set weights once there are enough closed-won and closed-lost records to learn from, usually two quarters of data.
Inputs fall into three groups. Fit is who the lead is: company size, industry, country, technology stack, role seniority, usually from enrichment. Intent is what the lead did: form fills, demo requests, replies, pricing-page views, documentation reads, from the CRM and web analytics. Usage, for products with a free tier or trial, is what the lead did inside the product: a second user invited, an integration connected, a plan limit hit. Product-led teams weight usage highest, often 50% of the points; sales-led teams weight fit and intent and may have no usage signal at all.
Three refinements separate a working model from a spreadsheet. Negative scoring subtracts points for disqualifiers: a student email domain, a competitor, a country you do not serve, a job title with no buying authority. Score decay reduces intent points as they age, for example halving a pricing-page visit after 30 days, so last quarter's curiosity does not outrank this week's demo request. Reachability is applied last and as a multiplier, not as points: 1.0 when the email address verified valid, 0.5 when it returned catch-all or unknown, 0 when it is invalid, disposable or role-based. The multiplier is BounceZero's recommended framing; the SaaS lead scoring guide works a full 100-point model through with these factors and shows how to calibrate the weights against closed deals.
Thresholds turn the score into hand-offs. A common 100-point layout is 0 to 29 nurture only, 30 to 59 marketing qualified, 60 to 79 sales qualified candidate with a call inside 24 hours, 80 and up sales-ready with an opportunity created. Set the cut-offs from your own conversion data: the marketing-qualified band should capture roughly a quarter of leads and at least two thirds of the deals that eventually close.
Two leads arrive the same morning. A 150-person SaaS company: head of growth viewed pricing twice (8 points), company size in range (10), industry in a top vertical (7), no product usage yet; fit and intent total 25, scaled to the model's 100-point range it sits at 85 on a sales-led scale. Its email returned catch-all, so the multiplier is 0.5 and the final score is 42. A 40-person agency: founder requested a demo (12), company slightly under the size band (0), industry match (7); fit and intent sit at 70 on the same scale, email confirmed valid, multiplier 1.0, final score 70. The demo request is worked first, which is the right order, and the catch-all contact waits for a verified address rather than consuming a rep's call slot.
Why the multiplier matters at scale: across 6.5 million business addresses verified by BounceZero between April and October 2026, 14.5% were invalid and 27% could not be confirmed. A model that scores every record as reachable spends roughly one in seven of its top-band calls on a dead address.
Verify a list or a single address. 100 free checks a month, unknowns refunded.
Verify Leads - FreeAyoub built BounceZero's 5-stage validation pipeline, its dedicated BGP-announced IP infrastructure, and the Patroni HA PostgreSQL cluster behind every verification. Previously built high-volume email delivery infrastructure. Trained at 1337 Benguerir (École 42 network, 2019). Open-source: bgp_analyzer.