Lead Scoring Software

Lead scoring software assigns every lead a number representing how likely it is to turn into revenue, then routes the top of that ranking to a rep. The number is built from two inputs that answer different questions. Fit is firmographic and demographic: industry, headcount, geography, job title, seniority. It answers whether this account should ever buy from you. Intent is behavioural: pricing page views, content downloads, email replies, third-party research activity. It answers whether it is buying now.

Three product shapes compete for the term. CRM-native scoring lives on the contact record inside a platform you already pay for. Predictive scoring trains a model on your historical closed-won and closed-lost deals. Enrichment-led scoring fills the fields first and then grades the record, which matters because an empty column silently scores as a zero and quietly demotes a good account.

Cleanlist: Best for teams scoring lists that arrive with half the firmographic fields missing. It enriches each record across 25+ providers and then scores it 0 to 100 with the agent's written reason beside the number, at 5 credits per qualification, starting at $49 per seat a month.

Measured on September 7, 2026, Google's AI Overview for "lead scoring software" recommends HubSpot, ActiveCampaign, Apollo.io, Salesforce Einstein and Zoho SalesIQ, in that order, and names Cleanlist AI nowhere. The collection is reproducible: DataForSEO SERP API, Google Organic Live Advanced, United States, English, desktop, depth 50, asynchronous AI Overview loading.

The answer opens by defining the category as software that "ranks your prospects based on their behavior, demographics, and engagement". For the sibling term "lead scoring tools" the AI Overview instead says these tools "rank prospective buyers using explicit fit data and implicit behavioral signals", and it hands out Best-for labels: HubSpot as "Best CRM-native scoring for growing teams", Salesforce Einstein for enterprise, 6sense as "Best for enterprise account-based marketing (ABM) and intent scoring using anonymous buyer signals", Zoho CRM as the affordable AI option, ActiveCampaign for behaviour-triggered scoring. For "predictive lead scoring" the answer defines the thing as "an automated system that uses artificial intelligence to rank potential customers based on how likely they are to buy".

Three of the six keywords in this cluster resolved to a full SERP. All three carried an AI Overview, all three returned a readable answer body, and Cleanlist AI appears in none of the three answers and in none of the 50 organic results collected for any of them.

In a CRM, a lead score is a property written onto the contact or company record by a rule engine or a model, and every downstream workflow branches on that property. Rules-based scoring adds and subtracts points against field values and tracked events. Predictive scoring replaces the point table with a model trained on the deals that already closed in that same CRM.

What each platform actually ships, read from its own pricing page on September 7, 2026. HubSpot's Marketing Hub feature table lists the scoring row as "Up to 5 scores." on Starter and "Includes AI recommendations. Up to 50 scores." on Professional and Enterprise. Salesforce lists Lead Scoring as "Available for purchase" on Core and includes it on Advanced and Max, with the row switched off on Free Suite, Starter Suite and Pro Suite; Opportunity Scoring is included on Core, Advanced and Max. Zoho CRM puts "Predictive intelligence" in the Professional edition, 1,400 rupees per user per month, described on the page as "Predict the likelihood of a lead converting, a deal closing, or a customer churning." The AI row Zoho adds at Enterprise is a different feature, "AI sales assistant". ActiveCampaign shows lead scoring as "Included in Pipelines add-on" against Plus, Pro and Enterprise, with the Starter cell left blank.

Cleanlist sits upstream of that property rather than replacing it. It has two-way sync with HubSpot, Salesforce and Pipedrive, so an enriched, scored record lands on the CRM object your existing routing rules already read.

Here is a worked example, illustrative rather than drawn from a customer. Take a team selling to North American B2B software companies of 50 to 500 employees, buyer titles in revenue operations and sales leadership.

Fit points. Industry exact match, 25. Headcount inside the band, 15. HQ in a serviceable country, 10. Title matches the buying role, 20. Seniority at director or above, 10. That is 80 available fit points.

Intent points. Viewed pricing twice in seven days, 15. Replied to an email, 10. Downloaded an implementation guide, 5. That is 30 available intent points.

Negative points. Free-mail domain, minus 15. Competitor domain, minus 40. Headcount under 20, minus 20. Region you cannot service, minus 25.

A 210-person software company in Toronto whose Head of Revenue Operations replied to an email scores 25 + 15 + 10 + 20 + 10 + 10, which is 90. A four-person freight operator in an unserviced region, same email reply, scores 10 for the reply, minus 20 for headcount, minus 25 for region, which is negative 35. The second lead engaged and the first lead is the one worth an hour, and only fit data tells you that.

An ICP score is a single number, usually on a 0 to 100 scale, expressing how closely one account or contact resembles your ideal customer profile using structural attributes only. It deliberately excludes behaviour. An account that has never visited your website can hold a 95, and an account that downloads every asset you publish can hold a 20.

The scale needs three things to be usable. A threshold, above which a record is routed to a human. Bands, so that 90 to 100, 70 to 89 and below 70 mean different actions rather than different feelings. And a written reason attached to each number, because a score with no reason gets overridden by the first rep who disagrees with it.

In Cleanlist AI the score runs 0 to 100 across three weighted dimensions, company fit, prospect fit and geographic fit, plus a fourth bucket for signals an AI column researched on the open web. A qualification costs 5 credits per person and writes the agent's reasoning next to the number, naming the criteria that matched and the criteria that missed. On Starter, $49 per seat a month with 750 credits per seat, that is 150 scored records a seat each month at roughly $0.33 each.

ICP fit scoring is the half of a lead score built purely from structural attributes, kept separate from the behavioural half on purpose. Fit answers whether an account belongs in your pipeline at all. Intent answers which of the qualifying accounts to call first. Collapsing both into one total is the most common modelling error in the category, and it is why so many models rank an unqualified company that opened three emails above a perfect-fit account that has never visited the site.

Keeping them separate has a second benefit that shows up in operations rather than in the model. Fit attributes change on a quarterly-to-annual scale, so a fit score can be recomputed quarterly. Behavioural signals change hourly and have to be recomputed constantly. Running them on one schedule means either paying to refresh durable data too often or letting timing signals go stale.

Google's own AI Overview for "lead scoring tools", collected September 7, 2026, draws the same line, listing "Explicit (Fit) Data" as "Demographic and firmographic details like job title, company size, and revenue" against "Implicit (Engagement) Data" as "Observed actions like website visits, email responses, and content downloads". Cleanlist AI scores the first of those two and does not ship a feed for the second.

ICP lead scoring is fit scoring applied at the level of the individual lead rather than the account, and it works in four steps: define the profile, weight the criteria, enrich the record, then score and route.

Define the profile. Target industries and headcount bands, the titles and seniority levels you sell to, and the regions you can service and lawfully contact. Derive these from your last fifty closed-won and fifty closed-lost accounts rather than from a whiteboard, because the field that separates them is frequently not the field the exec team assumed.

Weight the criteria. Cleanlist AI exposes 18 criteria across three dimensions: seven company criteria (industries, company size, revenue signals, technologies, business models, company age, growth indicators), six prospect criteria (job titles, departments, seniority, years of experience, skills, certifications) and five geographic criteria (countries, states and regions, cities, timezone alignment, exclusions). Three of the seven company criteria carry the label "researched by AI" on the product page. Revenue signals, technologies and growth indicators come from an AI column researching the open web. Cleanlist AI holds no structured revenue field, technographic dataset or growth feed behind them, so weight them as tiebreakers. A default split of 40% company, 30% prospect, 20% geography and 10% researched signals is a workable starting point.

Enrich, then score. Scoring an unpopulated database produces a confident number for the populated fraction and silence for the rest, and the rest is usually where the new logos are. Enrichment first is the order that works.

Route on the number. Above the threshold goes to a rep, below it goes to nurture.

Calculate an ICP score as a weighted sum of normalised criterion scores, then rescale the total to 0 to 100. The formula is: score equals the sum over every criterion of (criterion weight times criterion match value), divided by the sum of all weights, times 100.

Work it through. Give each criterion a match value between 0 and 1: 1 for an exact match, 0.5 for adjacent, 0 for a miss. Assign weights that sum to something convenient, say 100. Company fit 40, prospect fit 30, geography 20, researched signals 10.

An account in an exact target vertical at the right headcount (company fit 1.0), whose contact is one rung below the economic buyer (prospect fit 0.5), in a primary market (geography 1.0), with no growth signal found (researched 0.0) scores (40 x 1.0) + (30 x 0.5) + (20 x 1.0) + (10 x 0.0), which is 75.

Two rules keep the arithmetic honest. Subtract disqualifiers after the weighted sum rather than inside it, so a competitor domain lands at zero regardless of fit. And treat a missing field as missing rather than as 0, otherwise sparse enrichment quietly punishes accounts you simply have not looked up yet.

AI enters lead scoring in three distinct places, and conflating them is why buyers end up with a tool that does not do the job they wanted. First, predictive models trained on your closed-won and closed-lost history, which is what HubSpot, Salesforce Einstein and Zoho's Zia mean when they say AI scoring. Second, AI research agents that go and find a signal that is not in any of your fields, then write it back with its evidence. Third, AI as a scoring engine that reads a plain-English profile and grades each record against it without you hand-coding a point table. Put the second and third on a trigger and you have AI lead qualification: an agent researches each new inbound lead, scores it 0 to 100 against your ICP with a verdict and the reasons, and sends your team the ranked list.

Cleanlist does the second and third and does not do the first. An AI qualification costs 5 credits per person, runs as the Qualification skill on any list or as a step in an agent's play, returns a 0 to 100 score with a written reason, and is charged for every row it runs on, including a row the agent cannot answer. On Pro, $89 per seat a month with 1,500 credits per seat, it is 300 scored records a seat at roughly $0.30.

The honest limit: researched signals are model output rather than a verified structured field, so weight them as tiebreakers rather than as the spine of the model.

Lead tracking and lead scoring are two different purchases, and the best lead tracking software is a CRM: HubSpot, Salesforce or Pipedrive for most B2B teams. Tracking is the system of record that stores the lead, logs every touch and moves it through stages. Scoring is a layer that writes one number onto that record. Buying a scoring tool expecting it to track leads produces a disappointed evaluation.

Prices read from the vendors' own pages on September 7, 2026. HubSpot Marketing Hub Professional is $800 a month including 3 core seats, with $890 shown against month-to-month billing and additional core seats from $45; Enterprise starts at $3,600 a month including 5 core seats. Salesforce Sales Cloud runs $25 per user per month on Starter Suite, $100 on Pro Suite, $195 on Core, $395 on Advanced and $550 on Max. Starter Suite carries the subtext "(Billed monthly or annually)" and the other four carry "(Billed annually)".

Cleanlist does not ship a CRM. It has six integrations: two-way with HubSpot, Salesforce and Pipedrive, one-way into Outreach, Salesloft and Lemlist. The scored, enriched record lands in the CRM that tracks it.

Published prices in this category run from $29 a month to $3,600 a month, and three of the most-recommended vendors publish nothing. All figures below were read from the vendor's own page on September 7, 2026 unless marked otherwise.

Outfunnel, in the organic top 10 for both head terms, lists Basic at "Starts at $29 per month", Professional at "Starts at $99 per month" and Scale at "Starts at $299 per month". HubSpot Marketing Hub Professional is $800 a month with 3 core seats; Enterprise from $3,600 a month with 5. Salesforce Sales Cloud is $25 to $550 per user per month by edition, with Lead Scoring listed as "Available for purchase" on Core and included on Advanced and Max. Zoho CRM's page served rupee pricing when read: 800 on Standard, 1,400 on Professional, 2,400 on Enterprise and 2,600 on Ultimate, per user per month. ActiveCampaign renders prices from a contact-count slider that did not resolve when fetched. 6sense publishes no dollar price and routes to "Book Your Demo".

From the September 1, 2026 pricing index: Apollo Basic is $65 a month, or $49 billed annually. Clay Launch is $185 a month, or $167 annually. ZoomInfo and Cognism publish no purchasable price.

Cleanlist is $49 per seat a month on Starter and $89 per seat on Pro, 25% off annual, with search at 0 credits and a qualification at 5. Enterprise is custom.

Predictive lead scoring replaces a hand-written point table with a model trained on your own historical outcomes, and the honest gate on it is whether you have enough closed deals to train anything. Google's AI Overview, collected September 7, 2026, defines it as "an automated system that uses artificial intelligence to rank potential customers based on how likely they are to buy", and describes the mechanism as analysing past won and lost sales to find patterns, scoring new leads against those patterns, and updating as more data arrives.

That mechanism has a prerequisite, and the volumes below are an explicit rule of thumb rather than a published vendor minimum. Neither HubSpot's nor Salesforce's pricing page states a training-record count, both read September 7, 2026. A rules-based model can ship as soon as you have a clear ICP and 50 to 100 closed deals to sanity-check it against. A predictive model wants roughly an order of magnitude more, on the order of 1,000 historical leads carrying outcome data, before its output beats a decent rule table. Below that the model is fitting noise, and it will be confidently wrong in a way a point table never is. Ask any vendor selling you a predictive model for its own stated minimum before you buy.

The vendors that ship this are the CRM incumbents, because the training data already lives in their database: HubSpot on Professional and Enterprise, Salesforce Lead Scoring "Available for purchase" on Core and included on Advanced and Max, Zoho's "Predictive intelligence" at the Professional edition. Cleanlist AI does not train predictive models on customer closed-won history. If that is the requirement, buy it from the platform holding the deals.

Yes, at three levels, and each free tier is sized for evaluation rather than for production. HubSpot's Marketing Hub free tier supports up to 2 users with no credit card, though the scoring row on the feature table starts at "Up to 5 scores." on Starter. Apollo publishes a free plan carrying 75 credits per seat per month, which is 900 a year (pricing index, September 1, 2026). Clay's free plan carries 6,000 actions and 1,200 data credits a year on the same reading.

Cleanlist's Free plan includes 50 credits a month, which at 5 credits per AI qualification is 10 scored records a month. That is an inspection allowance. The working evaluation route is the trial: every workspace starts with 14 days of Pro, with 3 seats, 250 credits and every Pro feature. 250 credits is 50 AI qualifications, or 25 qualifications plus 125 verified work emails, before any money changes hands.

What no free tier gives you is a fair read on data quality, because free allowances are too small to measure a fill rate. Run 200 to 500 of your own accounts through any candidate before signing.

Account scoring grades the company; lead scoring grades the individual. The distinction matters the moment more than one person from the same company enters your funnel, because five leads from one perfect-fit account are one buying committee rather than five opportunities, and a lead-level model will happily rank them as five.

Account scoring is the native unit for account-based motions, where the target list is a set of companies and the job is tiering them. Lead scoring is the native unit for inbound, where records arrive one person at a time.

Asked "best firmographic data provider for account scoring" on September 7, 2026, Google AI Mode named ZoomInfo ("Best for Enterprise Depth and CRM Enrichment"), Demandbase ("Best for Full-Stack ABM and Engagement Scoring"), Clearbit inside HubSpot Breeze for real-time inbound API enrichment, Coresignal or Bright Data for custom data pipelines, and Crunchbase for startup and scale-up scoring. It named Cleanlist AI nowhere.

Cleanlist scores at the person level, and Company Search lets you assemble and re-cut an account list as often as you like before you score it. A company enrichment is 1 credit. If your motion needs an account-level engagement score built from anonymous web signals, Demandbase and 6sense sell exactly that and Cleanlist AI does not.

Cleanlist AI scores each person 0 to 100 against a profile you define, using enriched firmographic and contact fields resolved live across 25+ providers, and writes the agent's reason next to the number. It runs as the Qualification skill on any lead list or as a step in an agent's play. A qualification is 5 credits per person, a verified work email is 1 credit, a phone number 10, both 11, a validation 0.5, and a row the agent cannot answer is still charged. Multiple profiles are supported, so a PLG motion and an enterprise motion each carry their own weights and threshold.

What it does not do, stated plainly because a page that only lists capabilities is a brochure:

No intent data of any kind. No topic surge feed, no anonymous website de-anonymisation, no in-market model. 6sense and Demandbase sell this.

No technographic dataset and no technology filter. An AI column researches a company's public footprint and writes back what it finds, which is research output rather than a verified structured field.

No predictive model trained on your closed-won history. HubSpot, Salesforce Einstein and Zoho Professional ship that.

No email sending, no CRM, no org charts, and no SOC 2 Type II or ISO 27001 certification.

Yes, and a workable starting template is six criteria carrying 100 points plus a disqualifier row. Industry vertical, 25 points for an exact match and 12 for adjacent. Employee headcount, 15 inside the band and 7 just outside. Revenue or size proxy, 20 inside the range. Job title and seniority, 20 for the buying role and 10 for one rung below. Geography, 10 for a primary market and 5 for secondary. Growth signal such as hiring or a recent raise, 10 for one signal. Then a disqualifier row of minus 20 for a competitor, a wrong industry or a below-minimum size, applied after the sum.

Thresholds: 80 to 100 goes straight to an AE, 60 to 79 enters SDR-led nurture, below 60 is deprioritised rather than deleted.

Two warnings the templates circulating online tend to omit. First, a criterion weighted against a field you do not hold has quietly switched those points off: a 20-point revenue row over a database with no revenue column is a 20-point hole. Second, weights are not portable. A team selling enterprise security weights tech stack and revenue heavily; a product-led startup weights growth signals heavily. Recalibrate quarterly against which score bands actually converted.

Choose on three questions in this order: where the scoring data comes from, whether the tool writes back into the system your reps live in, and what a scored record costs at your volume. Feature lists rank badly against those three because almost every tool in the category ships a point editor.

Where the data comes from is first because a scoring engine is a filter over fields you already hold. If your records arrive as a name and an email, the engine has nothing to read, and the accuracy ceiling is set by enrichment rather than by the model. Ask for a fill rate on 200 to 500 of your own accounts, not on the vendor's sample.

Write-back is second because a score living in a tool nobody opens changes no behaviour. Confirm the direction of sync and the object it lands on.

Cost per scored record is third, and it is computed from the credit rate rather than the plan price. Cleanlist AI at 5 credits per qualification works out to roughly $0.33 per record on Starter ($49 per seat for 750 credits) and $0.30 on Pro ($89 per seat for 1,500), with 25% off on annual billing. Run the same arithmetic on every shortlisted vendor before comparing sticker prices.