Try Valley
Put your outbound workflow to work.
Saniya
Published:
Updated:
Published September 8, 2026 · Updated September 8, 2026
The answer up front: the standalone lead-scoring category quietly died. In eighteen months, MadKudu was acquired by HG Insights, Breadcrumbs was acquired by MadKudu, Clearbit became HubSpot's Breeze, Forwrd became Pendo Predict, Koala shut down, and Toplyne wound up - there is no credible self-serve pure-play left with public pricing. Lead scoring now lives inside your CRM (HubSpot requires the $3,600/month Enterprise hubs for predictive scoring; Salesforce sells Einstein scoring from its $195/user Enterprise tier; Zoho's Zia starts on Professional) or inside enterprise platforms (HG/MadKudu, ~$55K/yr reported median). Below: exactly which CRM tier unlocks scoring, the data minimums vendors publish (Salesforce's is 1,000 leads with 120 conversions), and why half of scoring projects collapse into MQL theater.
Disclosure: Valley (that is us) sells signal-based outbound - an adjacent lane, not an intent-data feed - so we profit if you conclude signals beat scores. The facts are sourced so you can check them.
How we know this category: we run signal-based outbound for 1,000+ LinkedIn accounts (30M+ messages sent). Every price below was verified September 8, 2026 against the vendor's live page, or labeled as reported where the vendor publishes nothing - buyer-reported medians come from Vendr's transaction marketplace.
Valley is an AI outbound platform that finds the people most likely to buy from you - from live buying signals, natural-language search, or your own lists - qualifies them against your ICP, researches each one, and sends personalized messages across email and LinkedIn that get 15-45% reply rates, against a category average of about 2%.
► Scoring ranks a list. Signals replace it: see the difference on your ICP - free for 7 days
The consolidation, in one table (all verified September 8, 2026)
What happened | When | Where it lives now |
|---|---|---|
MadKudu acquired by HG Insights | Aug 2025 | HG's enterprise suite; madkudu.com/pricing redirects to hginsights.com |
Breadcrumbs acquired by MadKudu | Oct 2024 | Folded in; its pricing page redirects to a MadKudu demo form |
Clearbit acquired by HubSpot | 2023-24 | Breeze Intelligence, credit-metered inside HubSpot; free Clearbit tools shut Apr 2025 |
Forwrd.ai acquired by Pendo | Jul 2025 | Pendo Predict; no public pricing |
Koala shut down | Sep 2025 | Team acqui-hired; product gone |
Toplyne wound down | Oct 2024 | Capital returned to investors |
Where predictive scoring actually lives now, and what it costs
Platform | Which tier has it | Price (verified Sep 8, 2026) |
|---|---|---|
HubSpot | Predictive 'likelihood to close' requires Marketing Hub Enterprise OR Sales Hub Enterprise - their own docs name exactly those two | Marketing Enterprise $3,600/mo (5 seats, $7,000 onboarding); Sales Enterprise $150/seat/mo |
Salesforce | Einstein Lead/Opportunity Scoring purchasable on Enterprise Core and Advanced, included in Max | Core $195 · Advanced $395 · Max $550 per user/mo |
Zoho CRM | Zia predictive intelligence from Professional up | Professional ~$23 · Enterprise ~$40 per user/mo (reported USD) |
Pipedrive | No true predictive scoring; win-probability on upper tiers | Reported $14-79/seat/mo annual across renamed 2026 plans |
MadKudu (HG Insights) | Enterprise sales only | Reported median ~$55,008/yr (Vendr) |
The data floor (the number vendors whisper)
Salesforce publishes it plainly: Einstein Lead Scoring needs at least 1,000 leads created in the last 180-200 days, of which at least 120 converted - per segment; below that it falls back to a generic model trained on other companies' data. HubSpot publishes no minimum and describes its model as "blackbox machine learning"; practitioners report ~25 converted contacts as the bare floor and ~1,000 contacts for reliability. The honest rule that falls out: under ~50 clean conversions, rule-based scoring is the correct choice, not a fallback - it is transparent, explainable to sales, and correct at low volume. Predictive earns its keep only on real conversion history, and silently degrades when your ICP or product shifts.
Why scoring projects fail (the documented patterns)
Score inflation: engagement-heavy models without decay let newsletter-openers outrank actual buyers - "half the database ends up above MQL" once new activities get points without re-weighting. MQL theater: when marketing is comped on MQL volume, the incentive shifts from finding buyers to manufacturing threshold-crossers - the intern researching a paper trips the score while the evaluating CFO sits at zero. Marketing-builds-alone: models built without sales input get their leads rejected, then ignored. Every one of these is a variant of the same root cause: a score is a proxy, and proxies drift from reality unless someone maintains them.
The alternative frame (our lane, declared)
Scoring exists because databases are too big to work: you buy a list of thousands and rank it. Signal-based outbound inverts the problem - start from the people already showing behavior (viewing your profile, engaging posts, hiring for the role you serve, raising money) and you rarely need a model to tell you who is warm; the behavior IS the score. Valley runs that motion end to end: signals in, ICP qualification, 200+ sources of research, messages in your voice, your approval on every send - 15-45% reply rates against a ~2% category average (methodology). If your database is big and your CRM is rich, score it. If your problem is starting conversations, work signals: the tools compared.
► The behavior is the score: run Valley free for 7 days and skip the model maintenance
How to choose in 30 seconds
On HubSpot/Salesforce with 1,000+ leads and real conversion history: use the native predictive tier you may already pay for. Smaller than that: rule-based, built WITH sales, reviewed quarterly. Enterprise fit+intent modeling: HG/MadKudu. Conversations rather than rankings: signals - intent data tools for accounts, Valley for named people.
FAQ
What is the best lead scoring software in 2026? For most teams: the scoring already inside your CRM - HubSpot Enterprise hubs or Salesforce Einstein (Enterprise Core and up). The standalone category consolidated away: MadKudu, Breadcrumbs, Clearbit, Forwrd, Koala and Toplyne were all acquired or shut between 2024 and 2025.
How much data does predictive lead scoring need? Salesforce publishes the clearest floor: 1,000+ leads in ~6 months with 120+ conversions, per segment. Below roughly 50 clean conversions, rule-based scoring outperforms in practice because it is explainable and does not overfit noise.
Rule-based or predictive? Rule-based: transparent, works at low volume, needs quarterly maintenance. Predictive: removes human bias at real volume, but opaque ('blackbox' is HubSpot's own word) and it degrades silently as your market shifts. Volume decides, not preference.
Why do lead scores stop being trusted? Inflation (points without decay), MQL theater (comp plans rewarding threshold-crossing over buying), and models built without sales input. A score is a proxy; unmaintained proxies drift until sales ignores them.
Is Valley a lead scoring tool? No - it removes the need for one in the outbound lane: instead of ranking a purchased list, it starts from live buying signals, qualifies against your ICP, researches each person, and drafts the outreach for your approval. $149/month billed quarterly, 7-day free trial.
► Rank less, converse more: start free, see a reply this week
