The ICP analysis you've always wanted to build. Done in three minutes.
You know the patterns are in the CRM data. You've just never had the tools to extract them properly. Telepath does the multi-dimensional clustering and weighted scoring that would take a data science team months — then hands you the output to operationalise across your stack.
See the ICP analysis your CRM data has been waiting for →Sound familiar?
Asked to “build an ICP framework” repeatedly but can't do weighted multi-dimensional analysis with pivot tables
CRM full of data nobody uses — hundreds of fields, half empty, insight locked inside
Spending time on data quality firefighting instead of extracting strategic intelligence
Surface-level segments by industry or company size that don't actually drive behaviour
Companies lose approximately 15% of revenue due to poor-quality data.
The ICP analysis you've been asked to build (and never quite finished)
If you’ve worked in RevOps for more than two years, this scene will be familiar. Your VP Sales pulls you into a meeting and asks, with the slightly hopeful tone of someone who’s been thinking about it on the train, whether you could pull together a proper ICP analysis. Look at our closed-won deals, find the patterns, build us a scoring system. We need to know which deals to chase and which to let go.
You say yes, because saying no isn’t really an option. You scope it out. You start exporting deal data from HubSpot or Salesforce. You realise the data quality is worse than you remembered. Industry fields are inconsistent. Company size is missing on a third of the records. Lead source has been rewritten by three different marketing platforms. You spend two weeks just on data hygiene before any analysis can start.
Then you pivot to the actual analysis. You start in Excel because that’s what’s available. You realise pivot tables can show you what won deals look like by industry, but they can’t show you which combinations of attributes predict wins. You try clustering in Python. You get something working but it takes three days and the output is hard to explain to anyone non-technical. You try a couple of off-the-shelf tools. They want £50k a year and a six-month implementation, and the output is a generic ICP framework rather than something specific to your company.
Six weeks have passed. The VP Sales asks for an update. You give them a partial answer — a chart by industry and a chart by company size. They look at it, nod politely, and then go back to forecasting on gut feel because that’s what’s actually in front of them.
This is the structural problem. Proper ICP analysis requires multi-dimensional clustering, statistical weighting, ongoing recalibration as the market changes, and an output format that operationalises into the CRM and lead routing rules. None of that is a one-person two-week project. Most RevOps teams know exactly how to do this in principle. The blocker is execution capacity, not knowledge.
Telepath is the analysis you’ve been asked to build. The maths is done. The output is operationalised. Start with a free ICP report and see what falls out of your data. You take the credit.
After Telepath
Multi-dimensional clustering and weighted scoring criteria done automatically
Output you can operationalise: T-Scores as CRM custom fields, scoring criteria feeding lead routing
CRM adoption improves because reps see WHY data matters — it feeds their scores
Move from operational support to strategic influence — you're driving the ICP intelligence
Integrates with HubSpot and Salesforce — reads existing data, writes scores back
What weighted scoring actually requires, and why most teams give up
When sales leaders ask for “deal scoring,” they usually picture something simpler than what’s actually involved. RevOps people know the gap intimately, because they’re the ones who’d have to build it.
A scoring system that genuinely predicts wins needs five things working together.
One: dimensional reduction across messy categorical data. Deal records contain dozens of fields, many of them text-based and inconsistently filled. You need a way to identify which fields actually carry signal and weight them accordingly. Industry alone explains maybe 15–20% of win variation. Industry combined with company size and tech stack explains 60–70%. Working out the right combinations is non-trivial.
Two: handling small datasets gracefully. Most B2B companies have hundreds, not millions, of closed-won deals. Standard ML approaches that need 10,000+ training examples don’t apply. You need techniques that work with 50–500 deals and produce statistically defensible outputs.
Three: segment identification rather than one-size-fits-all scoring. Most companies don’t have one ICP. They have two to four distinct customer segments — small/medium-sized professional services firms in one segment, mid-market technology companies in another, regulated enterprises in a third. A scoring system that averages across segments scores all of them mediocrely. A scoring system that recognises segments scores each one accurately.
Four: a writeback mechanism. A score that lives in a separate dashboard is information. A score that writes back to the CRM as a deal property and feeds lead routing is operational. The difference between the two is whether the analysis actually changes how the team works.
Five: continuous recalibration. The ICP that wins deals in 2024 isn’t necessarily the same as the one that wins in 2026. Markets shift. Buying behaviour changes. Your product evolves. A static scoring model degrades silently. A model that re-runs against the latest closed-won data adapts in step with reality. This is where ICP Drift Intelligence matters most.
Each of these alone is a week of work. All five together is a six-month project that most RevOps teams don’t have time for, even if they have the skill. Telepath is the entire stack of capability, productised, with sub-three-minute setup.
How Telepath fits into a RevOps stack
Telepath is built to operationalise into the tools and workflows you already run, not to replace them. It’s pipeline intelligence that slots into your existing stack.
On the CRM side, the integration is OAuth-based and minimal. HubSpot is fully live, Pipedrive is fully live, Salesforce is built but not yet pilot-tested with a real customer. The connection reads closed-won and open deal data; it writes back two custom properties (telepath_t_score and telepath_w_score) to deal records and a confidence label that sales reps can filter and sort on. Five additional telepath_ properties are written for diagnostic visibility. Nothing else in the CRM is modified. Full details on data handling are in the data processing agreement and security documentation.
On the lead routing and scoring side, the T-Score becomes an input you can use in any rule-based routing system. If your highest-scoring segment is mid-market PropTech in the UK, you can route those deals to your most experienced AE. If a deal scores below 30 across all six ICP dimensions, you can flag it for review or auto-disqualify. The score is just a number on a deal record; whatever your stack does with custom number properties, it can do with this one.
On the BI / dashboarding side, both T-Score and W-Score are queryable via the CRM API. If you’ve built reporting in Looker, Tableau, Mode, or any custom BI surface, you can include them as dimensions or filters without further integration. The most common use case is a “scored pipeline value” calculation that weights deal value by T-Score, producing a far more accurate forward-looking revenue view than raw pipeline.
On the GTM workflow side, Telepath produces a daily Slack brief per rep summarising their top-scoring deals, deals worth re-engaging, and deals worth pausing. This is where most sales reps interact with the product day-to-day. You don’t need to manage these briefs; they generate from the underlying data automatically.
On the analysis side, the ICP segments that Telepath identifies become the layer above industry/firmographic reporting. You’re no longer asking “what’s our win rate by industry” — you’re asking “what’s our win rate in segment 2, the mid-market FinTech segment with hiring signals and Series B funding,” which is a much more useful question. The segments are exportable so you can use them in Apollo searches, intent data filtering, or paid acquisition targeting.
The product was designed by someone with a sales background who used to wish RevOps could build this for them. The tool now exists. RevOps gets the credit for finally landing it.
How it works
Upload your closed-won deals
CSV from any CRM
AI analyses patterns
Across every deal in under three minutes
Get your weighted ICP scoring rubric
See what actually predicts a win
Questions RevOps people ask before evaluating Telepath
It's actual statistical analysis, not an LLM-wrapped guess. The pipeline runs unsupervised clustering on closed-won deals to identify segments, then computes weighted feature importance across each segment. LLMs are used in the data preparation layer (parsing inconsistent industry strings, normalising company size estimates from public data, enriching missing fields) but the scoring itself is mathematical rather than generative. We can produce a methodology document under NDA if your evaluation requires it.
HubSpot is fully supported and is our primary integration — read access to deals, contacts, companies, engagement events, and write access to seven telepath_ deal properties. Pipedrive is fully supported with the same capability set. Salesforce is built and code-complete but has not yet been tested with a paying customer's data, so it's currently labelled as beta. We don't yet support Microsoft Dynamics or Zoho, though we'd consider building either if there's a clear paying-customer use case.
Customer data is processed via the CRM connection and stored in our Supabase database (Postgres with row-level security). Database hosting is in the EU. Application infrastructure is on Vercel (SOC 2 Type II) and a Hostinger VPS in the Manchester data centre (ISO/IEC 27001:2022 certified). We're ICO-registered (number ZC101774). The DPA is at /data-processing-agreement and lists the full sub-processor chain including Apollo.io, Google Gemini, Brandfetch, Companies House, Sentry, Axiom, PostHog, Google Drive, and Hostinger.
For meaningful pattern extraction, you need a minimum of 20–30 closed-won deals. Below that the noise overwhelms the signal. Above 200 deals the patterns stabilise quickly. The product reads associated contact, company, and engagement records for additional dimensional richness, but the closed-won deal records themselves are the core input.
Partially. The core statistical analysis runs without any LLM involvement — that path is fully on our infrastructure with no third-party AI processing. The data enrichment layer (parsing fragmentary text fields) does use LLMs by default, but can be turned off if your security review requires it. Doing so reduces analysis quality on messy data but is a supported configuration.
Three things land in your hands. First, an ICP analysis report in PDF and HTML showing identified segments, weighted scoring criteria, and recommended actions (a sample is in the public materials). Second, two deal-level scores written back to the CRM as queryable properties. Third, an exportable segment definition you can use in any tool that accepts attribute-based filtering. You can build dashboards, routing rules, territory models, and quota plans on top of any of these.
Different problem space. 6sense is intent data: which accounts are researching topics in your category right now. Clari and BoostUp are forecasting and pipeline analytics platforms: probability-weighted forecasting on top of your existing pipeline. Telepath sits upstream of all three. It tells you which deals already in your pipeline match the pattern of your historical wins, regardless of intent signal or stage. You can run Telepath alongside any of those tools without overlap.
Three tiers: Reader at £500/month, Vision at £1,500/month, Oracle at £3,000/month. Pricing is by feature tier rather than per-seat, so the maths is straightforward. Most RevOps evaluations land on Vision because it includes the daily Slack briefs, ICP Drift Intelligence, and full CRM writeback. Reader is for read-only use cases (dashboards and reporting), and Oracle adds rep-level coaching analytics. There's no negotiation or annual lock-in pressure on the website tiers; commercial terms are flexible if you need a custom shape.
The ICP analysis you've always wanted to build. Done in three minutes.
See the ICP analysis your CRM data has been waiting for →Free. Three minutes. No signup required.
Your pipeline has a story. Learn to read it.
Weekly insights on pipeline intelligence, ICP strategy, and deal scoring — from someone who's spent 20 years in B2B sales.
No spam. Unsubscribe anytime. Read past issues →
Explore other perspectives