Index Graph vs Evidence Graph
What an evidence graph is, what goes wrong when a language model answers pipeline questions without one, and the four tests a fact should pass before a revenue leader acts on it.
Read the whitepaper →A 60 day services engagement inside your RevOps team. We connect your CRM, calls, email and Slack inside your own cloud, count what separates your wins from your losses. Your team runs it by day 60 and keeps everything we build, on the AI license you already pay for.
Most win/loss programs run on a dropdown of loss reasons and a handful of buyer interviews. Rev-Lens joins the calls, email, Slack threads and CRM history behind every closed deal, then counts which patterns separate your wins from your losses.
Shown on anonymised deal data. Names are changed. No customer is identified.
Two minutes, from the raw material to the findings: twelve months of closed deals and the calls recorded against them, then why deals were lost, why the won ones were won, and what is sitting in the live pipeline right now. The example comes from a real engagement.
Point a model at two years of sales calls and something has to decide which few passages it reads. Rev-Lens builds an evidence graph inside your own cloud that links every statement to who said it, on which call, on which deal, and how that deal ended. The model reasons over those connections, so each finding carries a count and the customer's own words, not a guess built from the passages that sounded closest.
Statement→Speaker→Call→Opportunity→Outcome
What an evidence graph is, what goes wrong when a language model answers pipeline questions without one, and the four tests a fact should pass before a revenue leader acts on it.
Read the whitepaper →Eight analyses read from one company's evidence graph, built as input to an annual go-to-market plan. Every finding is dated, quoted and counted, with the checks that show why the numbers hold. Names and figures are changed.
See what is in it →The agents in your stack flag which deal is at risk, one deal at a time. Rev-Lens joins every opportunity in your history and shows which behaviors across your GTM process are associated with winning and losing.
We work with what you run. Keep Gong for coaching, Salesforce for the record and your LLM for daily work; we join their history inside your infrastructure.
Built for B2B revenue teams with a working RevOps function and at least 80 closed deals in the period. Most have already tried pointing AI at their data and want the version that counts.
"Where and how do we actually win?"
"Which pipeline turns into revenue?"
"Is the market hearing what we built?"
"Every cross-system question lands in my queue."
"Three executives, three dashboards."
Agree three questions and the sources that can answer them. The 60 day clock starts once access is in place; most of the setup work is ours.
Calls, email, threads and CRM history joined to the same deals, inside your infrastructure. We show you whether your data can answer the questions. If it can't, we stop there.
A prioritized set of findings, each counted across your deals, traceable to its source records, with a recommended action and a named owner. Challenge any of them.
Your RevOps team reruns a finding without us. Then you decide whether you want more help. Everything we built stays with you either way.
Everything we build is yours to keep: the joined dataset in your warehouse, the agents on your LLM license, and the findings with their owners.
See the full 60 day scope →Gong is excellent at calls, coaching and deal-level insight, and its agents can now pull context from other systems through MCP. But they work one opportunity at a time, in Gong's cloud, and they don't count patterns across your pipeline. We use Gong's export as one of our inputs. You keep Gong. For the longer version, see how Rev-Lens compares to conversation intelligence, forecasting and win-loss research firms.
For a question about one deal, yes, and you should. For a question about 400 deals over a year, connectors return summaries one call at a time, with rate limits and no join between systems. You get a cited answer, not a counted one. We build the joined dataset so your LLM can count, and you keep it.
A fixed fee for the 60 days, agreed in writing before any access and set by scope: the questions and the sources. Not by company size or seat count. If the day 30 check shows your data can't answer the questions, we stop there.
Your own cloud storage and model usage, billed to your accounts as they are today. We don't resell infrastructure or mark up compute. We'll check on the first call that your model license includes API access, because many chat subscriptions don't.
Usually six months of closed deals for sales cycles of 30 to 45 days, and twelve months for longer cycles, depending on what your systems hold.
We'll tell you on the first call if we can see it. If it only shows up once the data is joined, the day 30 check catches it: we stop there.
When access is in place. Security reviews and approvals sit on your side, so we don't let them eat into the work. Most of the setup is ours.
One executive sponsor, one RevOps or data contact to grant access, and an hour from each leader whose question is in scope. We don't need rep time.
Access limited to the sources you name, a governed copy inside your own cloud account, your licensed model, no pooling, no training. We put the specifics in writing before any access is granted. The detail is in who owns what.
Everything we built stays with you: the data lake, the agents and a runbook, with your RevOps team trained to run them. Then you decide whether you want more help. If you do, it's month to month: regular reads and tuning, or new questions and sources. No long-term contract.
How many of the deals in your pipeline share this pattern right now?
A 30-minute call. We'll check whether your data can support the question. If it can't, we'll say so.
Case studies, white papers and the podcast: Resources →
We take on a small number of engagements each quarter, so each one gets the founders' time.