QuadSci

What it is

QuadSci is a prediction layer that sits under your customer success platform rather than replacing it. It reads raw product telemetry — sessions, feature adoption by role, module and API usage, workflow completion — alongside CRM, contract and call data, and classifies every account as growth, stable, contraction or churn, with the forecast pointed 9 to 18 months ahead of renewal. The output lands in the tools your team already works in: Gainsight CTAs, Salesforce records, Clari, Salesloft and Slack.

There are two products. Cohorts AI segments customers by how they actually use the software and ties each behavioural cohort to ARR. Growth AI does the account-level forecast and prescribes the intervention. On top sits Q-Chat, a conversational agent layer with Customer Success, Sales, Services and Product Marketing agents; the 3.0 release made it fully conversational (“which accounts are starting to slip?”), and an MCP layer exposes the same signals to Agentforce, Copilot, LangGraph, CrewAI or your own agents.

The company was founded in 2023 in Garden City, New York, by co-CEOs Sean Murray and Dan Harmeson, both previously at Elastic and MuleSoft. It raised an $8M Series A led by Crosslink Capital on 17 February 2026, with Alumni Ventures and Correlation Ventures participating. Third-party trackers put headcount near 40.

Why it shows up in CS stacks

• It answers the question health scores cannot. A CSP health score is a weighted blend of lagging inputs — logins, NPS, ticket volume, CSM sentiment — and it goes red once the customer is already leaving. QuadSci’s pitch is lead time: a signal early enough that a product gap can be closed before the renewal conversation starts.
• The case studies carry dollar figures, not adjectives. Boomi, managing 30,000+ customers and more than $5M a year of surprise churn, trained the model on 70 billion telemetry events from Salesforce, Gong, Marketo, Gainsight and product analytics. QuadSci reports $27M of at-risk ARR surfaced, 18% of it mitigated, and a 10% gain in six-month renewal-forecast accuracy, with predictions writing straight into Gainsight CTAs. At Clari, 6+ billion signals produced 90% accuracy on churn and growth and 80% on contractions six months out — and the headline finding was expansion, not risk.
• The data stays in your warehouse. The model deploys inside your environment on Snowflake, Databricks, BigQuery, Redshift or Synapse across AWS, Google Cloud or Azure, with no public endpoints, SSO-only access and read-only defaults. It holds SOC 2 Type II. For a security team that refuses to ship raw event streams to a startup, this is the deciding feature.
• It surfaces pipeline nobody opened. QuadSci says it finds, on average, 15% of a customer’s ARR sitting in a predicted-growth state with no open opportunity — which is why RevOps, not only CS, ends up owning the budget.

Pricing reality

There is no public rate card and no self-serve tier. Subscriptions are tiered by your ARR, not by seats, so a CS team of 12 and a CS team of 120 at the same company pay the same. No third-party price data exists yet. The one public anchor is indirect: Boomi’s case study claims 700%+ ROI against a retained $4.9M (18% of $27M), which puts that platform spend at no more than roughly $600K. Treat that as a ceiling for a large installed base, not a quote. Expect an enterprise-style annual contract, a data-integration phase before the first scored account, and a proof-of-concept backtest as the price of entry.

Best for

VPs of Customer Success and RevOps leaders at B2B SaaS companies with roughly $50M+ ARR, a large base of annual or multi-year contracts, and well-instrumented products (Pendo, Amplitude, Mixpanel, Heap, Segment or warehouse event tables) whose problem is surprise churn: renewals lost that the health score rated green. It fits best when a CSP such as Gainsight is already in place and the goal is to make it act earlier, not to replace it.

Skip it if you sell monthly self-serve plans (the 9-to-18-month horizon is longer than your contracts), if product events are not tied to account IDs, or if you have fewer than a few hundred accounts — the model needs enough historical renewals and churns to learn from.

Versus the alternatives

By installed base, Gainsight and ChurnZero own this budget line; both ship health scoring and risk alerts inside the CSM workspace. Stay there if your churn is visible 90 days out and the fix is CSM coverage. Planhat and Vitally are the pick when the complaint is workflow and data modelling, not prediction. Pendo and Amplitude give you the raw usage trends and let an analyst read them. The do-it-yourself route — a churn model built by your data team in the warehouse and synced into the CRM with Hightouch — wins when you already employ ML engineers and want to own the features. QuadSci wins only when lead time is the problem: you need 12 months of warning, not 90 days, and you want it without staffing a data-science team.

Watch-outs

• The accuracy figure moves between sources and is undefined. The Series A release says 94% at 12 to 18 months; the website and the June 2026 award release say 90%+ at 9 to 18 months. Neither says whether that is precision, recall or overall accuracy — and on a base with 8% annual churn, a model that predicts “renews” every time is 92% accurate. Guard: make the paid pilot a blind backtest on your own renewals from 18+ months ago, and get precision and recall for the churn class, by segment, in writing.
• No telemetry, no product. Every case study starts with billions of events mapped to accounts. Guard: before the first call, confirm that product events carry a stable account ID that joins to your CRM, and that you retain at least two renewal cycles of history.
• Trust takes longer than deployment. Boomi’s own team puts it at six months to get CSMs acting on the predictions. Guard: push signals into the existing CTA or task queue rather than a new dashboard, and set one measured outcome — for example, save rate on flagged accounts versus a holdout — before rollout.
• It is a young, small vendor. Founded in 2023, about 40 people and $8M of disclosed funding, selling into multi-year renewal decisions. Guard: the warehouse-native deployment keeps your data in place; also contract for export of scores, cohort definitions and feature importances, and keep the first term to 12 months.