Chattermill

What it is

Chattermill reads customer feedback that already exists across your channels and turns the free text into themes, metrics and quotes you can query. Its connector directory covers support desks (Zendesk, Intercom, Gorgias, Kustomer, Freshdesk, Gladly, Dixa, Talkdesk, Aircall), survey platforms (Qualtrics, Medallia, Forsta, SurveyMonkey, Typeform, Delighted, AskNicely, Survicate), review sites (Trustpilot, G2, Capterra, Google Play, the Shopify and Chrome stores, Tripadvisor, Yelp), call recordings via Gong and Modjo, ten social networks, and warehouses — Snowflake, Databricks, BigQuery, Amazon S3 — for the business context that sizes a theme against revenue. The vendor’s product page claims 65-plus channels; the homepage says 50-plus native integrations. Count the directory yourself against your own sources before you take either number into a business case.

The analysis engine is branded Lyra AI, and Chattermill is unusually specific about what is inside it: aspect-based sentiment analysis, supervised learning on fine-tuned specialist models, and purpose-tuned LLMs for open-ended reasoning. That mix matters commercially. Aspect-based sentiment scores the sentiment of each thing mentioned in a comment rather than the comment as a whole, so “delivery was fast but the app crashed twice” resolves into two labelled aspects instead of one neutral score. The named modules sitting on top are Precision Insights, Impact Analysis, Anomaly Detection, Reports & Dashboards, and Workflows & Automations, with PII redaction and translation across 100-plus languages in the ingestion layer.

Chattermill was founded in London in 2015 and raised a $26M Series B on 6 December 2022 led by Beringea, with DN Capital, Ventech, Runa Capital, btov, SVB and Blossom Street Ventures participating; Tracxn puts total funding at $40.5M across six rounds. Uber, Booking.com, Just Eat, HelloFresh, Tesco, Santander, J.P. Morgan, Qonto, E.ON Next and Zappos appear on the public customer list. SOC 2 Type II, ISO 27001:2022, GDPR and CCPA are covered, with SSO and per-user access controls.

Why it shows up in Customer Success stacks

• The MCP server is permission-scoped to the person asking, not to the workspace. Chattermill exposes feedback summaries by topic, segment and period; quantified insight — issue volumes, metrics, sentiment; raw verbatims filtered on your criteria; and cross-referencing against other connected systems. Agents inherit the authenticating user’s existing Chattermill permissions, so a CSM’s assistant sees the accounts that CSM already sees. Fifteen-plus clients are named — Claude, Claude Code, ChatGPT, Glean, Notion Agent, Cursor, Codex, Gemini CLI, OpenClaw, Manus, Dust, Mistral Le Chat, Microsoft Copilot, Grok. This is the part that changes the daily motion: the VoC corpus becomes answerable from the session where someone is already writing the QBR, instead of from a dashboard nobody opens between quarters.

• The consumer-channel coverage is the real moat. Twenty review sources and ten social networks is a different product bet from the support-ticket-first tools. If your customers complain on Trustpilot, in app-store reviews and on Reddit before they ever file a ticket, the tools that only read your helpdesk are reading the minority of your evidence.

• Anomaly Detection turns the corpus into an alert instead of a report. A quarterly readout tells you what went wrong last quarter. A spike alert on a theme tells you while the fix is still cheap, which is the difference between VoC as a reporting function and VoC as an operating one.

Pricing reality

Nothing is published; sales@chattermill.io and a demo form are the only route. Vendr’s transaction data puts the average Chattermill contract at $63,500/yr across more than five purchasers and five completed deals, with no floor or ceiling disclosed. The demo form segments prospects by feedback volume — under 1,000 per month, 1,001 to 5,000, above 5,000 — which tells you what the meter actually is. Price the deal on channels and volume, not seats, and assume the mid-five-figures band unless you are buying a single-market pilot. Multi-year commitments move the annual number; ask for the two-year and three-year quotes side by side at first contact rather than after you have anchored on one.

Best for

Consumer-scale CX and Customer Success leaders in retail, marketplace, delivery, travel, telco and consumer fintech, running above roughly 5,000 feedback items a month across four or more channels, in more than one language or market. The fit sharpens when reviews and app-store ratings are a real commercial signal and not a vanity metric, and when someone on the team is already working inside Claude or ChatGPT daily — the MCP path is worth little to a team that will only ever open the web app.

Do not buy it if

Your feedback lives in one system. If Zendesk is the whole surface, its native intelligence plus a saved view gets you most of the way for the integration work you avoid. The same holds below about 1,000 items a month, where a monthly export into a Claude project answers the same questions for the cost of an afternoon — at $63.5K, this is roughly half a fully loaded analyst, so the corpus has to be bigger than one person can read. It is also not a survey platform: Chattermill connects Qualtrics, Medallia and Typeform as sources, it does not run the survey program, own the case management or handle closed-loop follow-up. And if your board deck is built in PowerPoint and Excel, check the export path in the trial — reviewers name that specific gap.

Versus the alternatives

• Qualtrics — the largest install base in experience management, and it closed its $6.75B acquisition of Press Ganey Forsta on 18 May 2026, followed by job cuts in Seattle, Utah and overseas as it absorbs the deal. Pick Qualtrics when you need survey collection, case management and program governance on one contract. Pick Chattermill when the feedback already exists and reading it is the job.
• Medallia — the other incumbent by install base, and the one to price risk against. On 22 April 2026 a creditor group including Blackstone, KKR, Apollo Global and Antares Capital took control from Thoma Bravo through a debt-for-equity swap that wiped out roughly $5.1B of sponsor equity. The product runs and accounts are being served; what changed is who sets the roadmap. Get a written product-investment commitment before signing a multi-year term.
• Enterpret — the closest AI-native peer and the head-to-head you will actually run. Enterpret’s Adaptive Taxonomy re-learns your label set rather than asking you to maintain one, and its MCP server exposes seven explicitly named tools; its median contract runs $36,302/yr against Chattermill’s $63,500 average. Pick Enterpret for B2B SaaS where support tickets and product feedback are the corpus and the budget sits lower. Pick Chattermill when reviews, app stores and social carry your brand signal and you need multi-market language coverage.
• Unwrap.ai — the fastest-growing entrant in the segment, with a $12M Series A led by Scale Venture Partners in January 2025 and Microsoft, Lyft, jetBlue, Oura and Perplexity named as customers. Pick it when the buyer is a product team and the budget sits below Chattermill’s floor.
• Gainsight — a neighbor, not a substitute. It owns account health and the CSM motion; Chattermill supplies the verbatim evidence behind the score. Teams running both feed Chattermill’s anomaly signal into health rather than maintaining two opinions about the same account.

Watch-outs

• Implementation takes longer than the marketing implies. The vendor’s comparison content frames deployment as weeks; G2’s own time-to-implement figure for Chattermill is 2 months, and Capterra reviewers say it is “hard to configure before you can actually start using it.” Guard: write onboarding milestones into the order form with a go-live date, and validate the taxonomy against a hand-labelled sample of 200 records in month one. Label disagreement above roughly 10% is a scoping conversation, not a launch.
• The taxonomy resists change after go-live. Reviewers specifically raise difficulty restructuring themes and categories as the business evolves, which is the failure mode that turns a live model into a description of last year’s product. Guard: ask in the trial to see a category split and a merge performed on your own data, timed, before signature — not described in a deck.
• Volume metering means “connect everything” is how the renewal doubles. Guard: start with the four channels that already drive decisions, add one only when a specific unanswered question required it, and get the volume tier and the per-unit overage rate written into the first order form.
• Dashboards and exports are the weakest surface. Capterra reviewers call filters and dashboards difficult, exports “fiddly,” reports dated, and name the absent PowerPoint and Excel integration. Guard: rebuild your actual QBR chart inside the trial rather than accepting the demo dashboard, and treat the MCP server as the primary reporting path if the UI does not clear the bar.
• Account management is the recurring service complaint. One reviewer reports “days or weeks between responses to queries.” Chattermill rates 4.5 on G2 across 218 reviews and 4.5 on Capterra across 25, so this is a tail risk, not the median experience. Guard: name the CSM and a support-response SLA in the contract, with a mid-term review clause.
• The headline customer numbers are vendor-published. A 144% NPS increase at E.ON Next, a 42.8% drop in contacts per transaction at Footasylum and 80-plus hours saved at Qonto all come from Chattermill’s own site. Guard: treat them as directional and ask for a reference call with a company at your feedback volume and channel mix, not at Uber’s.