Somewhere inside almost every large company sits a spreadsheet nobody wants to open. It holds thousands of open-ended survey answers, support tickets and app-store reviews, all of it written by real customers, all of it more or less ignored. Reading it by hand takes weeks. Skimming it produces guesses. Thematic exists because its founder got tired of doing the reading.
Thematic is an AI feedback analytics platform. It takes the messy, unstructured stuff customers write - surveys, tickets, chat logs, call transcripts, reviews, social posts - and turns it into themes, sentiment and numbers a team can act on. The company describes itself as "the customer intelligence layer for CX," and the pitch is narrow on purpose: not another dashboard, but one traceable view of what customers are actually saying, assembled from their own words.
The company was founded in 2017 in Auckland, New Zealand, and went through Y Combinator's Summer 2017 batch. Its two co-founders, Dr. Alyona Medelyan and Nathan Holmberg, are married. Medelyan, the CEO, holds a PhD in natural language processing that was funded by Google and has been cited by more than 3,000 researchers. The idea traces back to 2015, when she was running an NLP consultancy and kept running into the same wall: clients were sitting on mountains of feedback and had no good way to read it.
There is something fitting about the name. "Thematic analysis" is a qualitative research method for finding recurring patterns in text - the kind of careful, slow work Medelyan trained in academically and then set out to automate. The company is less a pivot toward the hot thing and more a straight line from what its founder already knew how to do, which is part of why the product feels opinionated about how the reading should be done rather than agnostic about it.
What the software actually does
Most feedback tools start with a taxonomy - a fixed list of categories like "pricing," "shipping," "app crashes" - and then sort comments into those bins. It works until customers start talking about something the list didn't anticipate, at which point the tool goes quiet exactly when you need it. Thematic works the other way around. It reads the feedback first and builds the themes from the language people actually use, so a new complaint about, say, a checkout change shows up as its own theme rather than getting swept into a vague "other" pile.
On top of that theming sits a growing set of products. Themes & Taxonomy does the discovery. Impact Analysis ties each theme to business metrics like NPS and CSAT, so you can see which issues are actually dragging the numbers down instead of just which ones are loudest. Lenses stitch multiple data sources into one view. Workflows route findings to the teams that own them. In 2024 the company added Answers, a generative-AI tool where anyone can type a question in plain English and get back a written answer with charts and, importantly, the verbatim customer quotes behind it.
From raw feedback to confident action.Thematic's own framing of the job
That last detail - the quotes behind the answer - is the whole personality of the product. In a market full of AI that sounds confident and is sometimes wrong, Thematic's bet is on traceability: every insight links back to the exact comments that produced it. It is a less flashy promise than "ask the AI anything," and a more useful one when you are the person who has to stand in front of a VP and explain why a metric moved.
Who is actually using it
The customers are enterprise CX, insights and product teams - the people whose job is to know what customers think and who drown in the raw material. Named users include DoorDash, LinkedIn, Instacart, Vodafone, Atom Bank, Mitre 10, Woolworths and Albertsons. These are companies with feedback volume measured in the millions of comments, where hiring enough analysts to read all of it is not a real option.
The payoff the company points to comes from a Forrester Total Economic Impact study, which reported a 543% return on investment over three years for customers on the platform, with payback in under six months. Thematic also cites customers seeing insights arrive up to 92% faster and, in one case, a 69% reduction in contact-center calls after acting on what the feedback revealed. Numbers like these come from a vendor-commissioned study, so read them as the ceiling of a good outcome rather than a guarantee - but the direction is the point: less time spent reading, more time spent fixing.
The problem, stated plainly
Companies do not lack feedback. They lack the ability to read it at the speed decisions happen. A quarterly survey gets summarized into a slide; the free-text answers - where the real reasons live - get skimmed or skipped. Support tickets pile up as a cost center, not a signal. The result is a familiar gap: leaders sense something is wrong with the customer experience but cannot say precisely what, or how much it is costing, or which fix would move the needle most.
Thematic's job is to close that gap on two fronts at once. It handles scale, chewing through volumes no team could read manually, and it handles the qualitative-to-quantitative jump, converting "people keep mentioning the new login flow" into a sized, sentiment-scored, metric-linked theme you can rank against everything else on the list.
The pitch, in one chart: shrink the orange bar so the rest of the month can go to the work that matters. Figures illustrative.
How it differs from the incumbents
Thematic competes in a category dominated by large experience-management suites - Qualtrics and Medallia among them - along with newer text-analytics players like Chattermill and Keatext. The suites are broad: they run the surveys, host the dashboards and sell to the whole enterprise. Thematic is narrower and goes deeper on the one hard part, the reading. Where a suite hands you a dashboard and leaves interpretation to you, Thematic aims to hand you the interpretation, with the evidence attached.
The research pedigree is not just marketing. Thematic leans on its founder's academic background to argue for a "researcher's approach" to AI - one that shows its work rather than asking you to take the output on faith. In 2025 the company extended that idea outward, adding Model Context Protocol integrations so its customer intelligence can be pulled directly into assistants like Claude, ChatGPT and Microsoft Copilot. The hard-won knowledge about what customers think starts to live inside the AI tools teams already use all day.
One trusted view of your customer, built from the bottom up across every channel.The line on the homepage
The business behind it
Thematic sells the way most B2B software does: annual SaaS subscriptions to enterprise teams, generally priced by data volume, sources and seats, with an optional managed reporting service for programs that want the analysis delivered rather than run in-house. The company is small for its client list - around 20 people, fully remote across New Zealand and US time zones - and has stayed lean, raising a $1.2M seed in 2017 through Y Combinator and angels rather than a long chain of venture rounds. Third-party estimates put annual recurring revenue in the low single-digit millions.
The culture reads off its four stated values - transparency, ambition, collaboration and ownership - and the practical fact of being a day ahead of most of its customers. Being SOC 2 Type II certified matters here too; selling feedback analytics to banks and large retailers means the security posture is part of the product, not an afterthought.
Where it fits
Zoom out and Thematic sits at a useful seam in the market: between the survey tools that collect feedback and the BI dashboards that report on structured numbers, there is a layer that has to make sense of everything written in between. That layer is where Thematic lives. As more companies wire AI assistants into their workflows, the value of a clean, trustworthy, well-organized read of the customer voice goes up, not down - a chatbot is only as good as the ground truth it stands on.
The company's story is quieter than most AI headlines. A researcher who found a real problem, went deeper on it than most people cared to, and let the rigor be the pitch. The spreadsheet nobody wanted to open is still there. The difference is that now something reads it.