A customer writes, “I’m furious - there’s none left.” A conventional sentiment dashboard sees anger, paints a red bar, and nudges the brand toward damage control. A human sees something else: demand. The product is not being rejected. It is sold out. The sensible move is not an apology; it is another production run. Tellagence has spent most of its life in the expensive little gap between those two readings.
The Portland company calls that gap contextual intelligence. Its software ingests language from social media, creator communities, reviews, forums, surveys and other digital channels, then looks for the narratives, motivations and shifts beneath mention counts. The current product, Tellagence Discover, packages the work in dashboards, custom reports and an API, with researchers available to help design a study or turn the output into a decision.
This is not a fresh company riding a fresh acronym. Co-founders Matt Hixson and Nitin Mayande started Tellagence in 2011. Hixson had worked in enterprise software and run social initiatives at Tripwire. Mayande was researching how information moved through online communities. Their founding hunch was unfashionably specific: popularity is not influence, and language is not separable from the relationships around it.
The first product chased the quiet person in the room
In 2013, influencer marketing was busy turning follower counts into a personality test. Tellagence went the other way. It modelled how information passed from person to person and tried to identify the small group positioned to carry a message through a particular network. Hixson offered a vivid claim at the time: the right ten people might spread word of mouth to 4,000. The point was not that ten was magical. It was that the loudest account and the most useful bridge were often different people.
The company offered Tellagence for Twitter, then two related products. Community found people likely to engage and transmit a message. The original Discover inspected a topic across as many as 250,000 recent tweets, surfaced connected themes, let analysts drill into subthemes, and exported the supporting posts. It cost $350 per use, discounted to $315 during beta. That is the cleanest public answer to “what did it cost?” in the early era.
What failed first was the shortcut. Keyword lists assumed the marketer knew the vocabulary before the research began. Follower counts mistook visibility for transmission. Binary sentiment flattened sarcasm, fan hype, slang and scarcity into positive or negative buckets. The record does not contain a melodramatic founder confession or one doomed launch. It shows something more useful: the same limitations being attacked from successive angles.
“Context is where we believe you start.”Matt Hixson, 2013
The market finally learned the company’s favorite word
Tellagence did not so much change its mind as widen the aperture. Its early question was who moves a conversation. The new one is why the conversation is moving. Modern Discover reads far beyond Twitter and is pitched to brand, agency, market-research, communications, customer-experience and strategy teams. It works across 140 languages, according to the company’s August 2026 launch announcement.
The core sales example is almost comically simple. A scan shows 14 percent negative conversation around a product. A basic tool says brand-health problem. Discover clusters the language and finds frustration about stockouts. The metric stays put; the recommended action flips. This is where Tellagence sits in the market: deeper than monitoring, narrower than a giant customer-experience suite, and more opinionated than handing raw text to a general-purpose chatbot.
Mayande, Tellagence’s chief science officer, has worked on the underlying problem since 2008. In April 2026, Tellagence researchers and collaborators from Villanova University and Portland State University published two papers focused on consistency and contextual text categorization. The company says its weighted syntactic and semantic framework was tested across more than 433,000 real-world reviews, reduced analytical variability from roughly 25 percent to below 3 percent, and reached 96 percent alignment with human analysis.
Company-reported 2026 research result. The useful buyer question is not whether the demo looks clever, but whether repeated runs remain stable on your own data.
That positioning matters. A generic model can summarize a folder of comments quickly. Brandwatch, Sprinklr, Talkwalker, Meltwater, NetBase Quid, Sprout Social and YouScan offer broad listening or consumer-intelligence suites. Tellagence is selling a context layer and a repeatability argument, plus human analysts when the buyer needs more than a dashboard. It is a focused shop in a market full of department stores.
The business model is an anti-seat-tax experiment
The 2026 pricing page does not list dollar amounts. It does explain the mechanism. Customers buy tokens based on the number of completed reports they expect to run. Collaborators are unlimited, unused credits roll over, features are not held behind enterprise tiers, and volume lowers the unit cost. Large users can arrange monthly or quarterly invoicing. Expert research and strategy services sit alongside the software.
That packaging is worth stealing. Analysis is collaborative: researchers, strategists, executives and clients may all need to inspect the same result. Charging for every viewer taxes circulation and quietly limits the product’s reach inside an account. Tellagence instead meters the expensive act - completed analysis - while making distribution free.
The company reports that early customers saved about $9 in internal labor for every $1 spent, largely by automating ingestion, clustering and deck production. Treat that as a vendor benchmark, not a law of nature. The return depends on data volume, analyst salaries, workflow complexity and how often the output changes a real decision. Scout Lab co-founder Willow Hill publicly called Discover a strategic tool in client work. Tellagence also names a longtime relationship with YouTube and Google focused on social applications and analytics. Most global-brand customers remain unnamed.
The organization itself is compact. LinkedIn places Tellagence in the 2-to-10 employee company-size band while showing 29 associated profiles, an imperfect but useful reminder that private-company headcounts are fuzzy. Its public roster mixes AI engineers with brand strategists, social analysts, data managers and security specialists. The biographies are disarmingly specific: Hixson likes art and sneakers; Mayande likes Bugs Bunny and takes his Peruvian light roast without milk or sugar.
What the funding bought - and what founders can copy
Tellagence was not built on a spectacular capital bonfire. Public databases disagree on labels and totals, but Dealroom records $25,000 in May 2012 from Rogue Venture Partners and Portland Seed Fund, another $30,000 seed check that September, and a $350,000 debt financing involving Portland Seed Fund in February 2016. The supplied company record puts total funding at about $410,000. That modest financing appears to have supported a long mix of research, software and client work rather than a blitzscaled land grab.
The copyable part is the architecture of the journey. Preserve the hard thesis, but let the commercial wrapper move. Publish the science when trust becomes the purchasing objection. Turn topical research into public proof-of-work: Tellagence’s Discover Pulse reports aim the system at cultural moments, from sports sponsorship to product disputes, and show the gap between surface sentiment and narrative. Finally, price the job completed rather than the employee who logs in.
Sell a stable underlying insight through whatever format buyers currently understand: product, report, API or service.
Use public analyses as honest demos. Show the reader the surprising business decision, not a cloud of model jargon.
Fifteen years of persistence only works when services or customers fund the learning and the technical thesis keeps producing value.
Your dataset is tiny, the language is clean, the stakes are low, or all you need is a mention count and a rough directional read.
There are clear conditions under which the Tellagence approach will not earn its keep. If a team has fifty plainly worded survey responses and a researcher can read them before lunch, a context platform is overhead. If the buyer cannot connect themes to inventory, creative, product or communications decisions, richer analysis becomes decorative. Multilingual nuance also deserves testing on the exact markets and dialects that matter, regardless of a platform-wide language count.
But ambiguity scales faster than headcount. Brands now collect reviews, posts, creator chatter, call transcripts and open-text surveys faster than analysts can read them. General-purpose AI makes summarization cheap while making consistency a fresh anxiety. Tellagence’s bet is that the scarce output is no longer an answer. It is an answer that survives a second run, preserves the odd little cluster everyone else discarded, and tells a decision-maker whether to call crisis communications or the factory.