Repsense
Havel, the platform published by Lithuanian company Repsense, UAB, reads text, broadcast, social and video in 75+ languages. It scores the impact of every mention, detects coordinated campaigns and projects their trajectory for enterprises, defence institutions and governments.
What is Repsense?
Havel is the platform published by Repsense, UAB. The marketing site is repsense.io; the application itself runs on app.repsense.io. It ingests what the company calls the full information environment — text, broadcast, social, video, advertisements and search results — across 75+ languages and 60,000+ sources, with announced collection latency of under 15 minutes for web and social and under two hours for broadcast.
At the centre sits the Adler model, named after Alfred Adler. It combines Source Authority, Content Quality, Brand Prominence and Sentiment Amplitude into a single score running from -100% to +100%, and applies it only to content matching a narrative the client has defined in advance. The argument is deliberately contrarian: most media monitoring platforms collect content and count it; Havel collects content and evaluates it. Repsense illustrates the point with its own figures — roughly 6% of a betting brand's mentions match its narratives, around 40% for a bank.
Video is treated as primary material rather than a transcript. Three layers stack up: speech-to-text, a large language model (Gemini 2.5 Flash) prompted for the use case, and voice recognition with speaker diarisation, tracking up to 100 pre-recorded voices and auto-labelling the 1,000 most frequent ones. Each video yields six components: title, a 75-125 word summary, a transcript timestamped to 0.1 second, on-screen text, metadata, and ratings across five dimensions.
Coordination detection looks for synchronised posting, duplicated content across platforms and amplification architectures, separating organic conversation from a manufactured operation. Prediction is structural rather than volumetric: trajectory is projected from behaviour — coordination intensity, cross-platform diffusion, source authority — with 90% accuracy claimed on validated datasets.
Eight modules share one analytical base: Impact (Adler), Media Monitoring, Online Reputation (Bernays, which classifies Google, Bing, Yandex and Yahoo results by type), Politics, Disinformation, Broadcast Compliance, Advertising & Deep Analysis, and KPIs & Ambitions. Sentiment, labelled Context, is computed sentence by sentence on the same scale and presented as a secondary metric, while Communication Effectiveness is defined as Impact multiplied by Contacts. Output reaches users through the Havel web application (React, mobile-optimised), an embedded Metabase environment exporting CSV, XLSX, JSON, PNG and PDF, interactive HTML analyst reports, and digests and alerts pushed to Slack.
What it does
- Define what matters: countries, threats, brands, executives, influencers
- Track every medium and network, including the video other tools miss, with alerts and digests included
- Detect narratives, networks and inauthenticity, then predict where a story is heading
- Act: trace content forensically back to its source, commission AI analyst reports, run simulations
- Score the reputational impact of every mention on the Adler scale, from -100% to +100%
- Discover storylines automatically through unsupervised clustering
- Trigger smart alerts on a combination of entity, narrative and coordination
When to use Repsense / When not to
A quick filter to help you decide if Repsense is the right fit.
When to use Repsense
- StratCom, defence and government teams tracking foreign information manipulation and interference (FIMI) operations
- Corporate communications leaders and CCOs carrying live reputational risk, particularly where exposure plays out in short-form video (TikTok, Reels, Shorts) that text-only tools cannot read
- Public institutions measuring their own visibility inside a foreign media environment
- Agencies and consultancies reselling narrative intelligence to their own clients under their own brand
- Regulators and broadcasters that must evidence political neutrality and gender balance on air
When not to use Repsense
- Buyers who need transparent, self-service pricing: nothing is published and every path leads to a demonstration
- Individuals, freelancers and small businesses: the named references are NATO StratCom COE, the European Parliament, governments and large enterprises
- Anyone who needs a native mobile app: Havel is a React web application described as mobile-optimised, with no app store listing
- Teams that want narrative wargaming today: that module is a beta open only to selected clients
- Teams looking for a publishing or community management tool: Havel analyses and alerts, it does not post
How to use Repsense
A typical end-to-end flow, from setup to results.
- Request a demonstration from the home page anchor, or book one through the contact page: there is no self-service sign-up
- Define the entities to be tracked with the Repsense team — brand, executives, products, keywords
- Set the narratives that matter, either the seven supplied by default (employer branding, customer success, leadership, product success, business success, ESG, scandals) or custom ones
- Use the configuration switcher to follow several brands or several markets from the same account
- Sign in through the login page; the platform itself lives on app.repsense.io
- Work the live mention feed daily: filter by narrative, search entities and keywords with boolean operators, open the embedded video players
- Turn on deep video analysis where it is needed — the site describes it as activated on configuration
- Read the morning digest of QA-reviewed mentions, and set operative or smart alerts for what cannot wait
- Drill from any Metabase dashboard segment down to the underlying mentions, then export what you need
- Commission AI analyst reports, delivered after human prioritisation and supervision; partners can also export data via API and MCP
Pros & Cons
Pros
- Native video analysis: TikTok, Reels, Shorts, broadcast and advertising are read at content level, where most tools stop at text or a raw transcript
- Institutional references named openly on the site: NATO StratCom COE, the European Parliament, LRT the Lithuanian public broadcaster, and public bodies in Lithuania, Moldova, Germany, Romania and Armenia
- Prediction built on structural behaviour rather than volume extrapolation, with three days of lead time claimed in the Kapčiamiestis case of January 2026
- Narrative filtering that keeps decisions off the large majority of mentions the publisher itself counts as noise
- European publisher, GDPR compliance explicitly claimed, and analysis drawn only from publicly available sources
- Modular by design: eight modules on one shared analytical base, so scope can follow the use case
- Auditable dashboards: every chart is clickable down to the individual mentions that produced it
Cons
- No pricing is published at all: /pricing returns the site's standard 404 and the 180-URL sitemap carries no pricing page
- No terms and conditions: /terms and /legal return that same 404, leaving the privacy policy as the only legal document
- No public API documentation: /api and /docs return the 404 as well, and the API appears only as a partner benefit
- No security certification claimed — neither ISO 27001 nor SOC 2 — and /security does not exist either
- No list of sub-processors, only categories of recipients, and no hosting country or region declared anywhere
- No free trial and no free plan announced: the only way in is a demonstration
- The site's own FAQ is out of date, still describing “the platform Repsense.io” and referring to 2024 while the rest of the site says Havel; narrative wargaming also remains a restricted beta
Pricing & Plans
No price is published. Repsense operates a contact-sales model: there is no pricing page — /pricing returns exactly the same 404 as a made-up path, and the 180-URL sitemap contains none — and no amount is attached to the product anywhere on the site. No free trial and no free plan are announced. Every call to action leads to a conversation instead: “Request Demo”, “Book a Demo”, “Request a Sample” or “Apply for Beta Access”. A figure can therefore only be obtained by requesting a demonstration through the home page or the contact form and asking for a quotation. Note that the amounts visible elsewhere on the site belong to other subjects entirely: those in the FAQ are third-party market estimates used to illustrate an explanation, and those in the newsroom are investment rounds raised by the company. Neither set has any bearing on what Havel costs.
- the offer is described by module
- not by plan
- reputational impact scoring applied to every matching mention
- collection and tracking across text
- broadcast
- social and video
- search engine result analysis across Google
- Bing
- Yandex and Yahoo
- political and public affairs monitoring
- coordination and inauthentic behaviour detection
- neutrality
- speaker and gender analysis for broadcasters and regulators
- structured breakdown of TV
- radio and online advertising
- targets to reach in order to become or remain a leader
- a beta programme open on application
- not a commercial tier
- agency
- channel
- professional services
- risk and compliance
- media
- academic
- and referral
Data, GDPR & hosting
A consolidated view of how Repsense handles your data.
GDPR overview
GDPR compliance is claimed explicitly: the privacy policy states that Repsense “is guided by and complies with Regulation (EU) 2016/679”, alongside the Lithuanian law on the legal protection of personal data. Nine rights are listed — information, access, rectification, withdrawal of consent, erasure, restriction, objection, portability and complaint — exercised through support@repsense.io, with a one-month response and possible proof of identity. The supervisory authority is named: the Valstybinė duomenų apsaugos inspekcija (State Data Protection Inspectorate), L. Sapiegos g. 17, LT-10312 Vilnius, ada@ada.lt. Transfers outside the EEA are anticipated and framed by the European Commission's standard contractual clauses. The online version is effective from 4 April 2026. As the publisher is established in Lithuania, no Article 27 representative is required or designated. This is a stated commitment, not a certification.
Who owns the data?
Repsense, UAB acts as data controller for its own processing, while clients are described in the privacy policy as independent data controllers for the analytics, reports and dashboards delivered to them. The material analysed comes exclusively from publicly available sources on the internet: Repsense states it “does not intentionally collect data from private, restricted or non-public sources”. Account data — name, email, encrypted credentials, usage data — is processed to perform the contract under Article 6(1)(b) GDPR. No terms and conditions are published anywhere on the site, so no ownership or intellectual property clause covering client data can be read before signing.
Reuse rights
No terms of use are published, so what a client may do with the analytics, reports and dashboards delivered to them — reuse, republish, resell — cannot be established from the site. Only the privacy policy speaks, and it declares six purposes: monitoring and analytics services; client relationship management; communication; website operation and statistics; user accounts and platform access; and recruitment. Monitoring rests on legitimate interest under Article 6(1)(f) GDPR. Automated analytical tools, including AI-based solutions, are applied to publicly available information, and Repsense states that “such processing may qualify as profiling within the meaning of Article 4(4) of the GDPR” while adding that it “does not carry out automated decision-making within the meaning of Article 22”. Results are used only for aggregated analysis and client reporting, never to take a decision about an individual. Training models on customer data is never mentioned anywhere on the site: it is neither promised nor ruled out.
Data retention & training
Hosting summary
No hosting country or region is declared anywhere on the site. The privacy policy names only categories of recipients — “IT, hosting and cloud service providers”, “payment service providers”, “IT and CRM system providers” — and never an individual sub-processor; searching the collected pages for the term returns nothing at all. The policy does anticipate transfers outside the European Economic Area, stating that Repsense “may transfer your personal data to entities outside the European Economic Area (EEA) on a lawful basis and for a lawful purpose”, framed by the European Commission's standard contractual clauses. That is a transfer clause, not a hosting declaration: it says data may leave the EEA, not where it normally sits. Technical signals exist but are never confirmed by the publisher: the marketing site resolves to a Squarespace anycast address in the United States, and the application on app.repsense.io is served alongside a separate CDN hostname. Neither tells you where the Havel platform runs. A buyer with a data residency requirement will have to obtain that answer contractually.
Things to keep in mind
Risks and trade-offs to weigh before adopting Repsense.
- Self-declared performance: 90% trajectory accuracy, three days of lead time in the Kapčiamiestis case and sub-15-minute collection are the publisher's own figures, with no third-party audit. Treating a prediction as a measurement is the first risk here
- No terms and conditions are published anywhere on the site, so usage conditions, liability and client rights over delivered analytics remain unknowable until a contract is on the table
- Model training is never addressed: the site neither commits to anything nor offers an opt-out, and the subject does not appear once across its pages. Silence is not a guarantee
- The tool profiles public speech. Repsense states the processing may qualify as profiling under Article 4(4) GDPR while ruling out automated decision-making under Article 22, but a human still reads the dashboard, and a single score attached to a named person invites faster judgement than the underlying evidence supports
- Two divergent first-party addresses coexist: the contact page gives Saulėtekio alley 15, LT-10224 Vilnius, while the privacy policy gives the registered address V. Nagevičiaus str. 3, LT-08237 Vilnius
- The FAQ is a stale layer that still names “the platform Repsense.io” and refers to 2024, and its meta description promises answers on languages, deployment and security that appear nowhere in the page; the G2 badge on the home page could not be verified either
- Third-party statistics quoted on the site (Reuters Institute, World Economic Forum, Gartner) are not Repsense measurements, and Gemini 2.5 Flash is a technical sub-processor inside the video layer, not a client-facing integration: reading either as a product claim would be a mistake
Setup & Integrations
Technical difficulty
Low technical difficulty, but no autonomy. There is nothing to install: Havel is a web application reached at app.repsense.io. Setup is assisted — the client works with Repsense to define its entities and custom narratives, with seven standard narratives supplied as a starting point. Dashboards are built per client, module or use case, and hands-on onboarding, working sessions with analysts and methodological support are cited among partner benefits. The real effort is analytical rather than technical: deciding which narratives matter. Access itself is gated by a demonstration, so no one sets the tool up alone.
Deployment
Integrations
Behind Repsense
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Repsense.
Frequently asked questions
Who publishes Havel?
What is the Adler model?
How many languages and sources are covered?
Is video really analysed, or just transcribed?
How much does it cost?
Is there an API?
Is there a mobile app?
Is it GDPR compliant?
How long is data kept?
Who founded the company, and who uses the platform?
Should you pick Repsense?
Repsense is a European narrative intelligence proposition, built by a 25-person Lithuanian company founded in 2022 and shipped under the product name Havel. Its stated differentiator holds up: rather than counting mentions, it reads video content and hunts for coordination, which brings short-form platforms and broadcast inside the analysis instead of leaving them outside it. The Adler model, applied only to content matching a client-defined narrative, follows the same logic — it trades reach for relevance, and the company says so openly.
Credibility rests on named institutional deployments the site does not hide behind anonymity: NATO StratCom COE, the European Parliament, LRT, and public bodies across several countries. For a buyer in strategic communications, defence or public administration, that reference base weighs more than any feature list.
The main reservation is commercial opacity, and it is complete. No pricing, no terms and conditions, no API documentation and no security certification are published; the only legal document on the site is a privacy policy. A prospect cannot judge what they would be signing without entering a sales conversation, and an engineer cannot scope an integration without one either.
The performance figures deserve to be read as claims rather than measurements. Ninety percent trajectory accuracy, three days of lead time in the Kapčiamiestis case, collection in under fifteen minutes: all are stated by the publisher, none is audited publicly, and the G2 badge shown on the home page could not be verified. Havel is a serious option for organisations that already know they need this class of tool — but it is not a product anyone can evaluate, or even price, from the website alone.
- Choosing a selection results in a full page refresh.
- Opens in a new window.