The Havel Platform
Havel Overview
One European-built platform that ingests the full information environment — text, broadcast, social, video, advertisements, and search — across 75+ languages. It analyses narrative structure, coordination patterns, impact, and trajectory, then delivers intelligence through dashboards, alerts, and AI-generated reports. From raw media to actionable intelligence in one architecture.
<2hrs
colection speed (broadcast)
sources
60,000+
languages
75+
collection speed (web/social)
<15 min
How Havel Thinks
Not All Mentions Are Equal. That’s the Starting Point.
Most media monitoring platforms collect content and count it. Havel collects content and evaluates it. Every piece of content that enters the platform is scored across multiple dimensions before any human sees it: what is its source authority, how well is it constructed, how prominently does it feature the entity being tracked, and does it match any of the narratives that actually matter for the client’s decision-making?
This is the Adler Impact Model - Havel’s core analytical engine. Named after Alfred Adler. Impact combines Source Authority, Content Quality, Brand Prominence, and Sentiment Amplitude into a single score from –100% to +100%, but only for content that matches pre-defined narrative frameworks. Content that doesn’t match any narrative is still visible in the platform - but when it comes to the metrics that matter for strategic decisions, it is treated as what it is: noise.
A platform that distinguishes between an article entirely about your brand and a passing mention in a football recap. Between a TikTok review that shapes purchase intent and a bot repost. Between a hostile narrative that will damage reputation and a negative article that will be forgotten by tomorrow. In practice: not counting mentions, but evaluating what each mention means.
What Goes In
Every Source. Every Format. Every Language.
Traditional Media
News agencies, web articles (including paywalled content under agreement), television shows, radio shows, print media (120+ publications in Lithuania alone), forums, and podcasts. Metadata extracted down to author, punctuation, and content structure.
Social Media
Facebook pages and comments, YouTube (descriptions, comments, and deep-analysed Shorts), Instagram (posts, hashtags, and deep-analysed Reels), TikTok (descriptions, hashtags, and deep-analysed videos), Telegram, X, Reddit, and LinkedIn.
Deep Video & Audio
TikTok videos, Instagram Reels, YouTube Shorts, broadcast segments, podcasts, and advertisements - processed through speech-to-text, LLM-powered deep analysis, and voice recognition. Not just transcript extraction: structured intelligence including timestamped transcription, visual text, brand/product detection, persona identification, and narrative classification.
Deep Video & Audio
TikTok videos, Instagram Reels, YouTube Shorts, broadcast segments, podcasts, and advertisements - processed through speech-to-text, LLM-powered deep analysis, and voice recognition. Not just transcript extraction: structured intelligence including timestamped transcription, visual text, brand/product detection, persona identification, and narrative classification.
Go deeper: Deep Video Analysis
Advertisements
TV, radio, internet, and outdoor advertisements ingested and deep-analysed: brands, products, offers, personas, colours, production quality - structured and searchable alongside editorial content.
Search Engines
Google, Bing, Yandex, and Yahoo SERPs scraped and scored for online reputation analysis. Each search result classified by type (own, news, review, wiki, social, competitor) and scored for impact on entity perception.
What Happens Inside
Models, AI, and Data — Working Together
Narrative Analysis
Content transformed into text embeddings that capture semantic meaning. Matched against pre-defined client narratives using cosine similarity. Content that matches is scored for impact. Content that doesn’t is visible but deprioritised. This is the mechanism that solves the valuation problem: only ~6% of a betting brand’s mentions match communication narratives; for banking, ~40%. Without this filter, strategic decisions are based on 60–94% noise.
Coordination Detection
Behavioural analysis identifies coordinated inauthentic activity: synchronised posting, cross-platform content seeding, amplification patterns. Distinguishes organic conversation from manufactured operations. The same detection that identified state-sponsored FIMI campaigns, applied to commercial narrative threats.
Context (Sentiment)
Numeric sentiment from –100% to +100% on a continuous scale - not positive/neutral/negative buckets. Calculated at the sentence level, weighted toward sentences that mention relevant entities. We call it Context, not Sentiment, because it describes the surroundings of your entity, not a direct evaluation. A secondary metric, not the primary one.
Communication Effectiveness
Impact (qualitative) multiplied by Contacts (quantitative). The metric that answers: how many people were meaningfully reached - beyond potential exposure. This is what KPIs should be built on.
Storylines Discovery
Unsupervised clustering detects groups of semantically similar content forming in real time - topics and storylines emerging from the data without pre-definition. Each cluster is named, described, and summarised automatically. This is how Havel finds the stories you didn’t know to look for.
Predictive Intelligence
Narrative trajectory projected from structural behaviour - coordination intensity, platform spread, source authority distribution - not volume extrapolation. Early warning before conventional tools trigger. 90% trajectory accuracy across validated datasets.
Go deeper: Predictive Intelligence
Brand Prominence
Every mention scored for how prominently the entity features: Primary (the article is about you), Secondary (you play a meaningful role), or Episodic (you’re mentioned in passing). Competitors treat all three the same. Havel doesn’t.
What Comes Out
Intelligence, Not Dashboards
The Havel App
A web application with live mention feed, narrative filters, entity and keyword search with boolean operators, impact scoring on every mention, embedded video players, and mobile optimisation. Configuration switcher for multi-brand or multi-market monitoring. The interface a social media manager uses daily and a CCO reviews weekly.
Analytics Environment
Embedded dashboards with metrics, narratives, metadata, discovery, interactions, and storylines tabs. Every chart is interactive and transparent: click any segment to see the underlying mentions. Exportable to CSV, XLSX, JSON, PNG, or PDF. Dashboards personalised per client, module, or use case. The analytical depth of a BI tool, embedded in the monitoring application.
AI Analyst Reports
Structured intelligence reports generated from any narrative cluster, alert, or investigation - in minutes. Interactive HTML deliverables with narrative analysis, source attribution, sentiment trajectory, and recommendations. Not dashboard summaries. Ready to publish intelligence.
Go Deeper: AI Analyst Reports
Notifications & Alerts
Daily digests every morning with QA-reviewed mentions. Operative alerts on any new mention or smart alerts triggered by specific entity + narrative + coordination combinations. Weekly and monthly summary emails with key metrics, competitive comparison, and top storylines. Slack integration for real-time alerts. All configurable, all automated.
Modules
One Foundation, Many Applications
-
The core reputation and narrative impact model. Scores every mention for source authority, content quality, brand prominence, and sentiment. The foundation everything else builds on.
-
Real-time data collection and analysis across traditional and social media. The main commercial product. Companies track their brand, competitors, and industry keywords.
-
SERP analysis across Google, Bing, Yandex, Yahoo - scoring every search result for its impact on entity perception. Reputation index, competitive benchmarking, and actionable improvement insights.
-
Media monitoring with extra attention on social media discovery, party narratives, power ratings, and specialised landscape analytics. Election monitoring and political balance auditing.
-
Cross-border narrative tracking across 100+ foreign regions per country. Currently operational for Lithuania, Latvia, and Estonia. 80–90K mentions/month per configuration. Early threat detection with spike alerts and storyline analysis.
-
Voice recognition, speaker identification, gender classification, and political neutrality monitoring for television, radio, and podcasts. Deployed with LRT.
-
LLM-powered decomposition of TV, radio, and online advertisements into structured objects: brands, products, offers, personas, colours, and quality ratings.
-
Calculates what targets a brand needs to hit - mentions, contacts, narrative coverage, hero mentions - to become or remain the market leader. Including how many press releases it would take.
See Havel in Action
Request a demo to see the platform configured for your use case - media monitoring, narrative threat detection, broadcast compliance, or any combination.
Or explore specific capabilities:
How Havel comprehends video at the content level
Structural prediction and early warning
Live examples of intelligence report output