Comms Radar
An intelligence platform for communications professionals. This guide explains how the tool works, what each section does, and how to get the most out of it.
What is Comms Radar?
Comms Radar is a private intelligence publication for communications professionals. It automatically monitors hundreds of sources — trade press, newswires, vendor blogs, research outlets — and surfaces signals that matter to PR, corporate communications, and CommsTech practitioners.
It is not a feed reader. The goal is not to show you everything — it is to show you what matters, ranked, grouped, and interpreted. Every article receives a relevance score. Every cluster of related articles receives an editorial summary.
Comms Radar runs on a private database. All data stays local unless you explicitly move it to a VPS.
The Three Intelligence Pillars
All intelligence is organised into three thematic pillars. Each article is classified into one pillar at ingestion time.
Corporate Comms Moves
Leadership changes, agency appointments, and strategic communications shifts. Who is hiring whom, which brands are changing agencies, which CCOs are moving.
Score field used: A-Score (emphasises named people and institutional names in the comms space).
CommsTech & Competitive Intelligence
Cision, Meltwater, Muck Rack, and adjacent vendors — acquisitions, product launches, funding rounds, and market moves. Anything that changes the competitive landscape of the tools we use.
Score field used: B-Score (emphasises vendor names, product terms, and funding language).
Studies, Associations & Standards
Research reports, industry benchmarks, association announcements, and evolving measurement standards. Content that shapes how the industry defines best practice.
Score field used: C-Score (emphasises research language, association names, and methodology terms).
Relevance Scoring
Every article receives a relevance score from 0–100 at ingestion time. The score is calculated by the intelligence engine using keyword matching, entity recognition, source quality weighting, and freshness.
Strong match. Directly relevant to the pillar.
Adjacent or partial match. Worth monitoring.
Weak match. Usually filtered from main views.
The scoring system is not perfect. PRNewswire press releases can score high due to keyword matches even when the content is irrelevant (e.g. a real estate article mentioning "communications"). This is a known issue being addressed in Sprint 1.
The Relevance Explanation field on each article detail page shows exactly how the score was calculated — which keywords matched, which source multiplier was applied, and what freshness bonus was added.
The Newsroom
The Newsroom (/) is the front page. It is structured like an intelligence publication, not a feed.
Intelligence Pillars strip
Top-scoring articles from each of the three pillars, last 72 hours. The number in the top-right of each pillar column is the total signal count for that window.
Lead Story (black area)
The highest-scoring Intelligence Cluster from the last 72 hours. Shows the relevance score, what happened, why it matters, what to watch next, and which companies are involved. The signal trajectory badge (Breaking / Accelerating / Building / Emerging) reflects how fast coverage is growing.
Intelligence Clusters grid
The next 5 top clusters. Each card shows the score, the "Why it matters" interpretation, the trajectory, and the outlets covering it.
Emerging Signals
The most recent 20 articles across all pillars. High-priority items (score ≥55) show a one-line editorial interpretation below the headline.
Radar sidebar
Real-time widgets: newly detected vendors, trending topics (24h), recently discovered sources, and pillar coverage counts.
Intelligence Clusters
Clusters are automatically generated groups of articles covering the same story or theme. When multiple sources report on the same event, the system clusters them and assigns a combined relevance score.
A cluster with coverage from 3+ independent outlets is a strong signal. A cluster with a single source may be a press release or vendor announcement — treat with more caution.
The cluster detail page (/clusters/[id]) shows all articles in the cluster, the sources covering it, and the full summary. Use it to quickly assess whether a story deserves your attention.
Dashboard & Filtering
The Dashboard (/dashboard) shows the full article stream with filters.
Article categories
Categories are assigned by the intelligence engine at ingestion time. They are not always accurate — the scoring explanation on each article shows how the category was determined.
Discovery & Taxonomy
The Discovery section (/admin/discovery) manages the Known Universe — the set of companies, vendors, topics, and entities the system actively tracks.
What are taxonomy candidates?
Every time the ingestion engine encounters an entity (company name, vendor, topic) that is not yet in the Known Universe, it creates a candidate. Candidates appear in the Discovery page for review. You decide whether to promote them to Known Universe, add to Watch List, or reject them.
Status lifecycle
Promoting an entity to Known Universe does not retroactively rescore old articles. It affects scoring of articles ingested after the promotion.
Sources & Ingestion
Sources are the RSS feeds, websites, and APIs the system monitors. Manage them at /admin/sources.
How ingestion works
- Each source RSS feed is fetched
- New articles (not previously seen) are extracted
- The intelligence engine scores each article against all three pillars
- Entities are extracted and matched against the Known Universe
- Articles above the minimum threshold are stored
- Clustering runs to group related new articles with existing clusters
Source quality multiplier
Each source has a quality multiplier (0.3–1.5) that scales the raw relevance score. PRNewswire and press release aggregators have a low multiplier (0.4) because they publish high volumes of low-relevance content. Tier-1 trade press has a higher multiplier.
Ingestion currently runs manually via the Admin panel. Automated scheduled ingestion is on the roadmap (Sprint 4).
Admin Tools
The Admin panel contains several maintenance tools. Here is what each one does.
What it does: Fetches all active sources, scores new articles, extracts entities, and runs clustering. Takes 1–5 minutes depending on source count.
When to use: Run manually when you want fresh intelligence. Will be automated in a future release.
What it does: Removes articles below the minimum relevance threshold that may have been stored before the threshold was raised.
When to use: Run after raising the intelligence engine threshold to clean up old low-quality articles.
What it does: Runs entity extraction on all stored articles that have not yet had entities extracted. Used after adding new entity types.
When to use: Run after significant changes to the entity extraction configuration.
What it does: Analyses recent article content to surface new candidate entities for the Known Universe. Creates new Candidate entries in Discovery.
When to use: Run weekly to keep the Known Universe growing.
What it does: Pushes the current Known Universe state to the scoring engine so it uses the latest entity list for future ingestions.
When to use: Run after promoting or rejecting entities in the Discovery page.
What it does: Generates a summary report of the past week's intelligence across all pillars.
When to use: Run every Monday morning for a weekly briefing.
What it does: Scans known article content for new vendor names and source suggestions, adding them as discovery candidates.
When to use: Run when you want to expand source coverage.
Known Issues & Roadmap
This is an actively developed tool. The following issues are known and being worked on. See the Changelog for what has already been fixed.
- —PRNewswire false positives — press releases with zero relevance scoring high due to keyword matches
- —Jobs category showing People Moves, not actual job listings — paused pending source rework
- —'Why it matters' text describes tool relevance, not industry significance — needs editorial rethinking
- —Remaining German-language strings in the UI
- —Snippets and summaries not visible in all stream views
- —Newsroom density and hierarchy needs refinement
- —Thematic pillar pages show article lists, not editorial layouts
- —Inline relevance score editing with learning feedback loop
- —Admin tools need descriptions and confirmation dialogs
- —Crawl Logs need per-source error detail
- —KPI dashboard for LLM usage, database size, and costs
- —Companies admin: no UI to add suggestions directly
- —Source discovery workflow (suggest → review → add)
- —Automated scheduled ingestion (no manual Run Ingestion button)
- —Social media monitoring (Bluesky / Twitter account tracking)
- —Discovery page UX redesign — clarify that these are taxonomy candidates, not headlines
- —Auto-updating changelog after each deployment