How Media Intelligence Works
How Gumshoe finds, scores, and surfaces news coverage for the entities you verify — including how GDELT works, what tone scoring means, and why NA is normal for most private companies.
Every Gumshoe supplier verification includes a Media Intelligence tile — a 30-day snapshot of public news coverage for the entity you're verifying. Here's how it works.
What media intelligence does
The media tile searches for recent news coverage of the supplier and returns up to 12 articles, each with a source, date, headline, link, and tone score. If significant negative coverage is found, the tile flags WARN. If no coverage is found (the most common result for private companies), the tile returns NA.
The two sources: GDELT and Google News AU
GDELT (Global Database of Events, Language and Tone) is an open-source project that monitors news in real-time across 65,000+ global sources, including all major Australian outlets — ABC, AFR, The Australian, SMH, The Guardian Australia, Herald Sun. GDELT is the primary source. It includes a machine-generated tone score for each article.
Google News AU RSS is used as a supplemental source, scoped to Australian results. It helps capture regional and trade press that GDELT may index more slowly. Both sources run in parallel; results are merged and deduplicated by headline similarity before being returned.
Waterfall matching: chains and brand names
Legal entity names are rarely how journalists write about companies. "WOOLWORTHS ST HELENA PTY LTD" won't return news coverage — but "Woolworths" will. The media tile uses a waterfall search strategy:
- Search the normalised entity name (legal suffixes stripped)
- If fewer than 3 results, try the brand name (corporate words stripped, first 1–2 words only)
- If still fewer than 3 results and the entity looks like a chain branch, try the inferred parent brand
- Also check any ABR trading names for the entity
When coverage is found via a waterfall level other than the exact name, the tile displays a note explaining which level matched (e.g., "Coverage found via parent brand — no exact-match results for the registered entity name").
Tone scoring
GDELT assigns every article a tone score from -100 (very negative) to +100 (very positive). We use this as the primary sentiment signal. Where GDELT's tone data isn't available for an article (e.g., from the Google News RSS supplement), we apply keyword-based enrichment: articles mentioning fraud, penalty, ACCC, breach, recall, conviction, or similar terms are marked negative; articles mentioning awards, growth, or expansion are marked positive.
The tile aggregates tone across all articles and calculates an average. If 60% or more of articles have a negative tone score (below -2), the tile status becomes WARN. Otherwise it's PASS (coverage found) or NA (no coverage).
When to expect NA
NA is the most common result and is not a red flag. Private companies, sole traders, B2B service providers, and any entity without a public media profile will return NA. GDELT primarily indexes English-language outlets; very small or niche entities are unlikely to appear.
NA means: no relevant news coverage was found in the past 30 days across the sources we monitor. It does not mean the entity is risky or suspicious — it simply means the media hasn't written about them recently.
Limitations
- GDELT indexes primarily English-language content; non-English coverage is not surfaced
- The lookback window is 30 days — older coverage is not included in the current tile
- Very small or niche entities with no public media profile will almost always return NA
- Tone scoring is machine-generated and may misclassify nuanced or ironic coverage
- Google News RSS results are not personalised but are algorithmic — relevance can vary