Graph and LLM ranking of journalists and influencers for media relations

Media relations lists go stale. Beat journalists change desks; creators shift platforms; relevance to a club, federation, or event is dynamic.

Dantani ranks journalists and influencers with a graph of bylines, quotes, topic affinity, and audience overlap, then uses LLMs to summarise why a contact fits a given campaign—without fabricating follower KPIs. Sports PR remains the decision layer.

Spray-and-pray outreach wastes access. Ranked shortlists grounded in recent coverage behaviour beat static Excel lists every transfer window and every tournament.

Motivation

Communications teams need ranked shortlists grounded in recent coverage behaviour and topical fit. A transfer-window specialist who moved outlets still matters; a creator who left the sport's orbit does not. Graph recency captures that; follower screenshots do not.

Regional language journalists often cover federation policy more deeply than global wires. Ranking that ignores language and outlet tier systematically underserves those relationships.

Methodology

A bipartite graph links writers and creators to stories, outlets, and entities. Centrality and recency features feed a ranker. LLM summaries explain ranking drivers for human media managers who remain the final decision layer.

Journalist and influencer feature mixes differ by design. Outlet prestige and beat consistency weigh more for journalists; graph overlap with campaign audiences weighs more for creators. Dantani does not invent follower totals when platforms withhold them.

Suppression lists matter: adversarial or recently hostile contacts may still rank on topical fit and require proactive briefing rather than pitch spam. Ranking is relevance, not affection.

Desk workflow

• Campaign brief: topics, entities, languages, and do-not-contact rules.
• Rank: top contacts with explainable drivers.
• Human trim: media managers cut and reorder.
• Outreach: personalised angles from LLM fit summaries, edited by humans.
• Feedback: log responses to retrain operational priors—not vanity open rates as company KPIs.

Koba inside Eagle Eye dashboards drafts fit blurbs from graph features. Terminal, coming soon, will sync ranked lists into CRM workflows clubs already run.

Schemas and code

type ContactNode = {
  id: string;
  kind: "journalist" | "influencer";
  outlets: string[];
  languages: string[];
  lastActiveAt: string;
};

type RankExplanation = {
  contactId: string;
  score: number; // ranker output for ordering — not a public KPI
  drivers: string[];
  fitSummary: string;
};

def rank_contacts(graph, campaign_topics, llm):
    scores = graph.score(recency=True, topics=campaign_topics)
    explain = {
        c: llm.summarise_fit(c, campaign_topics)
        for c in scores.top(20)
    }
    return scores, explain

Failure modes

• Stale CRM: contacts ranked after they left the beat.
• Follower worship: missing specialists with smaller but decisive audiences.
• Language blind lists: ignoring regional journalists on federation stories.
• Hostile contact spam: high relevance without strategy.
• Fabricated reach: inventing audience size when platforms omit it.
• LLM flattery: fit summaries that overclaim relationship warmth.

Examples

• Transfer-window specialists who consistently break club-adjacent stories.
• Regional language journalists covering federation policy more deeply than global wires.
• Creators whose audience overlaps a tournament's core demographic without official accreditation.
• Investigative reporters who should receive proactive briefings during integrity stories.
• Podcast hosts whose transcripts move talk-radio narratives after coaching changes.

Where Dantani products fit

Sports PR runs outreach. Monitoring supplies the byline and entity graph. Terminal, coming soon, syncs ranks to CRM. Koba inside Eagle Eye dashboards writes explainable fit notes for media managers.

Relationship strategy on top of rank

Rank explains relevance; strategy explains contact. Investigative reporters covering integrity may top a topical rank and still require briefing rather than product pitches. Creators near a tournament demographic may need embargo discipline the graph cannot infer alone.

Feedback loops should log outcomes desks care about—briefings accepted, corrections landed, embargoes respected—without converting email opens into company-wide vanity KPIs. Eagle Eye supplies graph features; Sports PR supplies judgement.

Language and outlet tier filters belong in the campaign brief before ranking runs. A federation policy campaign that forgets regional-language journalists will over-index global wires every time. Dantani keeps those filters explicit so the graph does not quietly recreate Anglophone bias.

Takeaways

Media lists should be living ranked graphs. Dantani gives sports PR teams explainable shortlists so outreach matches beat reality.