Campaign sentiment that only watches short-form social misreads sports PR. News desks, talk radio transcripts, and fan forums often diverge in tone from short-form social.
Dantani builds cross-channel sentiment models that calibrate per channel, then roll up to campaign-level views with explicit caveats when channels disagree. Eagle Eye classifiers power the axes; Sports PR acts on divergence, not on a single blended vanity score.
A kit launch can look loved on social while specialist press remains sceptical. Communications leaders need disagreement made visible, not averaged away.
Motivation
Averaged sentiment hides the decision. If social is euphoric and news is cold, the next creative and the next press briefing need different moves. Federations announcing policy see the same pattern: wires read neutral while social polarises.
Esports and creator-heavy campaigns amplify the gap. Mainstream sports desks may ignore a partnership that community channels saturate. Cross-channel views stop desks from declaring victory on the loudest feed alone.
Methodology
Channel-specific classifiers handle news prose, social slang, and broadcast transcript style. Scores stay on separate axes until a reconciliation layer flags divergence—only after each channel index is scaled to a shared comparable range. LLMs summarise drivers of disagreement for analysts.
Calibration is per channel and per sport context. Lexicons and few-shot exemplars differ for match reaction versus policy announcement. Dantani does not publish a universal accuracy percentage; desks get divergence flags and driver summaries.
Campaign objects bind channels to a shared entity and time window so week-over-week views stay comparable without pretending raw channel scores share one native scale.
Desk workflow
- Campaign setup: entities, window, and channels in scope.
- Daily: per-channel indices with divergence flag after scaling.
- When flagged: LLM driver summary plus sample items for human read.
- Response: adjust creative, press angle, or community posts by channel—not one blanket line.
- Close-out: record where channels disagreed and what changed.
Koba drafts divergence narratives for campaign stand-ups. It narrates disagreement across axes; it does not manufacture a blended confidence score.
Schemas and code
Illustrative pseudocode
type ChannelScore = {
campaignId: string;
channel: "news" | "social" | "broadcast";
window: string;
index: number; // channel-local scale — not a cross-channel KPI
};
type DivergenceReport = {
campaignId: string;
spread: number; // on scaled axes
flag: boolean;
drivers: string[];
};Illustrative pseudocode
def scale_channel(index, channel_stats):
# Map each channel onto a shared comparable range before diverge checks.
loc, scale = channel_stats.loc, channel_stats.scale
if scale == 0:
return 0.0
return (index - loc) / scale
def reconcile_sentiment(channel_scores, channel_stats, threshold=1.5):
scaled = {
ch: scale_channel(v, channel_stats[ch])
for ch, v in channel_scores.items()
}
spread = max(scaled.values()) - min(scaled.values())
return {
"spread": spread,
"flag_divergence": spread >= threshold,
}
def driver_summary(campaign_id, llm, samples):
return llm.summarise_disagreement(campaign_id, samples)Failure modes
- Single-channel worship: declaring success from social alone.
- Naive averaging: blending incompatible scales into one "score."
- Irony and meme blindness: classifiers missing sarcastic praise.
- Transcript gaps: talk radio unmonitored while social dominates the view.
- Campaign window drift: comparing mismatched weeks across channels.
- Invented accuracy: selling sentiment models with fake F1 claims.
Illustrative examples
The patterns below are illustrative campaign shapes—not named production results.
- Kit-launch weeks where social hype outruns press quality critiques.
- Federation-policy announcements that read neutral in wires and polarised on social.
- Coaching appointments praised on talk radio while fan forums stay split.
- Esports partnership reveals that creator channels amplify while mainstream sports desks ignore.
- Crisis recoveries where news softens before social trust returns—or the reverse.
Where Dantani products fit
Sports PR campaigns run on divergence-aware views. Monitoring supplies channel-calibrated scores. Analytics exports stay per-channel so engineers never collapse axes into one marketed number.
Acting on divergence
When social leads and news lags, creative teams and press teams diverge on purpose: community posts can acknowledge energy while press briefings address quality critiques. When news softens and social stays cold after a crisis, trust work continues on community channels even if wire tone improved.
Broadcast-adjacent transcripts—talk radio, studio shows—often mediate between the two. Including them as a third axis stops desks from treating social-versus-news as a false binary. Illustrative charts in this article show the shape of divergence; they are not production sentiment KPIs.
Calibration without vanity metrics
Channel calibrations are living configuration: exemplars, sport-specific slang, and known irony patterns. When a campaign type is new—say a federation sustainability programme—desks seed exemplars before trusting divergence flags. Eagle Eye improves with those exemplars; Sports PR still reads samples when flags fire.
Never with invented accuracy percentages. Divergence flags and driver summaries are the product surface; blended vanity scores are not. Koba stays a narrator of disagreement, not a manufacturer of blended confidence theatre.
Takeaways
Cross-channel sentiment is a reconciliation problem. Dantani surfaces disagreement so sports PR campaigns adjust messaging with eyes open—never with invented accuracy percentages.
