Media Sentiment Score

Goal of the analysis:

The Media Sentiment Score measures the tone of earned media coverage about the company, brands, executives, and issues across print, online, and broadcast outlets. For Corporate Communications leaders, it serves as an early warning system for reputational risk, a gauge of message resonance, and a tangible KPI to inform executive decision-making, investor communications, and crisis management. Executives use this analysis to understand whether the narrative is moving in the desired direction, how their reputation compares to peers, and where to focus outreach, content, or corrective actions.

Data required:

• Media coverage and metadata:
  • Article/post details: headline, body text, URL, publication date/time, outlet, journalist/anchor.
  • Outlet attributes: type (online, print, broadcast, trade), geography, language, topic vertical, political bias (if available).
  • Reach/prominence: unique monthly visitors (UMV), circulation, audience rating points, placement (front page, homepage), headline vs. body mention.
  • Content type flags: news, feature, op-ed, review, press release, syndicated/wire.
• Sentiment and classification data:
  • Sentiment labels per mention (positive/neutral/negative) with model confidence score.
  • Manual coding samples for quality control and model calibration.
  • Language detection and translation for non-English content (if included).
• Entity and topic tagging:
  • Company, brand, product, executive names (and common variations).
  • Competitor entities and category keywords for share-of-voice comparisons.
  • Issue taxonomy (e.g., safety, DEI, sustainability, pricing, layoffs, privacy).
  • Message pillars/strategic themes for resonance analysis.
• Historical and benchmark data:
  • At least 12–24 months of historical coverage for baseline and seasonality.
  • Competitor coverage during the same periods.
  • Corporate events calendar (earnings, launches, incidents) for event alignment.
• Quality controls and deduplication:
  • Duplicate and syndication detection (wire stories, copy-paste reprints).
  • Bot/scraper filtering and outlet quality lists.

Detailed step-by-step instruction on how to conduct the analysis:

1. Define scope and taxonomy. Agree on entities (company, brands, executives), markets, languages, outlet types, and the issue/message taxonomy. Document inclusions/exclusions (e.g., exclude company press releases from earned sentiment).
2. Set up data collection. Use media monitoring platforms (e.g., Cision, Meltwater, Talkwalker, Brandwatch, LexisNexis, GDELT) to build Boolean queries. Validate precision/recall with test pulls and refine to reduce false positives (e.g., disambiguate brand names that are common words).
3. Ingest and clean data. Standardize fields, remove duplicates and syndicated duplicates (use normalized URLs, title+timestamp hashing), tag content types, and flag paywalled sources. Exclude low-quality outlets if policy dictates.
4. Apply sentiment scoring. Use platform NLP or a custom model to assign sentiment (positive/neutral/negative) and confidence. Calibrate with manual review: pull a stratified random sample across outlets/languages, compute accuracy, and refine rules (e.g., handle negations, sarcasm, domain-specific terms). Aim for ≥80–85% accuracy on positive vs. negative classification.
5. Compute core metrics.
   • Net Sentiment = (Positive Mentions – Negative Mentions) / Total Mentions. Range: –1 to +1 (multiply by 100 for percentage).
   • Reach-Weighted Net Sentiment = (Σ sentiment_score_i × reach_i) / Σ reach_i.
   • Prominence-Weighted Sentiment: apply weights for headline/front-page mentions vs. body mentions.
   • Positive/Negative Share of Voice (SoV): company’s positive/negative mentions as a share of category mentions.
   • Volume and velocity: total mentions per period and week-over-week change.
6. Segment the results. Break down sentiment and SoV by:
   • Outlet type (tier-1 national, trade, regional, broadcast) and geography.
   • Topics/issues and message pillars.
   • Product lines and named executives/spokespeople.
   • Journalists/outlets (top 20 by reach) to identify influencers and detractors.
7. Time-series and event analysis. Chart weekly/monthly Net Sentiment and Reach-Weighted Net Sentiment. Annotate key events (earnings, product launches, incidents) and detect deviations using control limits (e.g., 2–3 standard deviations).
8. Competitor and category benchmarking. Repeat the same process for peers. Compare Net Sentiment, Positive SoV, and volume; build matrices (e.g., SoV vs. Net Sentiment) to position the company relative to peers.
9. Extract drivers and narratives. For negative spikes or persistent gaps, analyze top contributing headlines, quotes, and topics. Use keyword co-occurrence and n-grams to identify drivers; capture exemplar headlines for executive context.
10. Synthesize insights and actions. Convert findings into implications: which messages land, which outlets to prioritize, which issues require mitigation, and what next actions to take. Agree on cadence (weekly pulse; monthly deep dive; quarterly board view).

Format of the output of analysis:

• Executive dashboard with KPIs: Net Sentiment, Reach-Weighted Net Sentiment, Positive/Negative SoV, volume, and velocity.
• Time-series line charts showing sentiment trends with event annotations.
• Stacked bar charts of positive/neutral/negative by outlet type, region, topic, and product.
• SoV vs. Net Sentiment quadrant comparing company and key competitors.
• Heatmaps of sentiment by topic and market; top 20 outlets/journalists impact table.
• Headline excerpts for context on major positive/negative drivers.
• Methodology appendix (definitions, weighting, deduplication, accuracy).

How to interpret results:

• Higher Net Sentiment indicates favorable tone; sustained levels above +20% suggest a positive narrative taking hold. Reach-weighted results are more indicative of real-world impact than unweighted.
• Low or negative Net Sentiment signals reputational risk. If negative coverage is concentrated in high-reach outlets or broadcast, prioritize rapid response and corrective engagement.
• High neutral share often reflects factual or financial reporting; if desired, sharpen the company’s point of view with data, quotes, and third-party validators.
• Positive SoV matters: strong sentiment with minimal share implies limited narrative impact; conversely, large share with negative tone is high risk.
• Compare across segments: negative sentiment clustered in a region or topic indicates targeted issues (e.g., regulatory, product quality). Outlet-level differences can surface relationship or understanding gaps.
• Trends over time: improving sentiment post-initiative indicates message resonance; volatility suggests issue management gaps or inconsistent messaging.
• Always read alongside volume: small volumes can produce noisy scores; apply minimum-volume thresholds before drawing conclusions.

Steps a company can take to improve on this measure:

• Narrative and content: Clarify a compelling corporate story with proof points; publish data-rich reports (e.g., sustainability, innovation); package assets (visuals, FAQs) that journalists can easily cite.
• Media relations and targeting: Build relationships with tier-1 and key trade journalists; offer embargoed briefings and exclusives; correct inaccuracies quickly with documented facts; pitch tailored angles per outlet.
• Issues and crisis readiness: Maintain playbooks, holding statements, and approval SLAs; establish social listening and media alerts to trigger rapid responses; rehearse scenarios and designate spokespeople.
• Spokesperson capability: Provide media training, message discipline, and bridging techniques; equip executives with talking points and data; align internal communications to avoid mixed messages.
• Digital and SEO amplification: Optimize newsroom pages (schema markup, fast load), publish quotable summaries, and distribute via owned channels; coordinate with paid to amplify high-value positive coverage.
• Data, systems, and modeling: Refine queries to reduce false positives; build a domain lexicon (e.g., “disruption” positive in tech, negative in operations); expand language coverage; periodically re-label samples to improve model accuracy.
• Governance and cadence: Weekly sentiment pulse with clear thresholds for escalation; monthly reviews with comms, legal, and business leaders; tie incentives to Positive SoV on priority topics.
• Scenario guidance: If Net Sentiment is high but reach is low, prioritize amplification and tier-1 pitching. If reach-weighted sentiment is negative while unweighted is positive, address a few high-reach outlets. If neutral share is high, craft stronger viewpoints and third-party endorsements.

Benchmark comparisons:

General benchmarks:

• Net Sentiment (unweighted): many companies range from -10% to +30% in steady-state coverage.
• High-performing reputations sustain +30% to +60% Net Sentiment outside of crises.
• Reach-weighted Net Sentiment typically trails unweighted by 5–15 points due to concentration in large, more neutral outlets.
• Positive Share of Voice: 25–50% is common; leaders achieve 50%+ on priority topics.
• During crises, Net Sentiment can drop below -50%; recovery to pre-crisis baseline within 4–8 weeks is a sign of effective response.

Segment- or industry-specific benchmarks:

• Regulated sectors (energy, pharma, utilities) often see lower steady-state sentiment (0% to +20%); consumer tech and luxury can sustain higher (+20% to +50%).
• Financial services and telecom are typically mid-range (+10% to +30%) with episodic volatility tied to rates, outages, or service issues.
• Broadcast coverage tends more neutral; trade media skews more positive if product news flows are strong.

If robust external benchmarks are unavailable, construct internal comparators: track quarterly rolling averages, compare business units or markets, and set “top quartile” internal performance targets. Use a consistent competitor set and alert thresholds (e.g., 3-sigma week-over-week negativity spike) to standardize interpretation and action.