GPT performs best with structure and specificity. The more detailed your prompt, the stronger and more accurate the result.
Why Use Custom Summaries
- Tailored output: Generate summaries that focus on what matters to you—reputational risk, sentiment, emerging themes, or actor analysis.
- Executive-ready reporting: Align tone and insights with stakeholder needs, from frontline analysts to the C-suite.
- Repeatable workflows: Reuse consistent, structured prompts for daily, weekly, or ad hoc reporting.
- Flexible and fast: Custom summaries allow you to adjust analysis across regions, timeframes, and focus areas without changing data ingestion workflows.
Example Use Cases:
Executive Briefings
Objective: Provide strategic overviews of public discourse for decision-makers.Prompt Style: High-level, narrative-driven, focused on reputational risk and external perception.
Audience: CEO, communications teams, external affairs leads.
Common Elements:
- Volume and tone of conversation
- Key narratives with implications
- Summary section with recommended actions or attention areas
Information Security Monitoring
Objective: Track risks such as phishing attempts, data leaks, hacktivism, or bot amplification.Prompt Style: Technical, structured, and aligned with threat intelligence reporting standards.
Audience: Security analysts, risk teams, incident response units.
Common Elements:
- Mentions of attack indicators (e.g., specific exploits or vulnerabilities)
- Bot or coordinated campaign detection
- High-reach or anomalous accounts
Geopolitical and Government Analysis
Objective: Summarize political or social discourse in target regions or during elections/conflicts.Prompt Style: Analytical, multi-lingual, aligned with diplomatic or policy perspectives.
Audience: Policy teams, defense contractors, foreign affairs staff.
Common Elements:
- Influential narratives by region
- Notable actors (state-affiliated media, journalists, politicians)
- Language-specific sentiment and coverage differences
Brand or Product Monitoring
Objective: Identify market feedback, brand sentiment, and unexpected narratives.Prompt Style: Customer-centric, concise, focused on product mentions, influencers, and virality.
Audience: Marketing, PR, product teams.
Common Elements:
- Spikes in conversation around product names or campaigns
- Key mentions from high-reach individuals or publications
- Sentiment shifts and thematic comparisons over time
Example Custom Summaries
| Summary Name | Description (Output Summary) | Actual Prompt Example |
|---|---|---|
| Discussion Themes | Groups conversations into themes. Each theme includes a favorability profile and examples. | Identify and describe the main discussion themes surrounding Ultra-Processed Foods. Group similar posts into themes and, for each theme, provide a summary, note the favorability distribution, and include representative examples. |
| Favorability Discourse | Compares social media, news, and blogs. Notes that social media sentiment is balanced, news coverage skews positive, and blogs are mostly neutral. | Compare discourse about Ultra-Processed Foods across social media, news, and blogs. For each channel, summarize the dominant narratives and sentiment, and explain how conversations differ across these platforms. |
| Favorability – Unfavorable Analysis | Summarizes negative narratives, including criticisms or reputational risks. | Identify and summarize narratives expressing unfavorable sentiment about Ultra-Processed Foods. Explain key criticisms (such as health risks or boycotts) and discuss their impact on its reputation. |
| Executive Brief | Summarizes total mentions, sentiment distribution, and key narratives. Offers strategies to manage perception. | Prepare a concise executive briefing on McDonald’s, including total mentions, sentiment distribution, key narratives, and recommended actions to manage public perception. |
| Bot Analysis | Reports on bot activity in discussions and identifies key narratives promoted by bots. | Assess bot involvement in social-media discourse about boycotts of McDonald’s. Identify bot-generated posts, classify them by likelihood, and summarize the key narratives bots are pushing (e.g., boycott comparisons, geopolitical context, corporate criticism). |
| Competitor Report | Generates a structured digest of mentions, trends, and predictive insights about competitors. | You are an insights analyst generating digest reports for McDonald’s. Generate a structured report summarizing mentions and insights from this workspace, focusing on McDonald’s competitors. Report Structure 1. Executive Summary – Provide a narrative overview of activity within the selected period. Highlight mention volume and notable spikes/clusters across social, blogs, and news. Summarize overall visibility of each competitor, noting which company drove the most attention and why. 2. Key Mentions & Clusters – Identify and summarize high-impact mentions (e.g., major news articles, Reddit threads, viral posts). Group recurring themes into clusters such as clinical trial results, partnerships/M&A, regulatory updates, market analysis, and publications. 3. Predictive & Speculative Insights – Surface forward-looking signals such as upcoming data releases, actions, safety events, policy/payer trends, and key dates. 4. Investment & Market Relevance – Translate findings into implications for McDonald’s positioning. Highlight risks, opportunities, and market signals. Use concise, scannable formatting and label predictive content clearly. |
Getting Started
To create a custom prompt:- Navigate to your PeakMetrics dashboard
- Select the AI Summary module
- Input your structured prompt
- Review and iterate based on output
- Contact the Insights Team for guidance or improvements
Collaboration with the Insights Team
We know prompt design can feel overwhelming at first. That’s why our Insights Team partners directly with you to:- Workshop and write prompt language tailored to your reporting style
- Set up prompt-based workflows that align with your internal cadence
- Review and refine outputs to ensure accuracy and impact
- Share best practices from other successful prompt implementations
Best Practices for Prompt Design
- Be specific: Define exactly what kind of mentions or analysis you want. Vague prompts lead to vague outputs.
- Structure the output: Use numbered or labeled sections to guide GPT’s organization.
- Use role-based framing: Start with “You are a [type of analyst]” to influence tone and depth.
- Reference filters: Remind the AI to stay within your applied filters and date range.
- Avoid generalizations: Instruct the AI not to speculate or introduce information outside the data.
For assistance, reach out to your PeakMetrics Customer Success contact or email support@peakmetrics.com.