> ## Documentation Index
> Fetch the complete documentation index at: https://learn.social.plus/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Social Insights

> Qualitative intelligence for topics, sentiment, search intent, and AI research—convert community conversations into actionable product, support, and content decisions.

<Info>
  This dashboard augments quantitative usage analytics with qualitative insight—surfacing what people talk about, how they feel, what they search (but cannot find), and why churn or satisfaction shifts are emerging.
</Info>

<Note>
  **Sentiment Analysis has moved.** The sentiment analysis capabilities described here have been expanded into a dedicated feature with AI-powered topic detection, per-topic summaries, and thread-level analysis. See [Sentiment Analysis](./sentiment-analysis) for the new experience.
</Note>

<CardGroup cols={3}>
  <Card title="Topic Mentions" icon="at">Volume & share of conversation</Card>
  <Card title="Sentiment" icon="face-smile">Emotional direction & skew</Card>
  <Card title="AI Summaries" icon="file">Condensed thematic reasoning</Card>
  <Card title="Recommendations" icon="lightbulb">Actionable remediation & leverage</Card>
  <Card title="AI Research" icon="magnifying-glass">Ad‑hoc natural language analysis</Card>
  <Card title="Search Intent & Gaps" icon="arrows-left-right-to-line">Unmet demand & content prioritization</Card>
</CardGroup>

## What Can You Answer in Minutes?

<Tabs>
  <Tab title="Product Feedback">Which features generate praise vs friction and their emerging pain themes.</Tab>
  <Tab title="Release Impact">Did the latest update shift topic mix or sentiment trajectory?</Tab>
  <Tab title="Support Deflection">Which high-frequency searches fail → candidate help docs / in‑app tips.</Tab>
  <Tab title="Content Strategy">Which gap topics warrant articles, videos, or onboarding flows.</Tab>
  <Tab title="Risk Monitoring">Identify negative clusters early (bugs, policy issues, safety concerns).</Tab>
</Tabs>

## Daily Operational Workflow

<Steps>
  <Step title="Open & Filter">Select date range, locale, platform/app segment, and (optionally) user cohort (new vs returning) to focus analysis.</Step>
  <Step title="Scan Topic Shifts">Compare Top Mentions list vs prior period % change to flag emerging themes.</Step>
  <Step title="Assess Sentiment">Check Overall Sentiment gauge; drill into topics with highest negative skew.</Step>
  <Step title="Read AI Summaries">Skim negative → neutral → positive summaries to capture root causes & delights.</Step>
  <Step title="Review Recommendations">Validate AI suggestions; convert high-confidence items into backlog tickets (label with source tag).</Step>
  <Step title="Run Research Query">Pose a targeted natural language question for deeper synthesis (e.g., "Main complaints about video uploads this week").</Step>
  <Step title="Evaluate Search Gaps">Open Search Insights tab; prioritize new content for highest Gap Index terms.</Step>
  <Step title="Log Actions">Record chosen remediations with owner & ETA; schedule follow-up check next cycle.</Step>
</Steps>

## Community & Topic Filtering

<Info>
  Refine any insight to the communities and topics that matter—filtering removes noise so trend, sentiment, and gap signals stay actionable.
</Info>

<CardGroup cols={3}>
  <Card title="Topic Filter" icon="hashtag">Focus on specific themes</Card>
  <Card title="Community Filter" icon="users">Isolate audience segments</Card>
  <Card title="Combined Scope" icon="bullseye-arrow">Intersect topic + community for precision</Card>
</CardGroup>

### Why Use Community-Level Filters?

* Targeted insights: Each community exhibits distinct engagement & sentiment patterns.
* Faster diagnosis: Narrow scope to confirm whether an issue is broad or localized.
* Segment strategies: Tailor interventions (content, moderation, onboarding) per cohort.
* Reduce noise: Avoid dilution from mega-communities overshadowing smaller niches.

### Getting Started

| Requirement        | Details                                               |
| ------------------ | ----------------------------------------------------- |
| Access             | Portal login with Social Insights permission          |
| Feature Enablement | Social Insights enabled (contact support if disabled) |
| Data Window        | Default range (e.g., last 30 days)—adjust as needed   |

<Info>Need Social Insights enabled? Email <a href="mailto:support@social.plus">[support@social.plus](mailto:support@social.plus)</a> with your tenant ID.</Info>

### Interface Tour

| Element            | Location             | Purpose                                         |
| ------------------ | -------------------- | ----------------------------------------------- |
| Filter Bar         | Top of insights view | Houses Topic & Community selectors + date range |
| Topic Dropdown     | Left selector        | Choose one or multiple topics to include        |
| Community Dropdown | Right selector       | Search & pick one or more communities           |
| Date Range         | Right side           | Adjust temporal scope for comparison            |

<AccordionGroup>
  <Accordion title="Default View">All communities & topics aggregated (e.g., total mentions count across full corpus) until any filter is applied.</Accordion>
  <Accordion title="Progressive Narrowing">Apply topic first to isolate theme, then add community filter to test audience-specific variance.</Accordion>
  <Accordion title="Reset Strategy">If results feel sparse, clear the most recently added filter first to widen context.</Accordion>
</AccordionGroup>

### Applying Filters (Step-by-Step)

<Steps>
  <Step title="Open Topic Menu">Click Topic: Select topic; search or scroll the list.</Step>
  <Step title="Select Topic(s)">Choose one (start simple) or multi-select to form a thematic cluster.</Step>
  <Step title="Observe Metrics">Note changes in Total Mentions, sentiment bars, emerging topics list.</Step>
  <Step title="Add Community Filter">Open Community: Select community; search e.g. "Fashion & Style"; select.</Step>
  <Step title="Compare Before/After">Is sentiment skew or gap profile materially different? Capture delta.</Step>
  <Step title="Iterate">Test additional communities to spot outliers; avoid over-filtering below statistically meaningful volume.</Step>
</Steps>

### Interpreting Filtered Results

<Tabs>
  <Tab title="Low Volume Caution">If filtered mentions fall below reliability threshold (internal baseline), treat insights as directional only.</Tab>
  <Tab title="Divergent Sentiment">A community showing negative skew vs global baseline indicates localized friction—investigate context posts.</Tab>
  <Tab title="Gap Localization">High Gap Index within one community reveals niche documentation/content deficit.</Tab>
  <Tab title="Adoption Variance">Topic share % shifts highlight uneven feature or content adoption patterns.</Tab>
  <Tab title="Escalation Routing">Community-specific negative themes can route to specialized support or product squads.</Tab>
</Tabs>

### Best Practice Patterns

<AccordionGroup>
  <Accordion title="Sequential Filtering">Apply topic FIRST to anchor analysis, then layer community to avoid premature fragmentation.</Accordion>
  <Accordion title="Benchmark Preservation">Keep an unfiltered baseline tab open for quick relative comparisons.</Accordion>
  <Accordion title="Volume Guardrail">Set a minimum mentions threshold (e.g., 300) for acting on sentiment shifts.</Accordion>
  <Accordion title="Change Logging">Record filter combinations used when generating recommendations to ensure reproducibility.</Accordion>
  <Accordion title="Periodic Reset">Once per week, review global view to avoid tunnel vision from niche segments.</Accordion>
</AccordionGroup>

### Example Workflow

1. Start unfiltered → identify emerging topic.
2. Apply topic filter → confirm growth & sentiment skew.
3. Add top 3 related communities → detect which exhibits worst negative skew.
4. Run AI Research prompt: "Root causes of negative sentiment about \<topic> in \<community>."
5. Validate references → create targeted remediation ticket.

<Info>Combining topic + community filters early in investigation accelerates root cause isolation but always validate sample size before action.</Info>

## Modules & Interpretation

### Mentions & Topic Breakdown

Highlights conversation concentration. A healthy distribution usually shows a balanced long-tail—extreme concentration may indicate a blocking issue OR a successful campaign.

### Trend Graph

Multi-line frequency lines reveal acceleration or decay. Sustained upward slope + positive sentiment → amplify; upward + negative → triage.

### Sentiment (Overall + Per Topic)

Overall gauge (0–100). Correlate dips with deployment timestamps or incident reports. Topic sentiment bars isolate outlier issues masked by aggregate positivity.

### AI Post Summaries

Condense thousands of posts into 5–10 key statements per sentiment polarity. Treat as directional—spot check references before broad decisions.

### AI Recommendations

Strategic remediation / opportunity list derived from the weighted intersection of volume, sentiment skew, and recency. Confidence score (if shown) reflects model certainty; low-confidence items may require manual verification.

### AI Research (Conversational Analyst)

Natural language Q\&A across indexed conversations. Returns a structured mini-report: Introduction, Methodology, Findings, References (canonical post links) to enable traceability.

### Search Insights & Content Gaps

Maps expressed intent (search queries) vs existing content coverage. Gap Index ranks unmet demand to drive documentation & self-service improvements.

## Metrics & Signals

| Metric                        | Definition                                                              | Why It Matters                                   | Action Trigger                                    |
| ----------------------------- | ----------------------------------------------------------------------- | ------------------------------------------------ | ------------------------------------------------- |
| Total Mentions                | Total topic references in filtered period                               | Measures conversation volume / engagement energy | Spike > X% without planned campaign               |
| Topic Share %                 | Mentions of topic / total mentions                                      | Identifies dominance or neglect                  | Single topic >40% share 2+ days                   |
| Emerging Topic                | Topic with high period-over-period growth and baseline low prior volume | Surfacing new opportunities or issues            | Growth rate > threshold (e.g., >150%)             |
| Overall Sentiment Score       | Normalized weighted sentiment across posts (0–100)                      | Health & morale indicator                        | Drop of >5 pts d/d or >8 pts w/w                  |
| Topic Sentiment Skew          | Deviation between topic sentiment and overall baseline                  | Prioritizes hotspots                             | Negative skew > (baseline -10)                    |
| Negative Summary Themes Count | Distinct negative themes extracted                                      | Complexity of issues / support load              | Rapid increase in count                           |
| Total Searches                | Volume of internal search queries                                       | Demand / intent scale                            | Sustained decline (possible UX findability issue) |
| Trending Searches Count       | # of search terms with rapid growth                                     | Emerging informational needs                     | Large increase → accelerate content creation      |
| Content Gap Count             | Unique queries with low/no results                                      | Self-service deficiency                          | > predetermined target                            |
| Gap Index                     | Weighted (query volume vs results available)                            | Prioritization of content production             | Top index unchanged across cycles (stale backlog) |
| Zero-Result Rate %            | (Queries with no result) / total searches                               | Support deflection efficiency                    | Rate > target (e.g., >8%)                         |

## Leveraging AI Research Effectively

<Tabs>
  <Tab title="Problem Investigation">"Why did sentiment drop after release 5.2?"</Tab>
  <Tab title="Feature Validation">"User perception of the new Stories feature this week."</Tab>
  <Tab title="Support Intelligence">"Top causes of failed media uploads last 7 days."</Tab>
  <Tab title="Adoption Barriers">"Reasons new users mention abandoning chat setup."</Tab>
  <Tab title="Opportunity Mining">"Unprompted requests related to live events."</Tab>
</Tabs>

## Prompt Patterns

<AccordionGroup>
  <Accordion title="Comparison">"Compare sentiment for video uploads vs live streaming this month."</Accordion>
  <Accordion title="Root Cause">"Summarize primary reasons behind negative sentiment about notifications."</Accordion>
  <Accordion title="Cohort Split">"Differences in topics between new and returning users last 14 days."</Accordion>
  <Accordion title="Temporal Shift">"What changed in top 5 topics compared to previous week?"</Accordion>
  <Accordion title="Opportunity">"Identify feature improvement opportunities mentioned positively but low volume."</Accordion>
</AccordionGroup>

## Related

<CardGroup cols={2}>
  <Card title="Sentiment Analysis" icon="face-smile" href="./sentiment-analysis">AI-powered sentiment scoring, topic detection, and thread-level analysis</Card>
  <Card title="Activity Analytics" icon="chart-line" href="./activity-analytics">Quantitative usage metrics</Card>
  <Card title="Raw Data Export" icon="download" href="./raw-data-export">Deeper custom analysis</Card>
  <Card title="User History" icon="user" href="../../console/user-and-content-management/user-social-history">Drill into individual behavior</Card>
</CardGroup>

<Info>Need advanced taxonomy tuning, extended retention windows, or custom sentiment domains? Contact support for enhanced AI configuration options.</Info>