Skip to main content
Sentiment Analysis uses AI to examine threads across your application, detect conversation topics, classify each mention as positive, neutral, or negative, and produce summaries with actionable recommendations — all without manual tagging or surveys.

​
Features

Sentiment Scoring

Overall sentiment score (−100 to +100) with period-over-period comparison to track community health trends

Topic Detection

AI automatically identifies and tracks conversation topics across all analyzed threads

AI Summaries

Per-topic summaries of negative, neutral, and positive mentions with actionable recommendations

Thread-Level Analysis

Browse individual analyzed threads with detected topics and full conversation context

Trend Charts

Sentiment over time and mentions over time visualizations with percentage and absolute views

Settings

Define the custom topics the AI tracks, and re-run the current configuration over past threads with retroactive tagging

​
Getting Started

Need Sentiment Analysis enabled? Email [email protected] with your network ID.

​
Credit & Analysis Coverage

At the top of the page a credit banner shows how many threads have been analyzed this calendar month, out of your monthly credit allocation.
  • Coverage percentage — A circular gauge (e.g., 95%) indicates how much of your monthly allocation has been consumed.
  • Thread count — For example, 9520 of 10000 threads analyzed.
  • Manage credit — Click to view or adjust your credit allocation.
The credit banner always reflects the current calendar month and resets at the start of each month. It is not affected by the date range or filters you select on the page — changing the date picker will not change these numbers.
If the number of threads created or updated in your application exceeds your credit, not all threads will be analyzed by AI.

​
Filtering & Date Range

​
Date Range

Use the date picker in the top-right corner to select the analysis period. Comparison metrics (e.g., +2 pts, +80%) automatically reference the equivalent prior period.

​
Filters

Click the Filter button to narrow analysis scope:
1

Open Filter Panel

Click the Filter button next to the date picker.
2

Select Filters

Choose communities and/or user tags from the dropdowns.
3

Apply

Click Apply to refresh all tabs with the filtered data. Use Clear all to reset.

​
Overview Tab

The Overview tab provides a high-level snapshot of sentiment across your entire application (or filtered scope).

​
Key Metrics

Sentiment Score

Net sentiment score from −100 to +100, where 0 means positive and negative mentions are balanced. Labeled Positive (above 0), Neutral (exactly 0), or Negative (below 0). Shows point change vs. the prior comparison period.

Total Topics Detected

Number of distinct topics the AI identified in the selected period. Shows percentage change vs. prior period.

Total Mentions

Aggregate count of all topic mentions across analyzed threads. Shows percentage change vs. prior period.

​
How the Sentiment Score Is Calculated

The score is the net balance between positive and negative mentions, expressed on a scale from −100 to +100:
Worked example — 12 positive, 15 negative, and 63 neutral mentions (90 total):
Neutral mentions are included in the total (the denominator), but never in the positive or negative counts.This means neutral mentions can never change whether a score is positive or negative — they only pull it closer to zero. A period that is overwhelmingly neutral will score near 0 even when the small non-neutral remainder leans one way:A small negative score such as −3 therefore means “essentially balanced, leaning very slightly negative” — not “3 out of 100”.
A mention is one topic classification on one thread — not one thread, and not one post.If a single thread is classified under three topics, it contributes three mentions, each with its own sentiment. A thread that is negative about several topics therefore carries more weight in the score than a single-topic thread.Because of this, the Total Mentions figure is normally higher than the number of threads analyzed, and will not match the thread count shown in the credit banner.
Mentions are grouped by day across the selected range, then summed. The score is calculated once on the period totals — it is not an average of the daily scores, so higher-volume days carry proportionally more weight.The score is rounded to the nearest whole number for display only; the underlying value keeps full precision.Any manual sentiment overrides your team has applied take priority over the AI classification and are reflected in all charts and totals.
The comparison period is the window of equal length immediately preceding your selection. Selecting a 14-day range compares it against the 14 days directly before it.The change is shown in points on the −100 to +100 scale, not as a percentage — if the current period scores −3 and the change reads −25 pts, the earlier period scored +22.
The score is highly sensitive when mention volume is low. A period with just 3 positive and 2 neutral mentions scores +60. Large point swings between periods usually reflect a small sample rather than a real shift in community sentiment — check the mention counts for both periods before acting on a big change.

​
Top Negative & Positive Topics

Two side-by-side panels highlight the topics with the highest concentration of negative and positive sentiment:
  • Top topic callout — Shows the topic name, total mention count, and what percentage was negative or positive.
  • Horizontal bar chart — Ranks the top 5 topics by total mentions. Negative chart bars are shown in red/orange; positive chart bars are shown in blue.
Hover over any bar in the chart to see a breakdown of negative, neutral, and positive mention counts for that topic.

​
Sentiment Breakdown

A donut chart showing the distribution of all mentions across three categories:
  • Negative mentions (red) — count and percentage
  • Neutral mentions (orange/yellow) — count and percentage
  • Positive mentions (blue) — count and percentage

​
Sentiment Over Time

A line chart tracking how sentiment proportions change over the selected date range. Toggle between:
  • Percentage — Shows each sentiment category as a percentage of daily total
  • Number — Shows absolute mention counts per day

​
Topic Analysis Tab

The Topic Analysis tab provides a sortable table of all detected topics with detailed metrics.

​
Topic Detail Page

Clicking View insights on any topic opens a dedicated page with deep analysis for that specific topic.
  • Sentiment score — Topic-specific score (−100 to +100) with comparison to prior period
  • Total mentions — Mention count for this topic with percentage change
The right-side panel provides AI-generated insights organized by sentiment:
  • Negative mentions — Summary of key negative themes with a View threads button to see source posts
  • Neutral mentions — Summary of neutral/mixed feedback
  • Positive mentions — Summary of positive themes and praise
  • Recommendation — Actionable suggestions derived from the analysis (e.g., “Address concerns urgently. Consider user feedback surveys.”)
Line chart showing how mention volume for this topic changed over the selected date range.
Donut chart showing negative, neutral, and positive distribution for this specific topic.
Line chart with Percentage and Number toggle showing sentiment trend for this topic.

​
Analyzed Threads Tab

The Analyzed Threads tab lets you browse the actual posts and comments that were analyzed.

​
Thread List (Left Panel)

A scrollable list of analyzed threads showing:
  • Author name and avatar
  • Timestamp (e.g., 2 Apr 2026, 17:35)
  • Community name where the thread was posted
  • Content preview — Truncated text of the original post
  • Engagement — Reaction count and comment count

​
Thread Detail (Right Panel)

Selecting a thread displays:
  • Topic detected — Tags showing which topics the AI identified in this thread (editable via tag icons)
  • Full post content — Complete text including hashtags, mentions, and links
  • Engagement metrics — Reactions and comments count
  • Comment thread — All comments and replies with author, timestamp, and nested replies
Use the Analyzed Threads tab to validate AI-generated summaries by reviewing the actual source posts that informed them.

​
Settings

Access Settings via the settings gear icon (⚙) in the top-right corner of the Sentiment Analysis page. Settings has two tabs:

​
Topic Configuration

This configuration applies to new analyses only. Changes take effect from the date they are saved, and previously completed analyses are not affected. To update past threads, use Retroactive Tagging.

​
Managing Topics

1

Open Configuration

Click the settings gear icon on the Sentiment Analysis page to navigate to Topic configuration.
2

Review Existing Topics

View the table of configured topics with their names and date added. The table is sortable by Date added.
3

Add a Topic

Click + Add topic to define a new topic for the AI to track in future analyses.
4

Remove a Topic

Click the delete icon (🗑) in the Action column to remove a topic from future analyses.
Removing a topic stops future analysis for that topic but does not delete historical data. Historical results remain available for the date range when the topic was active.

​
Retroactive Tagging

Topic configuration changes only apply going forward, so threads analyzed before a topic existed carry no mentions of it. Retroactive tagging closes that gap: it reprocesses historical threads in a date range you choose using the current topic configuration, then updates those threads with the new results. Open it via settings gear icon (⚙) → Retroactive tagging.
Typical use: you add a topic today and want last month’s conversations counted against it, or the AI missed an emerging theme you have since configured.

​
Job History

The tab lists every retroactive job for the network, newest first.

​
Job Details

Click View details on any row to open a side panel containing:
  • Date range as the panel title, with a note that historical threads in that range were reprocessed using the topic snapshot
  • Threads found — the thread count for that job
  • Topic in this snapshot — the exact topic list the run used, with each topic’s Date added, sortable
Because the panel stores the snapshot, you can tell which topics existed at the time of a past run — useful when two jobs over the same range return different results.

​
Running a Retroactive Job

1

Confirm Your Topics

On the Topic configuration tab, verify the topic list is what you want applied. The job uses the configuration as it exists at submission time.
2

Start the Flow

On the Retroactive tagging tab, click + Apply retroactively.
3

Pick a Date Range

Choose the window to reprocess. Presets cover Last 7 days, Last 14 days, and Last 30 days; a two-month calendar handles custom ranges. Future dates are disabled, and preset ranges end on the previous day.
4

Review the Estimate

Click Estimate. The review step shows the data range with its day count, the number of topics being applied, the approximate Threads found, the Credit required for this job, and the Credit remaining after it runs.
5

Start the Job

Click Start retroactive. The job is queued and begins processing when capacity is available; affected threads are updated once it completes.
Retroactive jobs cannot be undone. A running job can’t be canceled, and a completed job can’t be changed or deleted. Confirm the date range and credit cost on the review step before starting.

​
Credit Cost

Retroactive jobs draw on the same thread credit pool as ongoing analysis — one credit per thread analyzed. Threads found on the review step is an estimate, and the review step is the only place the cost appears before the job starts.

​
Daily Operational Workflow

1

Check Overview

Open Sentiment Analysis and review the three KPI cards for any significant changes in sentiment score, topic count, or mention volume.
2

Identify Problem Areas

Check the Top negative topics panel. If any topic shows a high negative percentage, click through to investigate.
3

Review AI Summaries

On the topic detail page, read the AI Summary panel’s negative mentions and recommendations.
4

Validate with Source Data

Click View threads to review actual posts behind the AI summaries. Confirm findings before taking action.
5

Monitor Trends

Use the Sentiment over time chart to check whether negative sentiment is a spike or sustained trend.
6

Take Action

Convert high-confidence recommendations into backlog tickets, content changes, or moderation actions.

​
Best Practices

  • Monitor the credit banner regularly — if coverage drops below 80%, consider applying filters to prioritize high-value communities.
  • Use community and user tag filters to focus credits on segments that matter most.
  • Contact support to increase credit limits if your application consistently exceeds allocation.
  • Start with 5–10 broad topics and refine as patterns emerge from the AI analysis.
  • Review the Topic Analysis tab monthly to identify if any configured topics have zero mentions — consider replacing them.
  • Add new topics when you notice emerging themes in the Analyzed Threads tab.
  • Remember that topic changes only affect future analyses — use retroactive tagging to apply them to past threads.
  • Finish adding topics before submitting a job — each run snapshots the configuration, so a forgotten topic means paying for the same range twice.
  • Start with a short range to measure credit cost per day, then widen it.
  • Check Credit remaining after on the review step; a wide range over a busy period can consume a large share of the pool.
  • Use each job’s detail panel to compare topic snapshots when two runs over the same range disagree.
  • A sentiment score of 0 indicates perfectly balanced sentiment — not necessarily good or bad. Scores near 0 (such as −4 or +3) mean the community is broadly balanced, not that sentiment is poor.
  • Focus on change over time rather than absolute scores — a 5+ point drop warrants investigation.
  • Cross-reference sentiment shifts with product releases, incidents, or campaigns for causal analysis.
  • Low mention volumes can cause score volatility — set a minimum threshold before acting on changes.
  • Always validate AI summaries by reviewing source threads before escalating issues.
  • Use the Recommendation section as a starting point, not a final directive — apply business context.
  • Compare negative topics against positive topics to understand relative priority.
  • Track whether actions taken improve sentiment in subsequent analysis periods.

AI Research

Ad-hoc natural language analysis across community conversations

Social Analytics

Quantitative engagement and usage metrics for social features

Raw Data Export

Export raw data for deeper custom analysis and reporting