How Ovation turns public audience signals into a decision brief

1. Scope and window

Each Narrative Report uses a rolling seven-day window and the plan’s published source caps. Closely timed reports can contain overlapping comments and are not statistically independent samples. Focus Scans inspect one eligible public post or video owned by the saved creator identity.

Ovation retains derived findings, aggregate coverage, limitations, and public supporting-post links. Raw comments are processed in bounded worker memory and are not saved in Ovation report history.

2. Supported sources

• Instagram and TikTok: server-side public-profile collection for the saved creator.
• YouTube: public channel and video data through a server-only YouTube Data API key. The user supplies a public channel URL, handle, or channel ID.
• X: revocable read-only authorization for the creator’s public posts and replies.

Public YouTube comments are excluded unless Swaynexa has recorded the required provider approval and enabled the server-side gate. Saving a public channel does not grant access to a Google account.

3. Narrative Score v2

The 0–100 Narrative Score is a deterministic composition of four model-estimated drivers:

• Audience affinity, 35%: favorable versus unfavorable audience response.
• Message clarity, 25%: whether repeated misunderstandings are present.
• Constructive momentum, 20%: the balance of useful positive demand and negative friction.
• Risk pressure, 20%: repeated risks weighted by severity, prevalence, and change; lower is better.

The score is round(0.35 × affinity + 0.25 × clarity + 0.20 × momentum + 0.20 × (100 − risk pressure)). It is a directional model estimate, not a poll, census, probability, or guaranteed forecast. Confidence and source coverage are reported separately.

4. Comparison and meaningful change

Ovation shows movement only when the prior report uses the same score method and source coverage remains comparable. Movement is withheld when provider mix, sample volume, or collection completeness changes materially.

A meaningful-change alert requires at least one documented threshold: an absolute score move of eight points, a qualifying new or rising risk, a theme moving at least two prevalence bands, or a theme becoming dominant. The first report is a baseline and cannot trigger a change alert.

5. Action Log and report comparison

The creator can save a recommendation and mark it planned, used, partly used, or skipped. Notes and state-specific timestamps help preserve the creator’s own record of what happened.

Later reports may describe change only when methodology and source coverage are comparable. Those report-to-report comparisons are directional context and do not establish that a recorded action caused the later result.

6. Limits

• Available public content can be incomplete, delayed, removed, or unavailable.
• Language, sarcasm, slang, spam, and coordinated behavior can reduce classification quality.
• Estimated prevalence labels are ordinal, not exact audience percentages.
• Evidence links are representative support, not a complete archive.
• Users should review evidence and limitations before acting on a recommendation.