Socialseal opportunity analysis

SocialSeal Opportunity Analysis

Overview

SocialSeal opportunity analysis answers: where should the brand create or improve content because people are searching and the current surfaced content leaves a gap?

This skill replaces a vague prompt like “analyze this SocialSeal data.” It tells the agent what to inspect, how to define an opportunity, what metrics are safe to use, and what a good deliverable looks like.

Inputs

Required:

• SocialSeal export files or SocialSeal MCP/CLI access
• workspace ID and group ID(s), or exported CSV/JSON files
• brand/entity definition and owned handles
• platform, market, language, keyword/topic scope, and date range

Good to have:

• competitor list
• current content pillars
• business priority or target audience
• previous-period export for comparison

Data Access

Use enriched ranked search rows for keyword-level analysis:

npx -y @socialseal/cli data export-search-results \
  --group-ids <group-id-1>,<group-id-2> \
  --workspace-id <workspace-id> \
  --out ./exports/search-results.csv \
  --timeout 120000

Use group evidence when you need a safer unified export across social and Google AI groups:

npx -y @socialseal/cli data export-group-evidence \
  --group-id <group-id> \
  --workspace-id <workspace-id> \
  --out ./exports/group-evidence.csv \
  --timeout 120000

MCP-first: there is no export-group-evidence/export-search-results MCP tool. Reach enriched ranked rows via socialseal_export_report (reportType: "search_results_enriched", payload: { "groupIds": [<group-id>] }) or socialseal_export_tracking_data. See references/mcp-and-cli-usage.md.

Attribution: cite surfaced results in human-readable terms, the "keyword" [market, platform], the video title or video_url, and @author_handle, with where it ranked. Keep video_uid/search_result_id only as an internal traceability note, and respect the timestamp rules (no platform-age claims when published_at is blank). See references/socialseal-data-contract.md.

Evidence tiers: discoverability, coverage, and share of voice are hard observations, not estimates; report them plainly. But every figure is computed over a biased sample (only high-ranking videos for the tracked queries), so it describes "what surfaces for these queries," not total market demand. See references/evidence-and-confidence.md.

Workflow

1. Inspect the export. Identify columns, grain, platform, market, date range, keyword field, URL/media ID, account/entity fields, metrics, and analysis fields.
2. Define the denominator. For each group, count qualified keywords/searches and qualified surfaced results. Exclude irrelevant rows before calculating.
3. Separate entities. Tag owned brand, competitors, creators/media/partners, and irrelevant/noise. If owned handles are unknown, ask or create a conservative placeholder column.
4. Calculate safe metrics. Use keyword coverage, discoverability by topic, surfaced attention/share where qualified metrics exist, and owned-vs-creator split. Do not lead with raw row counts.
5. Find gaps. A strong opportunity usually has search intent, relevant surfaced content, weak/absent owned presence, and a content job the brand can answer.
6. Classify content jobs. Use practical tags such as teach, walkthrough, compare, plan, reassure, show mood, show detail, or answer first-timer questions.
7. Prioritize. Rank by relevance, search intent strength, competitive/creator activity, expected business usefulness, and production feasibility.
8. Select evidence. For each recommendation, include metrics plus 1-3 examples that show what currently surfaces, cited by video title/URL and @handle. Label each recommendation's confidence basis (measured / scoped statistic / indicative pattern).

Output

Create an opportunity table with:

• priority
• platform / market / keyword or topic
• current brand presence
• who/what surfaces now
• content job
• why this matters
• recommended content direction
• evidence examples (video title/URL, @handle, "keyword" [market, platform])
• caveat/confidence (measured / scoped statistic / indicative pattern)

Also include a short action summary: what to brief, what to track, what to inspect further, and what not to pursue.

Do / Don't

Do:

• use percentages and denominators
• distinguish “no brand presence” from “no meaningful search demand”
• cite examples by video title/URL, @handle, and "keyword" [market, platform]
• keep recommendations shootable or operationally actionable

Don't:

• call every zero a whitespace opportunity
• recommend topics outside the brand’s plausible scope
• claim total market demand from sampled social-search results (selection bias)
• make the user read internal IDs or raw exports as the deliverable

Troubleshooting

• If exports lack video/source identifiers, run export-group-evidence or refresh the export before video-level analysis.
• If brand tagging is ambiguous, tag conservatively and add a review-needed column.
• If keyword language is mixed, split analysis by language/market before ranking opportunities.
• If the export is stale, refresh before making recommendations.

Verification Checklist

• Export grain and scope are stated.
• Denominators are visible.
• Owned, competitor, creator, and irrelevant rows are separated.
• Opportunities include evidence and a content job.
• Recommendations are actionable and within brand scope.