Performance Analyzer

Performance Analyzer

Analyze influencer campaign performance past surface metrics — score results vs target/benchmark, rank platforms/creators/content, read engagement quality and sentiment, attribute conversions, and write ranked learnings.

Cross-discipline (paid ads): this is also the cross-channel paid-ads scorecard/anomaly lens — account-wide metric rollups vs target/benchmark that feed ad-test-designer (what to test) and paid-measurement-loop (what to read back). Save paid runs under memory/ad/performance-analyzer/.

Quick Start

Analyze performance of [campaign name] influencer campaign

Compare creators within one campaign:

Compare performance of these influencers from [campaign]: @handle1, @handle2, @handle3

Skill Contract

• Reads: campaign name and date range; native platform analytics (reach, views, engagement); influencer-supplied reports or screenshots; website/GA traffic and conversion data; sales and promo-code redemption data; targets, benchmarks, and the preregistered decision rule/readback window if supplied; the optional lightweight campaign tracker and its evidence_refs; and any ROI/ROAS artifact already computed by roi-calculator. Reuse each explicit upstream opaque creator_ref or a verified creator-registry aggregate ID; a raw handle/name/URL/provider ID is transient lookup input only and never becomes a saved identity. Per-creator baselines come from memory/creators/<aggregate-id>.md only when an authorized artifact or verified registry link resolves that ref. Never derive the path from a raw locator.
• Writes: return the performance analysis inline by default. When a current non-forked tracker-state artifact proves measured or closed, include the compact Campaign Retro Card from step 8 bound to that campaign, creator, measurement contract, and decision rule. Save the analysis and card together to memory/influencer/performance-analyzer/YYYY-MM-DD-<campaign>.md only with exact WARM-save authorization; saved tables, headings, evidence, and handoffs use creator_ref plus opaque source refs, never raw handles, names, profile URLs, email addresses, or provider IDs.
• Promotes: only with separate exact authorization, promote durable evidence-backed campaign facts (verified metric results and descriptive format/platform associations) to memory/hot-cache.md; any ROI/ROAS value remains tied to its exact roi-calculator artifact. The Retro Card's qualitative renew | retest | retire | unknown decision, rationale, next hypothesis, and limitations remain WARM and are never promoted as registry truth. This skill makes no creator-registry proposal: after a creator row is closed, the existing boundary still permits only a separately authorized, evidence-backed actual rate, signed rights window/expiry, or measured performance baseline to be proposed by the owning workflow; creator-registry alone decides whether it becomes canonical.
• Done when:
• Core metrics are compared against compatible source-dated targets/benchmarks. Missing or incompatible context is Unknown/NOT_SCORED, never an invented /10 score or adjective verdict.
• Creators/platforms/content are ranked only under a declared metric, compatible window/basis, complete candidate set, and preregistered decision rule; descriptive associations and causal hypotheses stay visibly separate.
• Conversions use one declared attribution model with deduplicated, mutually exclusive counted buckets; overlapping promo/UTM/direct observations remain reconciliation evidence, and modeled influence stays Estimated outside the counted total.
• With verified current measured or closed state, each requested next-cycle decision has a scope-bound Campaign Retro Card with campaign/creator/state/measurement/decision-rule refs, evidence-backed rationale, evidence_refs, next-campaign hypothesis, and unresolved limitations; insufficient decision evidence resolves to unknown, while missing/forked state blocks the card.
• Primary next skill: roi-calculator — turn measured performance into dollar-level return.

Handoff Summary

Emit the standard shape from skill-contract.md §Handoff Summary Format.

Data Sources

This family needs no live integrations (Tier 1). The skill runs entirely on inputs you provide — paste platform exports, influencer report screenshots, GA numbers, and promo-code redemption counts, and it analyzes the supported fields. Missing inputs do not block a partial descriptive read, but any dependent score, verdict, rank, causal explanation, attribution total, or decision becomes Unknown/NOT_SCORED/NEEDS_INPUT rather than being filled in.

Where a connector could speed the work, the skill marks it with a ~~ placeholder:

• ~~social platform analytics — native reach/engagement/video metrics per post.
• ~~web analytics — site traffic, click-through, and on-site conversion data.

Measured YouTube post-performance (free key): when campaign content lives on YouTube, python3 "${CLAUDE_PLUGIN_ROOT}/scripts/connectors/youtube.py" videos @creator --limit 20 pulls the actual per-video views/likes/comments for the campaign window — Measured platform metrics without waiting for the creator's screenshot export. Keep both labels honest: API numbers are Measured, creator-supplied numbers are User-provided, and the two can legitimately disagree (display rounding, timing). Free YOUTUBE_API_KEY. See scripts/connectors/README.md.

• ~~ecommerce / sales platform — revenue, orders, AOV, promo-code redemptions.
• ~~influencer database — historical creator benchmarks for comparison.

No placeholder is required to run. See CONNECTORS.md for the verified free/keyless data recipe per category.

Instructions

Work the steps below as one dependency-aware pass. Each fill-in template lives in references/analysis-templates.md. Build the Step 2 shell after intake, but run Step 7 before populating or publishing Step 2 Conversions, Revenue, or any rate/cost that depends on them; those fields must cite Step 7's reconciled counted total or remain Unknown/NEEDS_INPUT.

1. Gather performance data — log campaign/period/influencers/platforms and the available sources (native analytics, influencer reports, web analytics, sales, promo codes). Template: step 1.
2. Analyze core metrics — compare reach, impressions, engagements, ER, video views, clicks, promo uses, conversions, and revenue against compatible source-dated targets/benchmarks. Emit field-level comparison states; do not invent an aggregate score or adjective verdict. Template: step 2.
3. Analyze by platform — compare platforms on compatible reach/ER/click/conversion/CPA windows and state observed differences. Put any explanation in a separately labeled hypothesis unless a designed comparison supports it. Template: step 3.
4. Analyze by creator — use opaque creator_ref; rank only comparable rows under the declared rule. Consume ROI/ROAS only from a cited roi-calculator artifact, do not compute it here, and separate observed content anatomy from causal hypotheses. A renew/retest/retire call comes only from the Retro decision gate. Template: step 4.
5. Content performance analysis — compare formats/themes under compatible exposure and attribution bases. Name observed higher/lower associations; describe a hook/message/visual as causal or "winning" only when the supplied design clears the measurement protocol. Template: step 5.