Creator Performance Audit

SKILL.md

---
name: creator-performance-audit
description: Audit creator/influencer performance for commerce teams — score creators on reach quality, engagement authenticity, conversion contribution, and cost efficiency, then output a keep/scale/renegotiate/drop decision with evidence.
---

# Creator Performance Audit

Turn scattered creator campaign data into a clear keep / scale / renegotiate / drop decision.

## Quick Reference

| Decision | Strong | Acceptable | Weak (avoid) |
|---|---|---|---|
| Success metric | Revenue or profit per creator dollar (tracked links/codes) | Engaged-view CPM + assisted conversions | Follower count, raw likes |
| Attribution | Dedicated codes/links + post-purchase survey | Platform attribution with stated window | "Sales went up that week" |
| Engagement check | Engagement rate vs account's own 90-day baseline | Vs category benchmark | Vs global average across niches |
| Authenticity | Comment quality sampling + follower growth curve | Spot-check comments | Trust the media kit |
| Cost basis | Full cost: fee + product + shipping + usage rights + agency | Fee + product cost | Fee only |
| Verdict | One of keep/scale/renegotiate/drop per creator, with thresholds | Ranked tiers | "Mixed results, continue monitoring" |
| Sample window | Last 3-6 campaigns or 90 days | Last 2 campaigns | Single post |

## Solves

1. **Vanity-metric decisions** — creators get renewed for likes and views that never convert.
2. **No cost truth** — fees look fine until product, shipping, usage rights, and agency cuts are added.
3. **Fake or drifted audiences** — inflated engagement, bought followers, or an audience that no longer matches the buyer profile.
4. **One-post judgments** — a viral fluke or a flop drives the renewal decision instead of a stable sample.
5. **Un-comparable creators** — TikTok CPMs vs Instagram saves vs YouTube AVD, with no common scorecard.
6. **No action at the end** — audits that end in "interesting" instead of keep/scale/renegotiate/drop.

## Use when

- Deciding whether to renew, scale, renegotiate, or drop specific creators
- Building a quarterly creator portfolio review
- Comparing creators across platforms on one scorecard
- Vetting a creator's real performance before a bigger commitment

## Do not use when

- Prospecting brand-new creators with no campaign history (that's vetting, not auditing — this skill only covers pre-commitment vetting checks partially)
- You need raw data extraction without synthesis
- The question is creative strategy, not performance

## Inputs

- Creator list with platform(s) and campaign history (posts, dates, deliverables)
- Cost data: fees, product cost, shipping, usage rights, agency commissions
- Performance data: impressions/views, engagement, clicks, conversions, revenue (tracked links/codes if available)
- Business context: target audience, margin structure, campaign goals
- Optional: category benchmarks, past audit results

## Workflow

### Step 1 — Fix the question and the window

Define what "performing" means for this business (profit per dollar, CAC vs blended CAC, content library value) and set the sample: minimum 2-3 campaigns or 90 days per creator. Note any creator being judged on a single post — flag as low-confidence.

### Step 2 — Assemble full cost per creator

Fee + product cost + shipping + usage/whitelisting rights + agency commission + internal handling. This is the denominator for everything. See references/metrics-benchmarks.md for the cost checklist.

### Step 3 — Normalize performance across platforms

Compute the common scorecard (references/metrics-benchmarks.md): engaged-view CPM, engagement rate vs the creator's own baseline, CTR, conversion rate, revenue per creator dollar. Platform-specific metrics stay as secondary evidence.

### Step 4 — Run authenticity checks

Follower growth curve (spikes = suspicion), engagement-to-follower ratio vs category norms, comment quality sampling (generic emoji walls vs real questions), audience geo/demo match to buyer profile. See references/audit-framework.md.

### Step 5 — Score and classify

Score each creator on the four axes (reach quality, engagement authenticity, conversion contribution, cost efficiency) and map to a verdict: **keep / scale / renegotiate / drop**, using the thresholds in references/audit-framework.md.

### Step 6 — Write the action plan

Per creator: verdict, evidence (2-3 numbers), and the concrete next step (new rate, different deliverable, usage-rights renegotiation, replacement candidate). Portfolio-level: concentration risk, budget reallocation.

### Step 7 — Self-check

Run assets/audit-checklist.md before delivering. Output follows references/output-template.md.

## Worked Example 1 — Three creators, one budget

**Input (summary):** DTC skincare brand, 3 creators over 90 days.

| Creator | Platform | Full cost | Views | Eng. rate | Tracked revenue |
|---|---|---|---|---|---|
| A | TikTok | $6,800 (3 posts) | 2.1M | 6.2% | $4,100 |
| B | Instagram | $4,500 (4 posts) | 380K | 4.8% | $9,700 |
| C | TikTok | $9,000 (3 posts) | 4.6M | 1.1% | $2,300 |

**Output (verdict section):**

> **B — SCALE.** Revenue per dollar 2.16, best in portfolio; engagement steady vs her 90-day baseline (4.8% vs 4.5%); comments show purchase-intent questions. Action: propose 6-post quarterly package before her rates adjust; add usage rights for ads — her post #2 outperformed our best paid creative.
> **A — KEEP (watch).** RPD 0.60 looks weak, but 71% of tracked conversions came from the one post with a direct product demo; the two "lifestyle" posts converted ~0. Action: renew at same rate but change the brief to demo-format only; re-audit after 2 posts.
> **C — DROP.** RPD 0.26 with the portfolio's highest cost. Engagement 1.1% vs TikTok category norm ~4%; follower curve shows two +80K single-day spikes (likely purchased); comments are generic emoji strings. The 4.6M views are not reaching buyers. Action: do not renew; reallocate $9K → B's package ($6K) + test one new demo-style creator ($3K).

## Worked Example 2 — Single creator renewal call

**Input (summary):** Gadget brand, creator on YouTube, $12K for 2 dedicated videos + 1 short. Views 640K total, AVD 4:10 on 9-minute videos, 3,900 clicks, 210 orders tracked ($18,900 revenue), gross margin 38%.

**Output (verdict section):**

> **RENEGOTIATE.** Revenue per dollar 1.58 but margin-adjusted return is 0.60 — below the 0.8 renewal bar. The content itself performs (AVD 46%, CTR 0.61%, CVR 5.4% — all healthy), so the problem is price, not the creator. Action: counter at $8.5K for the same package (breakeven at current conversion is $7.2K) or same fee with usage rights + a dedicated coupon push in the first 48h, where 63% of his conversions happened. Low-confidence note: 2 videos is a thin sample; lock a 2-video test at the new rate before any annual deal.

## Common Mistakes

1. **Judging on follower count or raw views** — reach that doesn't hit buyers is a cost, not an asset.
2. **Fee-only cost math** — product, shipping, usage rights, and agency cuts routinely add 30-60% to true cost.
3. **Comparing engagement across niches** — a 2% rate can be great in one category and terrible in another; compare to the creator's own baseline first.
4. **Ignoring the content variable** — a "bad creator" may be a bad brief; check whether specific formats converted before dropping.
5. **Single-post verdicts** — always state sample size and mark thin samples low-confidence.
6. **Skipping authenticity checks** — follower spikes and emoji-wall comments invalidate every downstream metric.
7. **No margin adjustment** — revenue per dollar above 1.0 still loses money on thin margins.
8. **Ending without a verdict** — every audited creator gets keep/scale/renegotiate/drop plus a next step.
9. **Ignoring content library value** — usage rights for a top-performing creative can justify an otherwise marginal deal; price it explicitly.

## Resources

- `references/output-template.md` — standard audit report structure
- `references/metrics-benchmarks.md` — scorecard metrics, formulas, and cost checklist
- `references/audit-framework.md` — four-axis scoring, authenticity checks, verdict thresholds
- `assets/audit-checklist.md` — pre-delivery quality checklist