Lumen Botanics: creator risk and a collapsing reach engine

This is an operating-model failure, not a single creator's error. The brief allowed a claim the product cannot lawfully make, and nothing caught it for 30 hours.

Immediate, week one: freeze new onboarding; audit all posts from the last six months for therapeutic claims, before-and-after imagery and missing disclosure; remove or edit the eczema claim on every channel including creator-owned accounts; log timestamps of removal. Keep a written record - regulators assess whether a system existed, not whether one mistake happened.

Contractual rewrite. A claim library of pre-approved phrases; an explicit ban on medical or therapeutic claims and on before-and-after imagery where category rules disallow it; mandatory platform-native paid-partnership disclosure plus a hashtag; script pre-approval for any claim-bearing content; takedown rights with a stated window, such as two hours for a violation; perpetual paid usage rights; a defined exclusivity period; audit rights and a termination clause.

Process by risk tier. Large-reach creators and anyone handling claims get script and first-cut approval. Smaller creators get approved talking points and post-publication review. Document the brief, the approval and the review for each.

Monitoring. A named owner reviews new posts daily, uses platform brand-safety tools, keeps a shared tracker and a 24-hour SLA for violations.

Accept the cost. Tighter terms will lose some creators and slow output. Fewer, better-paid creators with clean rights is also the right answer commercially - the €31 cost per new customer already says the volume model is failing.

Diagnose before reallocating. Three candidates, tested separately:

• Dilution or saturation: 180 creators repeating similar hooks. If reach per post falls while total reach holds, it is dilution, not platform decline.
• Distribution shift: the platform is favouring different formats or watch-time signals. Check performance by format and length.
• Creative fatigue: cohort posts by hook and track reach decay curves against post age.

A quick test distinguishes them: pull the top decile of creators by new customers and compare their reach trend against the long tail. If only the tail is decaying, the programme has scaled quantity, not quality.

Rebalance. Stop paying for post volume. Retain roughly the top 20% of creators by incremental new customers on retainer terms with usage rights, and reduce the rest to a small always-on test group. Raise paid amplification above €300k but with a rule: only boost creative that beat the organic median on hold rate and click-through, and cap frequency. Boosting underperforming creator content buys reach nobody wanted.

Build the owned hedge. Discord and the newsletter reach the audience without platform permission and convert at a fraction of TikTok's cost per customer. One person and no budget is the actual strategic problem. Fund it, set a subscriber and activation target, and pipe creator content into it so the audience accumulates somewhere the brand controls.

Set portfolio targets: maximum share of new customers from any single platform, cost per new customer by source, and owned-audience growth, reviewed monthly rather than quarterly.

Start by naming the bias. Creator codes and last-click overstate TikTok when TikTok is the last touch and understate it when discovery happens there and conversion happens elsewhere. Neither number answers the budget question.

Build three layers:

1. Descriptive: creator code redemption, new versus returning customers, first-touch versus last-touch, and blended CAC alongside platform-reported CAC. Report the gap explicitly; it is usually large and it is the number that changes behaviour.
2. Causal: geo holdouts. Match regions on prior revenue, suppress paid TikTok and pause creator codes in half for four to six weeks, and compare total revenue and new-customer counts. With 62% of acquisition attributed to one platform, this is the single most important test the brand can run - it either validates the engine or reveals dangerous concentration. Expect leakage through organic TikTok, so treat the estimate as a lower bound.
3. Value: cohort LTV by acquisition source at 90 and 180 days. A channel with a €31 CAC may still be cheap if repeat rate is high, and a €19 CAC may be expensive if the cohort never repurchases. Skincare has enough repeat purchase to make this measurable within two quarters.

Then set the decision rule. If the holdout shows TikTok driving less incremental revenue than attributed, shift budget toward owned channels and paid amplification of proven creative. If it confirms the attribution, the concentration risk is real and still needs a second platform tested - but as a hedge with its own budget, not as a replacement.