Planet Crusher: preserve a learning record when creative refresh is automatic

Gryphon's Planet Crusher case describes automated sourcing and refresh. The practical question is how a team retains an understandable experiment history when the asset pool keeps changing.

What the public case establishes

TikTok reports a Planet Crusher campaign running from 23 October to 5 November 2025, using Creative for Smart+ to source, rank and refresh creator material. The account spans install, purchase and value objectives. Its comparative ROAS and CTR figures are brand-reported, with no public raw dataset supplied. TikTok for Business case study.

This is a source-led creative casefile, not a frame-by-frame viewing or a campaign we ran. The diagram below is our own schematic. It does not reproduce the advertiser’s footage, and we have not established reuse permission for that footage.

The decision worth examining

A changing pool can make yesterday's control disappear. If both the set of available assets and their delivery weights move, an aggregate improvement does not explain which concept should be reproduced. The creative team needs a record of eligibility and exposure over time, not just a final list sorted by spend or the last snapshot of an automated dashboard.

The proposed register keeps three dates per asset: admission to the eligible pool, first delivery and removal. It also records why a file left: permission expiry, factual change, manual retirement or automatic refresh. These are our suggested fields. They distinguish a production failure from an asset that was never meaningfully explored and preserve evidence after the optimiser moves on.

A usable next check

For a bounded learning period, retain an approved reference pool while allowing a separate refresh lane. Compare the concept coverage and delivered exposure of both lanes using the same campaign objective. Analyse results only after the relevant cohorts mature. A constantly expanding pool should not be presented as a conventional two-cell creative experiment.

Download the populated case worksheet. Its control, variant and decision rules are a proposed experiment, not an account of results already measured. Keep the source URL with the worksheet when circulating it so reported findings cannot become unattributed team benchmarks.

Evidence boundary

The public source supports the account's workflow description, not an isolated causal effect of automatic refresh. We did not observe the ads, run the campaign or verify the reported comparative figures against an advertiser export.

For background, see the creative brief is a hypothesis document and post campaign creative debrief process. These are existing archive discussions, not independent certification of this case. No first-hand campaign result is asserted.