A small ad-reference board: observation is not evidence that an idea will win

[LMSlay]

For AFFiNE readers who turn reference videos into a creative brief, here is a small manual board outline, not an integration or a tested AFFiNE template. The useful handoff is not just collecting examples: it is keeping what was observed separate from what someone inferred and what the team wants to try.

AFFiNE's core concepts describe Page and Edgeless views of a doc. Its getting-started guide demonstrates reference notes beside a brief. That offers a place to arrange the following four labels manually.

Entirely synthetic example — not a real ad, Actor output or campaign result:

Board label	Example note	What it does NOT establish
Source observation	Fictional reference A, playback 00:04: a hand folds a bag into a pocket. Add the source link and the date you viewed it when using a real reference.	Who bought it, why they bought it, or whether the ad performed well.
Interpretation, including AI if used	“This may emphasize portability.” Keep it labelled as an interpretation and check it against the source.	A verified audience motivation or a guaranteed field in any product's output.
Our creative hypothesis	“For our own bag, try showing the packed size before listing its features.”	Permission to reuse the reference's footage, copy or other assets.
Checks still needed	Confirm the observed moment and surrounding context; verify our product claim and asset rights; define what a later, separately authorized evaluation would measure.	Completed testing, spend, conversions or a winning creative.

If you connect the notes visually, label the connection “suggests a question”, not “proves.” Preserve the decision and its uncertainty in the Page text as well: the official guide cautions that canvas layout and visibility are not the same as the linear page.

The companion four-part creative research brief expands the observation / interpretation / original hypothesis / review distinction. This outline does not automatically import video or invoke a model; AI is not required to use the labels. It has not been tested as an AFFiNE workflow, and it makes no performance or privacy guarantee.

Disclosure: I maintain TubeText, which offers optional paid Apify tools for video analysis. The linked guide is public; no purchase is needed to use this manual outline. This post was prepared with AI assistance.