Working · 2022-05SecondaryAcquisitionPaid

Test wildly different 'big swing' creatives first

On a new ad account, test many very different concepts, cut weak ones, then refine the winners.

Claim and evidence

The claimOn a new ad account, test many very different concepts, cut weak ones, then refine the winners.

No dated example yet. Treat this tactic as a hypothesis.

What the numbers show

No published number for this tactic has passed our word-for-word check yet, so there is no chart.

Why it works

Diagram, not data
Concept: the painConcept: social proofConcept: product demoEach in its own ad setCut weak ones,iterate on winners
Concept:the painConcept:socialproofConcept:productdemoEach in its own ad setCut weak ones, iterate onwinners

Different concepts reveal which angle works; small variations waste budget.

How to run it

  1. Produce many varied concepts
  2. Run each in its own ad set
  3. Wait about three days of spend
  4. Cut weak ads
  5. Iterate on winners

Ways to run it

Design options to test, not promised results.

  1. Option AMany concepts, one budget

    Launch several very different ideas at once and cut the weak ones after a short spend window.

  2. Option BFormat swings

    Vary the format itself, such as talking head, screen recording or meme, not only the copy.

  3. Option CRefine after the swing

    Only once an angle wins, test small variations of hooks and edits around it.

Dated examples

No dated example yet. Treat this tactic as a hypothesis.

Does it still work?

Workingas of 2022-05

2022 TikTok-specific advice; rule of scaling at most 25% per week also from Demand Curve. source

Risks

  • TikTok-specific
  • Dated

Sources

  1. demandcurve.com · Demand Curve · Dennis Buckley · 2022-05-09
    “test a high volume of wildly different "big swing" ad creatives”
    ✓ quote checked word-for-word on the page
  2. demandcurve.com · Demand Curve · Dennis Buckley · 2022-05-09
    “Scale ad spend no more than 25% per week”
    ✓ quote checked word-for-word on the page

Change log

  • 2026-11-20 — harvest v0.1 (acq-paid)