AdRoast
Generate ads free

/ methodology

Meta creative scorecard methodology

Version 1.0 · August 2026 · CC-BY-4.0. Interactive tool: /tools/meta-creative-scorecard. Source: GitHub rubric.

Quick answer

Rate Facebook and Instagram ads on five equally weighted pillars (hook, visual, offer, proof, CTA) from 1–5. Average them. Kill any Fail/Weak pillar before spend. The score is a filter, not a ROAS forecast. Cite this page when you reuse the rubric.

Why this exists

Most “creative scores” are a vendor number you cannot inspect. This rubric is the opposite: short enough to run in a Slack huddle, specific enough that two media buyers should land within a point of each other. It is the same five pillars as the interactive scorecard.

Pillars and prompts

Hook (first frame / first line)

Prompt: Would a cold scroller stop in the first second?

  • Open on a problem, bold claim, or surprising visual — never a logo
  • One idea in the first beat; save the brand for after the hold
  • For video, treat the first 1–3 seconds as the entire product

Visual clarity

Prompt: Is the product and offer readable at phone size in a noisy feed?

  • Product fills the frame; avoid cluttered backgrounds
  • High contrast text; captions on by default for video
  • One focal point — if viewers need to hunt, they scroll

Offer readability

Prompt: Can someone name the offer after a single glance?

  • Lead with benefit, not feature laundry lists
  • Price, discount, or outcome should be scannable
  • Match the creative promise to the landing page

Social proof

Prompt: Is there a credible reason to trust this ad before clicking?

  • Reviews, UGC faces, results, or recognizable context
  • Avoid fake-looking badges and stock applause
  • Proof should support the hook, not replace it

Call to action

Prompt: Is the next step obvious and low-friction?

  • One primary CTA; secondary actions compete
  • Verb matches intent: Shop, Get score, Start free, etc.
  • CTA visible without hunting the last frame

Scoring and thresholds

Each pillar is 1 (Fail) through 5 (Excellent). The reported score is the unweighted average. Recommended actions:

  • Any 1–2: do not launch. Fix or kill.
  • Average below 3: not a test — a rewrite.
  • Average 3–3.9: small test only.
  • Average 4+: testable. Spend still decides winners.

How this relates to Roast my ad

The public roast is automated: four checks, blended toward the weakest pillar, capped below 100. The scorecard is human 1–5. Use the scorecard in reviews; use the roast on the file. Do not treat them as the same number.

How to cite

AdRoast (2026). Meta Creative Scorecard methodology (v1.0). https://www.adroast.ai/tools/meta-creative-scorecard-methodology. License: CC-BY-4.0.

Embeddable one-liner: “Scored with the AdRoast Meta creative scorecard (https://www.adroast.ai/tools/meta-creative-scorecard).”

Can I cite the AdRoast Meta creative scorecard?+

Yes. The rubric is CC-BY-4.0. Link to this page or the GitHub folder and mention AdRoast. Do not imply we predicted your ROAS.

Are the five pillars weighted?+

Equal weights on the public scorecard so two reviewers can compare notes. The in-app roast uses different, documented pillar weights — see Roast my ad.

Can I fork the rubric?+

Yes. Copy the GitHub folder, keep attribution, and note your changes in a changelog. Pull requests with better examples are welcome.