How Olvyen scores creative

What Olvyen evaluates and what it does not, how a score is produced, and what has and has not been tested. This page describes Olvyen Methodology 1.9.0.

1. What a result contains

Every result has four parts: a composite score from 0 to 100; one of four verdicts; six dimension scores, each weighted and each with a reason; and up to three fixes in priority order, marked on the creative where the location matters. When the evaluation finds a defect, the result names it in place of the verdict's description, and its fix is fix 1.

2. The six dimensions

Each dimension is scored from 0 to 100. The composite combines them by these weights, which are published and stamped on every score as rubric 1.9.0. The weights are the same as in 1.8.0.

Hook Strength is tension, not noise. A hook earns nothing for being loud. These earn nothing: starbursts, repeated exclamation marks, stock sale templates, rainbow gradients, more than two typefaces, and a phone number as the primary element. A creative that stops the scroll because it looks like spam has not hooked anyone; it has been skipped faster.

Platform Fit on an organic placement is judged on that surface's own conventions, not an ad's. An organic post should read as a post, and the caption and the reply count as the call to action: CTA Strength there is judged on the native mechanism (a reply, a sticker, a caption), never on a button. Each organic placement is judged on these terms:

Calibration for organic placements can only come from the insights Instagram and TikTok expose for organic posts; Facebook groups and WhatsApp expose none, so scores on those placements cannot be calibrated against outcomes.

3. The four verdicts

A creative that carries a defect (section 4) is never Test or Scale: its verdict is capped at Rework. The composite is shown as computed, so a 64 with a defect reads Rework beside its number, and the result says why: "The creative carries unreadable text; nothing else is judged until that is fixed, and the verdict is capped at Rework."

4. How a score is produced

Each creative receives three separate evaluations at fixed settings, reconciled into one result: the composite and each dimension score are the median of the three. The same file under the same conditions returns the same result.

The defect check. Before any dimension is scored, each evaluation looks for a defect that breaks trust:

• Unreadable or gibberish text
• Lorem ipsum or placeholder text
• A broken, stretched or missing image
• A visible template artifact
• The creative shown inside a mockup (on a device, a wall, a shadowed card or another frame) instead of as the asset itself

Each evaluation is also asked directly whether the creative is framed this way. A defect counts when at least two of the three evaluations find it. A defect caps Message Clarity at 40 and Platform Fit at 50, its fix becomes fix 1, and the verdict is capped at Rework (section 3). For a mockup the fix is always the same: upload the asset itself at its native size, without the device, wall, card or other frame around it.

The two rules every fix follows.

• A fix is achievable in the asset's medium. A still image never receives motion, video or animation advice. On a video cover, the fix addresses the cover frame.
• A fix never invents a claim, number, testimonial, award or statistic. It may say to add a proof point if you have one, or to state the benefit the product actually delivers; it never supplies the content of the claim.

The check after every evaluation. Each evaluation also transcribes the text it reads in the image. Then every fix is checked against the two rules and the placement's rules (section 5), and is rejected when it:

• Asks for motion, video or animation on any placement that is not a video cover
• States a number as a claim (a proof such as "trusted by" or "rated", a duration, or an outcome such as "20% off") that neither the submitted copy nor the text in the image contains
• Prescribes an element the placement forbids (a fix that removes one is allowed)

An evaluation with a rejected fix is run once more, with the violation named. A fix that still breaks a rule is dropped, and so is a fix that repeats the defect's fix or an earlier one, so a result can carry fewer than three fixes rather than a broken or repeated one.

A new evaluation runs only when one of these changes:

• The file
• The placement
• Paid or organic
• The creative intent
• The carousel card role
• The brand profile
• The methodology version

5. Declaring what a creative is for

A creative is judged against the job it is doing. The placement sets the platform, the asset type and whether the creative is paid or organic. Intent follows from the placement and can be changed: direct response judges all six dimensions; brand awareness removes CTA Strength, and a missing call to action is not a defect. Organic placements default to brand awareness, except a Facebook Marketplace listing and a WhatsApp status, which default to direct response. A carousel card declares its role: opening, supporting or closing. A supporting card does not carry the call to action, so CTA Strength is removed there too.

Placement rules. Some placements have rules of their own: elements the surface does not allow, and dimensions it cannot carry. The evaluation receives the placement's rules as written, with one sentence of what the surface is judged on (for example: an Amazon main image is judged on product rendering, isolation and legibility at thumbnail size; Hook Strength there means the product itself stops the scroll in search results). It never prescribes a forbidden element and never scores a dimension the placement excludes.

Every other paid placement has no rules of its own and is judged on its platform and asset type, as before 1.9.0.

A removed dimension is not scored as zero and is not shown. The remaining weights are rescaled so they account for the whole score, and the creative can still reach 100.

6. What Olvyen does not judge

• Audience quality or targeting
• Bid strategy
• Campaign structure
• Offer economics
• Landing-page performance
• Product-market fit
• Media saturation
• Actual return on ad spend

7. What has been tested, and what has not

Olvyen judges readiness. It does not guarantee campaign performance.

• Tested against real ads within a single platform; agreement with an expert reviewer was measured. Calibration against live campaign outcomes has not been done.
• Scores do not yet compare across platforms. A 60 on Meta and a 60 on Amazon are not yet the same 60, so compare creatives within one placement.

8. Versions

Every score records the methodology version it was produced under. When the methodology changes, past scores keep their version and are not altered, and two versions of a creative scored under different methodology versions are not compared fix by fix.

• 1.9.0, 2026-10-12. Placements gained rules of their own: what each surface forbids, and the dimensions it cannot carry, which are removed rather than scored. Hook Strength now rewards tension, specificity and surprise and gives nothing for volume. A defect such as unreadable or placeholder text, a broken image, a template artifact or a mockup frame caps Message Clarity and Platform Fit, becomes fix 1 and keeps the verdict at Rework or below. Fixes can no longer ask for motion on a still image, invent a claim or number, or prescribe an element the placement forbids, and every fix is checked for this after the evaluation. Organic placements were added, with Platform Fit judged on each surface's own conventions. The weights did not change.
• 1.8.0, 2026-07-31. A creative declares its intent and, on a carousel, its card role. Brand awareness and a supporting carousel card remove CTA Strength, and the remaining weights are rescaled to account for the whole score.