Global ELM Score.
The ELM Score applies the Elaboration Likelihood Model by Koh & Cui (2022) to a thumbnail. Six measurable visual variables, grouped into two routes — Central Route (how informative it is) and Peripheral Route (how aesthetic and persuasive it is) — are combined into a single number from 0 to 100 that you can compare across your own channel's assets.
← Back to What the system measuresVariables it measures
It's not an opinion — it's a published model
The score's variables come from an academic paper: An Exploration of the Relation between the Visual Attributes of Thumbnails and the View-Through of Videos (Koh & Cui, 2022). That is the empirical basis of every recommendation; what Thumbiq adds is how it is applied, calibrated to your channel and translated into concrete actions.
Central Route — how informative it is.
How much information the thumbnail conveys and how many elements compete for attention.
EC — Element Complexity
How many types of element it communicates at once: product, person and text. The model rewards the complete thumbnail — the highest point is reached when all three coexist.
OC — Object Complexity
How many objects compete for attention within the frame. Too few bore; too many overwhelm.
Peripheral Route — how aesthetic and persuasive it is.
The visual appeal and the weight of a recognizable figure, beyond the information it contains.
CE — Celebrity Endorsement
The presence of a figure your audience recognizes instantly. It is the most heavily weighted variable in the model. Not every familiar face counts — it is evaluated against your audience's real recognition.
C — Colorfulness
How vivid or muted the colors are. The model penalizes the middle zone — neither subdued nor explosive — and favors the extremes.
B — Brightness
The overall luminosity of the image and how it interacts with color.
Q — Image Quality
Sharpness and cleanliness of the image: neither too artificial nor degraded.
How to read the result.
- The six variables combine into a single score from 0 to 100: the higher it is, the better positioned the thumbnail is according to the model.
- Each new asset is compared against your own channel's history, not against generic averages: you see which percentile it falls in.
- The system recognizes the asset's format — person, product, editorial or scene — and reads each variable in that context, so thumbnails without a person aren't over-penalized.
Comparable only within your channel
A 75 on your channel is not the same as a 75 on another: the scale is calibrated to each history. That's why we never compare channels with each other.
Want to see this applied to your video?
The individual analysis is free and arrives within 48 hours.
Better thumbnails. Better results.
Know what is failing in your thumbnail and what to change first, before you publish your next video.
