AdLens benchmark methodology

How AdLens benchmarks video ad scores

Benchmarks place an AdLens score within an eligible comparison set. They describe relative position—not guaranteed campaign performance—and are kept separate from the models that score the creative.

1. The score is produced

Multimodal AI evaluates the creative’s visual, verbal, audio, brand, and time-based signals across six metrics. That process produces a 0–100 score with rationale and evidence.

How scoring works

2. The score is compared

Eligible completed analyses with a valid score form a comparison distribution for that metric. AdLens can then calculate the average, percentile thresholds, and the selected score’s percentile position.

Comparison scope

Global report benchmark and filtered explorer

Analysis report: Global benchmark

The benchmark shown inside an analysis report uses the eligible global comparison set for each metric. It is not automatically filtered to the video’s category or language.

This keeps the report label honest while contextual cohort rules are developed.

Benchmark explorer

The Benchmark page lets users explore available category and language filters separately. Its results apply only to the active filter selection.

Placement is not treated as a default cohort dimension because it is optional and currently supplied less consistently.

What averages and percentiles mean

Average

The arithmetic mean of valid scores in the active eligible comparison set.

75th percentile

A score at this threshold is at least as high as approximately 75% of scores in the comparison set.

Percentile rank

The approximate share of eligible scores at or below the selected score. It is relative to the cohort, not a probability of success.

Cohort-quality rules

AdLens calculates sample size internally to decide whether a benchmark is suitable to show. Publishing the exact count is a separate product decision and may depend on account access.

Fewer than 30

Benchmark withheld

AdLens does not present the cohort as a reliable comparison. The creative score can still be reviewed on its own.

30–99

Limited comparison sample

The benchmark can be shown with a visible caution because percentile and average estimates can move more easily.

100 or more

Standard presentation

No special sample-quality warning is required, though all benchmark limitations still apply.

The five public demo reports are not treated as market statistics. The benchmark preview directs users to the eligible live dataset instead.

Privacy

Benchmarks use score-level comparison data. Other users cannot open your uploaded video or private report through a benchmark. Public demo reports are separate, intentionally curated examples.

Limitations

Benchmarks can shift as the eligible comparison set grows or filters change.

They do not control for media spend, targeting quality, placement delivery, brand equity, price, or landing-page experience.

Compare compatible analysis releases and review creative evidence alongside the relative score.

Explore the full live benchmark

Create an account to explore available metrics, categories, languages, percentile thresholds, and time ranges in the eligible live dataset.

Open full benchmark