Metrics: Cross-reference with other datasets (popularity indicator)

[lukepayyapilli]

Context

Related to #2336 (Metrics for downloads) and #2563 (tracking start date). Raw view counts lack context - is 1,000 views good or bad? Comparing against similar datasets would provide meaningful context.

Problem

• Absolute view counts don't indicate relative popularity
• A new dataset with 500 views in 1 month may be more "popular" than an old dataset with 2,000 views over 5 years
• Users have no way to gauge how a dataset compares to others in its category

Proposed Approach

Popularity Indicator

• Calculate a "popularity score" relative to datasets of similar age or category
• Display a subtle indicator (e.g., "Popular", "Highly viewed", or percentile ranking)
• Only show for datasets that meet a threshold (avoid labeling everything)

Calculation Options

1. Age-normalized views: Views per month since publication
2. Percentile ranking: "More views than X% of datasets published in the same year"
3. Category comparison: Compare within resource type (database, software, challenge)

Implementation Details

• Add method to calculate relative popularity
• Cache the calculation to avoid expensive queries on every page load
• Update periodically (e.g., daily) via management command or celery task
• Display only for datasets above a threshold (e.g., top 25%)

Alternatives Considered

1. Show raw percentile: "87th percentile" - precise but may not resonate with users
2. Star rating visual: 1-5 stars based on popularity - familiar but may imply quality judgment
3. No indicator, just comparison table: Show how this dataset ranks among similar ones - more information but more complex UI
4. Trending indicator: Show if views are increasing - different metric but could complement

Design Considerations

• Must not imply quality or endorsement, only popularity
• Should account for the tracking start date issue (Metrics: Display view count tracking start date #2563)
• Consider whether to compare across all datasets or within categories
• Badge/label should be subtle and not dominate the page

Acceptance Criteria

• Popularity is calculated relative to comparable datasets
• Indicator is displayed for datasets meeting threshold
• Calculation accounts for dataset age and tracking limitations
• Clear that this indicates views, not quality
• Performance is acceptable (caching implemented if needed)

[tompollard]

I love the idea of the "Trending indicator", and then being able to display trending resources on the front page: https://physionet.org/ (e.g. instead of "Featured resources", we have "Trending resources").