Servo BWA Feature Usage & Prioritization Report

How often are web platform features actually used in the real world? This report explores a framework for answering that question, applied to the 415 features in the W3C’s Baseline “Widely Available” (BWA) set — the features that the core browser set supports. We combine data from 6 independent sources: HTTP Archive page scans (~10.9M websites), Chrome telemetry (all Chrome page loads), and Firefox telemetry (desktop and mobile). Each feature gets a composite score estimating the fraction of web pages that use it (0–100%), plus a confidence level based on how many sources agree.

Data Source Coverage

Each number shows how many of the 415 BWA features that source can measure. The weight (w) controls how much each source influences the composite score.

383

Have at least one signal — 92.3% of 415 BWA features

No data from any source

107

HTTP Archive (specific) — page content scans of ~10.9M sites · w=0.35

HTTP Archive (generic) — indirect parent-property matches · w=0.03

HTTP Archive (Blink) — JS API detection on crawled pages · w=0.20

383

ChromeStatus — Chrome telemetry across all page loads · w=0.15

Firefox Desktop — desktop telemetry · w=0.15

Firefox Fenix — mobile telemetry (only mobile signal) · w=0.15

Composite Score Tiers

The composite score estimates the fraction of web pages that use a feature. Features are grouped into tiers by that score. The right column shows what share of all 415 BWA features fall into each tier.

>50% of pages	14 features	3.4% of BWA
10–50% of pages	93 features	22.4% of BWA
1–10% of pages	100 features	24.1% of BWA
<1% of pages	176 features	42.4% of BWA
No data	32 features	7.7% of BWA

Confidence Distribution

Confidence reflects how well-corroborated a score is. More independent sources that agree = higher confidence. Confidence does not filter — all features are shown regardless of level.

High	3+ sources, all within 20 percentage points of each other	67
Medium	2 sources, or 3+ sources that disagree by more than 20 percentage points	150
Low	1 source only	166
None	No data from any source	32

Cross-validation: 83 features have both Chrome and Firefox data

Composite Score Distribution

How many features fall into each composite-score range. Most features are used on fewer than 10% of pages — only a small set of widely-used CSS properties and APIs appear on more than half of all pages.

Top 30 Features by Composite Score

The highest-usage features across all sources. Bar color indicates confidence: green = high (3+ agreeing sources), amber = medium (2 sources or wider disagreement), blue = low (single source).

Weighted Composite Scoring Methodology

All sources contribute independently. Each source tells a different part of the story: site-weighted crawl data, Chrome telemetry, Firefox telemetry, mobile telemetry. The composite score blends them using measurement-class weights.
Weights by measurement class: HTTP Archive page scans (specific)=0.35, HTTP Archive (generic)=0.03, HTTP Archive Blink detection=0.20, ChromeStatus=0.15, Firefox desktop=0.15, Firefox Fenix=0.15. When a source is missing, its weight is redistributed proportionally among available sources.
Generic observables: Some features can only be detected via a parent property (e.g., subgrid is detected through grid-template-columns). These carry minimal weight (0.03) and are flagged, since the measurement is an upper bound — it confirms the parent property is used, not necessarily the specific sub-feature.
Blink artifact discount: When Blink UseCounters report that canvas, webgl, or media interfaces appear on more than 45% of crawled pages, we halve that value before weighting. These interfaces show suspicious clustering (many report ~53.5% of pages) that likely reflects detection methodology rather than true usage.
Firefox desktop and Fenix are separate signals, each weighted 0.15 — the same as ChromeStatus. Fenix is the only mobile usage signal in the dataset and carries equal weight to any other single telemetry source.
All features are included regardless of confidence. High/medium/low/none confidence levels and flags annotate data quality so experts can evaluate the backing data themselves.
Scale: All raw values are fractions from 0 to 1 (displayed as 0–100%). HTTP Archive values = fraction of ~10.9M crawled websites that contain the feature. ChromeStatus values = fraction of all Chrome page loads globally that trigger the feature. Firefox values = fraction of all Firefox page loads that trigger the feature. These denominators differ, so raw values across sources are not directly comparable — the composite score uses measurement-class weights to blend them.

Every Baseline “Widely Available” feature is listed below. Score is the weighted composite estimate of the fraction of web pages that use the feature. The individual source columns show raw values from each measurement system — these use different denominators and are not directly comparable. HA % = % of ~10.9M crawled websites containing the feature. Chrome % = % of all Chrome page loads triggering the feature. FF Desktop/Mobile % = % of Firefox page loads triggering the feature. Click any column header to sort.

#	Grp	Feature	Score	Conf	HA%	Chrome%	FF Desk%	FF Mob%	Src	Flags

For 83 features, we have usage data from both Chrome-ecosystem sources (HTTP Archive or ChromeStatus) and Firefox telemetry. This table compares them side by side. Delta is the absolute difference between the best Chrome signal and Firefox desktop. When Chrome and Firefox independently agree on usage levels, we can be more confident in the estimate. Large divergences may indicate measurement differences, browser-specific implementations, or features with genuinely different adoption across browsers. Sorted by biggest disagreements first.