data trove accessibility audit tools
Reading
This dataset is a web accessibility audit of approximately 100 top websites, covering 92 rows with metrics on error counts, error density, popularity rank, and automated WAVE tool scores. The most striking finding is that both 'errors' and 'error_density' are heavily right-skewed with extreme outliers — the median error count is just 5, but the max reaches 364, suggesting a small cluster of sites are dramatically worse than the rest. A second angle worth exploring is the 'notes' column, where 'Low contrast text' dominates as the most common accessibility issue (12 occurrences), pointing to a systemic problem across high-traffic sites. The near-uniform distribution of 'popularity_rank' suggests the sample spans the full top-100 range evenly, making comparisons across popularity tiers feasible.
citing: errors.median · errors.max · errors.skew · errors.n_outliers · error_density.skew · error_density.n_outliers · notes.top_value · notes.top_rate · popularity_rank.skew · popularity_rank.min · popularity_rank.max
Charts the summary said to look at first
Show data table
| bin | count |
|---|---|
| 0 – 40.44 | 70 |
| 40.44 – 80.89 | 7 |
| 80.89 – 121.3 | 1 |
| 121.3 – 161.8 | 2 |
| 161.8 – 202.2 | 0 |
| 202.2 – 242.7 | 2 |
| 242.7 – 283.1 | 0 |
| 283.1 – 323.6 | 0 |
| 323.6 – 364 | 1 |
Show data table
| bin | count |
|---|---|
| 0 – 0.03367 | 65 |
| 0.03367 – 0.06733 | 10 |
| 0.06733 – 0.101 | 2 |
| 0.101 – 0.1347 | 2 |
| 0.1347 – 0.1683 | 0 |
| 0.1683 – 0.202 | 1 |
| 0.202 – 0.2357 | 2 |
| 0.2357 – 0.2693 | 0 |
| 0.2693 – 0.303 | 1 |
Show data table
| value | count | share |
|---|---|---|
| Low contrast text | 12 | 13.0% |
| Missing form input label | 9 | 9.8% |
| Low contrast text, missing alt text | 9 | 9.8% |
| Missing alt text | 5 | 5.4% |
| Asia-based | 5 | 5.4% |
| No detected errors | 5 | 5.4% |
| Low contrast text, empty link | 4 | 4.3% |
| Low contrast text, missing alt text, empty link, empty button | 4 | 4.3% |
| No data | 3 | 3.3% |
| Low contrast text, missing alt text, missing labels | 3 | 3.3% |
| Low contrast text, missing alt text, empty button | 3 | 3.3% |
| Missing document language | 2 | 2.2% |
| Missing alt text, missing document language | 2 | 2.2% |
| Missing form input label, empty button | 2 | 2.2% |
| Low contrast text, missing alt text, empty link, missing form labels, empty button | 1 | 1.1% |
| Low contrast text, missing alt text, empty links | 1 | 1.1% |
| Missing alt text, missing form labels, empty buttons | 1 | 1.1% |
| Missing form input label, missing document language | 1 | 1.1% |
| Low contrast text, missing alt text, missing language | 1 | 1.1% |
| Missing alt text, missing form labels | 1 | 1.1% |
Show data table
| bin | count |
|---|---|
| 1 – 12 | 11 |
| 12 – 23 | 5 |
| 23 – 34 | 10 |
| 34 – 45 | 10 |
| 45 – 56 | 9 |
| 56 – 67 | 9 |
| 67 – 78 | 9 |
| 78 – 89 | 9 |
| 89 – 100 | 11 |
Show data table
| bin | count |
|---|---|
| 1183 – 1.112e+05 | 27 |
| 1.112e+05 – 2.212e+05 | 17 |
| 2.212e+05 – 3.312e+05 | 10 |
| 3.312e+05 – 4.411e+05 | 5 |
| 4.411e+05 – 5.511e+05 | 7 |
| 5.511e+05 – 6.611e+05 | 6 |
| 6.611e+05 – 7.711e+05 | 2 |
| 7.711e+05 – 8.811e+05 | 4 |
| 8.811e+05 – 9.911e+05 | 5 |
Schema
6 columns| Alerts | ||||
|---|---|---|---|---|
| domain | categorical | 0.0% | 92 |
long_tail
|
| wave_rank | numeric | 9.8% | 83 |
|
| popularity_rank | numeric | 9.8% | 82 |
|
| errors | numeric | 9.8% | 36 |
high_skew
outliers
|
| error_density | numeric | 9.8% | 64 |
high_skew
outliers
|
| notes | categorical | 0.0% | 38 |
long_tail
|
domain
categorical identifier long_tailThis column contains internet domain names (e.g., google.com, youtube.com, wikipedia.org), functioning as a unique identifier for each row in the dataset. With 92 rows and 92 unique values, cardinality is 100% and entropy_ratio is exactly 1.0, meaning every domain appears exactly once — the top_rate of 0.0109 confirms this. There is no distributional signal here beyond the list itself, making the column unsuitable as a categorical feature without further engineering. Treatment: Use as a row key or for joins; extract TLD / domain-level features (e.g., suffix, brand) before any modelling.
- n
- 92
- nulls
- 0 (0.0%)
- unique
- 92
- top_value
- google.com
- top_rate
- 0.01087
- cardinality
- 92
- entropy
- 6.524
- entropy_ratio
- 1
wave_rank
numeric featureThis column appears to be a numeric rank or score assigned within a 'wave' (likely a survey wave or data collection round), with values ranging from 1,183 to 991,094 — a span suggesting population-scale ranking rather than a small cohort. The distribution is moderately right-skewed (skew ≈ 0.96) with a wide IQR of 395,414.5 and a mean (301,136) well above the median (203,569), indicating many lower-ranked entries but a long tail of high-rank values. Notably, 83 of 92 non-null values are unique, consistent with a rank-like identifier, yet ~9.8% of rows are null, which warrants investigation before use. Treatment: Investigate nulls (9.78% missing) before modelling; consider rank-normalization or log-transform to reduce skew prior to regression.
- n
- 92
- nulls
- 9 (9.8%)
- unique
- 83
- min
- 1,183
- max
- 991,094
- mean
- 3.011e+05
- median
- 203,569
- std
- 2.746e+05
- q1
- 8.221e+04
- q3
- 477,628
- iqr
- 3.954e+05
- skew
- 0.9633
- kurtosis
- -0.1876
- n_outliers
- 0
- outlier_rate
- 0
- zero_rate
- 0
popularity_rank
numeric featureThis column is a popularity rank, almost certainly an ordinal score from 1 to 100 assigned to each record. Its distribution is strikingly uniform: mean 51.2, median 52.0, IQR of 49.0 spanning Q1=27.5 to Q3=76.5, near-zero skew (-0.05), and a platykurtic shape (kurtosis -1.19), all consistent with an approximately flat distribution across the full 1–100 range. With 82 unique values out of 92 non-null rows, some ranks are shared between records, suggesting either ties or binned scoring rather than a strict unique ranking. The 9.78% null rate warrants attention—records missing this field may represent unranked or newly added items. Treatment: Use as-is or normalize to [0,1]; investigate null records to determine if missingness is informative before imputing.
- n
- 92
- nulls
- 9 (9.8%)
- unique
- 82
- min
- 1
- max
- 100
- mean
- 51.23
- median
- 52
- std
- 29.39
- q1
- 27.5
- q3
- 76.5
- iqr
- 49
- skew
- -0.05046
- kurtosis
- -1.192
- n_outliers
- 0
- outlier_rate
- 0
- zero_rate
- 0
errors
numeric feature high_skew outliersThis column likely represents an error count per observation (e.g., system errors, validation failures, or defects), with values ranging from 0 to 364. The distribution is severely right-skewed (skew = 3.80, kurtosis = 16.12): the median is only 5.0 while the mean is 27.17 and the std is 57.20, indicating a small number of extreme cases are pulling the average dramatically upward. Eight outliers (9.6% of non-null rows) drive this effect, and ~9.8% of rows are null — both figures warrant investigation before modelling. Treatment: Log-transform (log1p) or apply a robust scaler before modelling; investigate and impute or flag the 9.78% nulls and 8 extreme outliers separately.
- n
- 92
- nulls
- 9 (9.8%)
- unique
- 36
- min
- 0
- max
- 364
- mean
- 27.17
- median
- 5
- std
- 57.2
- q1
- 2
- q3
- 26.5
- iqr
- 24.5
- skew
- 3.803
- kurtosis
- 16.12
- n_outliers
- 8
- outlier_rate
- 0.09639
- zero_rate
- 0.06024
error_density
numeric feature high_skew outliersThis column represents a density or rate of errors — likely errors per unit (e.g., per line of code, per request, or per token) — given its name and bounded [0, 0.303] range. The distribution is severely right-skewed (skew=3.23, kurtosis=10.95): the median is only 0.0057 while the mean is 0.027, and 8 outliers (9.6% of rows) pull the tail toward the maximum of 0.303. With ~9.8% nulls and ~6% zeros, the bulk of observations cluster near zero but a meaningful minority exhibit substantially elevated error rates. Treatment: Log-transform (e.g., log1p) before modelling to reduce skew; impute or flag nulls separately; investigate the 8 outliers for data quality issues.
- n
- 92
- nulls
- 9 (9.8%)
- unique
- 64
- min
- 0
- max
- 0.303
- mean
- 0.02725
- median
- 0.0057
- std
- 0.05376
- q1
- 0.0025
- q3
- 0.02545
- iqr
- 0.02295
- skew
- 3.229
- kurtosis
- 10.95
- n_outliers
- 8
- outlier_rate
- 0.09639
- zero_rate
- 0.06024
notes
categorical label long_tailThis column contains free-text accessibility audit notes describing detected WCAG/accessibility violations (e.g., 'Low contrast text', 'Missing alt text', 'Missing form input label') for 92 records. The top value 'Low contrast text' appears 12 times (13% of rows), and with 38 unique values across 92 rows the entropy ratio is high at 0.89, flagging a long-tail distribution of compound violation combinations. Notably, values like 'Asia-based' and 'No data' appear alongside structured violation strings, suggesting the column conflates geographic metadata and data-availability flags with accessibility findings — an inconsistency an analyst should investigate. No nulls exist, but the semantic mixing across categories limits direct modelling use. Treatment: Parse and split compound violation strings (comma-separated) into multi-label binary indicators; handle 'Asia-based' and 'No data' as separate flags.
- n
- 92
- nulls
- 0 (0.0%)
- unique
- 38
- top_value
- Low contrast text
- top_rate
- 0.1304
- cardinality
- 38
- entropy
- 4.663
- entropy_ratio
- 0.8885