# Data-quality review of `raw.Website_Analytics` (62,496 rows, 11 websites, 2014–2024) The data is structurally clean, but it has one large logical inconsistency (entrances exceeding page views) and one extreme outlier site. ## Main issues **1. Entrances greater than page views (impossible, widespread).** - 24,648 rows (about 39%) report more `ENTRANCES` than `PAGE VIEWS`. Every entrance is a page view, so this should never happen. - It is concentrated in a few sites (step-1): - `www.brla.gov`: 11,830 of 24,597 rows. - `brgov.com`: 9,211 of 17,595 rows. - `data.brla.gov`: 2,958 of 14,379 rows. - `city.brla.gov`: 243 of 404 rows. - This points to a systematic cause, such as how entrances are attributed or how rows were exported or aggregated, rather than random errors. Treat entrance-based conclusions with caution, such as which pages to optimise as entry points. **2. Extreme outliers in volume.** - The median page views per row is 8, the 99th percentile is about 64,542, and the maximum is 228,326,194 (step-0). - That maximum belongs to `city.brla.gov`, which has only 404 rows. Its maximum entrances value is 227,064,076, so the page views and entrances are consistent with each other but wildly out of scale. - This is most likely a root or aggregate row, or a tracking artefact, and it would dominate any sum or average. - `www.brla.gov` also has a very large maximum (17.6M). `my.brla.gov` reaches 2.36M despite having only 191 rows. **3. Duplicate keys.** - 21 website/year/URL keys appear twice (42 rows, at most 2 repeats per key), which breaks the stated grain (step-2). The count is small, but it will slightly double-count those pages. **4. Zero values.** - 7,319 rows (about 12%) have an average time on page of 0 seconds. This is plausibly a real result for exit or bounce pages, but it is also a typical "not measured" placeholder, so it can bias averages downward. - 6 rows have zero page views: 4 on `budget.brla.gov` and 2 on `my.brla.gov` (step-1). These are rows with no activity. **5. Minor format inconsistencies.** - `www.brla.gov` has 2 rows flagged for whitespace issues and 1 flagged as having a non-http URL (step-1). This is tiny, but it can split or drop entities in joins and grouping. ## Checks that came back clean (step-0) - No negative values anywhere. - Unique views never exceed page views. - Bounce rate and exit rate are never outside the valid range. - There are no nulls in page path, page URL, bounce rate or exit rate. - Average time on page tops out at 2,794 seconds (about 47 minutes), so there is no sign of mixed units (for example, milliseconds mixed with seconds). ## Recommendations - Exclude or separately flag `city.brla.gov` extremes before summing. - Deduplicate the 21 repeated keys. - Use medians or trimmed stats for volume metrics. - Do not rely on entrances without first clarifying why they exceed page views.