**The most counterintuitive finding is that entrances exceed page views, which should be impossible.** An entrance is a page view that starts a session, so entrances can never exceed page views for the same page. In this data they often do. - **Scale:** 24,648 of 62,496 page-rows (about 39%) have more entrances than page views. This comes from the per-site counts in the step-1 and step-2 results. - **Where it happens:** Every site except checkbook.brla.gov has such rows. For example, www.brla.gov has 11,830 of its 24,597 rows, and brgov.com has 9,211 of 17,595. - **Worst case:** gis.brla.gov logged 5,219,822 entrances against only 897,623 page views, a ratio of 5.82 entrances per view. my.brla.gov (1.36), budget.brla.gov (1.36) and 311.brla.gov (0.96) are also at or above 1 in total, and for all three some individual rows are above 1. - **Other sites:** Most other sites have ratios between 0.19 and 0.87 in total, but many individual rows still exceed 1. This suggests the metrics come from different counting rules, such as sessions versus hits, filtered versus unfiltered, or mismatched periods. They are not one consistent funnel. Treat entrance-based rankings, such as which entry pages to optimise, with caution until the definitions are reconciled. **Related oddities that point to the same data-quality concern:** - **Traffic concentration:** city.brla.gov has just 404 rows but about 720.8M page views, roughly 72% of the dataset's total of about 994M. Any cross-site comparison is dominated by it. - **Suspicious round number:** 311.brla.gov shows exactly 700,000 total page views across 35 rows, which looks like a cap or a placeholder rather than an observed count. - **Constant values:** checkbook.brla.gov's three rows all have bounce and exit rates of exactly 1.0, with no variation. - **Bounce and exit rates differ by site:** Several sites have maximums of 100, but gisdata.brla.gov's maximums are 2 (bounce) and 1 (exit). That suggests inconsistent units or scales (for example fractions versus percentages) between sites. These are all based on the per-site aggregates in the step-0 to step-2 results. I did not trace individual rows to find the cause of the entrance anomaly.