# Individual railroad weekly measures The individual explorers for BNSF, Union Pacific, Norfolk Southern, CPKC and CN contain four STB-reported measures for eight weeks ending July 24–September 11, 2026. CSX separately retains its reviewed history since March 2017, including source-review gaps. The additional values come from the same 48 original weekly workbooks used for the speed review, retrieved September 20, 2026. Each downloaded observation includes its official URL, source worksheet/cells, retrieval time and SHA-256 hash. ## Reading the measures Train speed is an operating measure, not end-to-end delivery speed. System terminal dwell is reported separately; it is never calculated by averaging the ten terminal values. Cars online is a weekly average inventory, not carloads delivered or a service-quality score. Trains held is an average number per day, not a weekly count of unique delayed trains or hours of delay. One selected measure uses original units. Two or more selected measures use a starting-week index of 100. Weekly change is the difference from the preceding reported week; percent change divides that difference by the preceding value. Zero or missing baselines leave the applicable index/percentage unavailable. Displayed values are rounded while calculations retain source precision. Clicking or tapping a week opens its details without changing the chart range. A measure with no values in the loaded coverage is grayed out and cannot be selected. A missing week within an available series remains a gap. ## Source differences retained The eight-week audit checked 2,464 compilation cells, 664 component/total relationships and independently read 144 headline values from workbook XML. All 48 workbook hashes were verified. The current six STB methodology PDFs and compilation match the preserved copies. Some August 19 compilation cells contain one-decimal values where original weekly workbooks retain more precision. The original values and their notes are preserved. CN's train speed for the week ending August 14 is 22.49 mph in the weekly source and 22.5 in the compilation; neither is silently substituted for the other. NS's car-type sums are 1–4 cars below its reported totals in all eight weeks. Some NS held rows, cause totals and train-type totals differ by one from their corresponding reported totals. The reason is unconfirmed. Its headline charts use the reported totals, and its held bars show overall totals without stacking causes. Separate cause values and original train-type tables remain available with a visible explanation. Figures are not rebalanced. For the latest week, NS's website reports 162,045 cars while both STB sources report 162,044; this site retains the STB value and notes the difference. ## Railroad-specific definitions - [BNSF](https://www.stb.gov/wp-content/uploads/BNSF_Methodology.pdf): dwell excludes run-through, stored, bad-order and maintenance-of-way cars. Cars online uses the AAR/Railinc inventory framework. Held snapshots can include planned customer/connection holds; one cause is assigned manually. - [Union Pacific](https://www.stb.gov/wp-content/uploads/UP-Methodology-EP-724-03.01.22.pdf): dwell has specified arrival/release/interchange start events and departure/placement/interchange end events. Inventory includes stored cars and cars on shortlines/customer facilities. Held snapshots exclude yard/local trains; clock/threshold are not specified in the reviewed document. - [Norfolk Southern](https://www.stb.gov/wp-content/uploads/NS_Methodology.pdf): dwell includes additional exclusions and ambiguous formula wording; the reported hours are preserved without attempting a correction. Cars online uses Railinc inventory. Held observations occur at 6 a.m., using a four-hour in-route threshold plus an origin-terminal process. Non-additive totals are explained above. - [CPKC](https://www.stb.gov/wp-content/uploads/CPKC-Methodology.pdf): dwell is processed dwell for available U.S. yards; cars online uses U.S.-network snapshots rounded to whole cars. Held trains meet a six-hour single-location threshold, with the longest-duration cause assigned. Legacy CP/KCS records are not spliced into these series. - [CN](https://www.stb.gov/wp-content/uploads/CN_Methodology.pdf): speed/dwell retain explicit U.S.-system labels. Cars online counts active cars, excluding stored/heavy-bad-order equipment. Held trains exceed their scheduled time at the snapshot location by one hour and are counted at 00:01 Eastern. Cars/holds retain the scope of CN's STB report without asserting entire North American coverage. These are individual reported trends. They do not establish equivalent populations, clocks or collection methods across railroads. The current comparison page remains limited to its reviewed train-speed groups. Longer histories and additional comparison features require separate implementation and validation. ## Downloads [BNSF JSON](https://www.railstats.org/data/weekly-service-bnsf.json) · [Union Pacific JSON](https://www.railstats.org/data/weekly-service-up.json) · [Norfolk Southern JSON](https://www.railstats.org/data/weekly-service-ns.json) · [CPKC JSON](https://www.railstats.org/data/weekly-service-cpkc.json) · [CN JSON](https://www.railstats.org/data/weekly-service-cn.json). All original reports are linked from the [STB Rail Service Data page](https://www.stb.gov/reports-data/rail-service-data/). No smoothing, interpolation, inferred zero, adjustment factor or composite score is applied.