Four cases in a thousand, drawn a third as long as the 782
Showing the misleading chart
A sales slide we drew for an invented trial-prep software firm shows how every federal civil case that ended in the year to September 2025 ended: 407,379 cases, four outcomes, a zero-based axis and a label on every bar. It says trial is not the exception. A zigzag in the axis deletes everything from 3,000 to 20,000 cases, and the 1,681 cases that reached trial come out a third as long as the 318,461 where the court acted but the case ended before pretrial. Counted, they are one 189th of that bar, and 0.4% of all the cases that ended.
01The claim
Trial is not the exception: it is a real slice of the federal civil docket, so a litigation team should budget for it. That is the pitch on a sales slide of the kind a trial-preparation software company might show a law firm, which we drew from the federal judiciary’s own statistics for an invented vendor. It takes every civil case that ended in the US district courts in the twelve months to 30 September 2025, all 407,379 of them, and sorts them by the stage they ended at, as the judiciary’s Table C-4 does: 60,674 that ended with no court action, 318,461 where the court acted but the case was over before pretrial (the table has headed that column “Before Trial” since the 2017 edition; earlier editions say “Before Pretrial”), 26,563 that ended during or after pretrial but before trial, and 1,681 that ended during or after trial. Four bars, one axis starting at zero, a number on every bar. The trial bar is gold, and it is about a third as long as the longest bar and half as long as the other two.
02The trick
No label on the slide is wrong, and the axis does begin at zero. What it does not do is continue. The lower scale runs 0 to 3,000; then come two zigzags; then ticks at 100,000, 200,000 and 300,000. Step those ticks back to the zigzag and the upper scale turns out to begin at 20,000, a value printed nowhere, so 17,000 cases of scale have been removed — 5.7 times the whole of the scale shown beneath the break. (Read the nearest printed ticks, 3,000 and 100,000, and you would guess 97,000; the guess overstates the gap but points the same way.) Nor do the two sides share a rate. Left of the zigzag, a thousand cases buy 286 pixels of the published image; right of it, 2.1. The trial bar is the only one that lives entirely on the generous side, and every case the other three carry past the break is drawn at a 135th of its rate. Put a ruler on the image and the consequence is plain. Trial measures 481 pixels and the before-pretrial bar 1,536, a drawn ratio of 1 to 3.2 for two counts whose real ratio is 1 to 189. Against no court action the drawing gives 1 to 2.1 where the counts give 1 to 36, and against during-or-after-pretrial 1 to 1.9 where they give 1 to 16. That matters more than usual here, because this is a part-to-whole question. Nobody looking at the slide reads 1,681 off its label and divides; they judge how much of the picture is gold, and the position of the break set that share.
03The fix
Draw all four outcomes on one zero-based axis and the gold bar shrinks to a sliver: on the redraw’s 0-to-350,000 panel it is 7 pixels wide, against 1,274 for the before-pretrial bar. The sliver is the result. Per thousand federal civil cases that ended that year, about 149 ended with no court action, 782 with court action before pretrial, 65 during or after pretrial and 4 during or after trial — roughly one in 242, the 0.4% the table itself prints in its “percent reaching trial” column. For a share like this, a grid of a thousand dots beats any bar: there is no axis to interrupt, and anyone can count four gold dots. Two honest cautions belong beside it. First, one line dominates the year: diversity cases filed as “other personal injury” ended 162,125 cases, only 115 of them during or after trial. Leave that line out and 1,566 of the remaining 245,254 cases ended at trial, 0.64%, about one in 157 — higher, still well under one in a hundred, so the finding does not rest on it. Second, the column counts cases that ended during or after trial, not the number of trials held: one trial can close several consolidated cases, and a trial still running at year end waits until its case ends. The slide’s “1,681 federal civil trials” treats the two as the same thing, which a careful version would not. As for spotting a broken axis in the wild, establish where each side of the zigzag really stops and starts, stepping unlabelled ticks back to the break where you have to. Here that is 3,000 and 20,000. If the gap swallows more scale than is drawn below it, as 17,000 against 3,000 does, the bars have become decoration, and the printed numbers are the chart.