The 44% of a truck mile that arrives as two ribbons
Showing the misleading chart
A fleet cost briefing draws the $2.336 it costs to run a Class 8 truck for one mile as a flow diagram: nine ribbons, every line item shown, all to one scale, summing exactly to the published total. Read as shapes, the biggest single band is 35% and the money looks spread across the vehicle. Driver wages and driver benefits are the top band and the bottom band, seven ribbons apart; added, they are $1.028 a mile — 44.0% of it, 2.13 times fuel, and a figure the publisher prints itself. Then a second, smaller thing: a ribbon carries its value as vertical thickness, so the width across it is that thickness times the cosine of its slope, and the slope is set by the stacking order and the shape of the box rather than by any number.
01The claim
One number, broken out nine ways, drawn as a flow diagram: $2.336, the industry-average marginal cost of running a Class 8 truck for one mile in 2025 — 3.4% above 2024’s $2.260 and the highest in the report’s history. The nine line items are the publisher’s own and they close exactly on that total. Every ribbon is drawn to one scale with its thickness set by its own line item. Nothing is indexed, annualised, smoothed, projected or rebased. No category has been merged into an “other” bucket and none has been dropped. No axis has been truncated, broken or capped. Read off the diagram, the mile is diversified: the largest single band is driver wages at 35.0%, nothing else reaches 21%, and six of the nine bands are under a tenth of the mile each. Read-out for the operations committee: that shape argues against hunting for one big saving and in favour of working the vehicle side, where the ribbons cluster — fuel, equipment payments, maintenance and tires are four of the nine bands and $1.151 of every mile, 49.3%, against 35.0% for the largest band on the chart. The priority list follows the diagram down: fuel economy, equipment cycle, maintenance intervals, insurance, tires, routing. Driver wages sit outside it — the band is large, but pay is set by a market the fleet does not control.
02The trick
A flow diagram asks the eye to rank shapes, and two of these nine shapes are the same cost. Driver wages are $0.818 a mile and driver benefits are $0.210; together, $1.028 — 44.0% of the mile, 2.13 times fuel, and more than fuel and equipment payments put together ($0.886, 37.9%). Everything that is not the driver is $1.308, or 56.0%. Both figures are printed on the briefing and neither is wrong. What is missing is a shape: the diagram draws the largest cost in trucking as its top band and its bottom band, seven ribbons apart, and nobody adds two ribbons by eye. A category drawn as two shapes reads smaller than its total whatever anyone intended by it, and the move survives every audit — nothing omitted, nothing rescaled, both labels carrying their own number. The publisher, for what it is worth, prints the $1.028 itself, as the cost to employ a driver; the briefing had it available and drew the two halves. Then the geometry takes a second bite, smaller than the first and worth doing carefully, because a Sankey is the one common chart whose marks are sheared. A ribbon encodes its value as the vertical distance between its two edges. Slope it and that vertical thickness is untouched — but the width across it, which is the dimension there actually is to look at, becomes the drawn thickness times the cosine of the slope. The ink does not change; a sheared parallelogram keeps its base times its height. What changes is the split between width and length, and the reader was asked for the width. Two things set that slope and neither is in the data. The first is where the band lands in the stacking order, which puts the steepest ribbons at the edges of the fan: here the driver’s two bands are the outermost, so they run at 26.0° at the waist — where a Sankey’s cubic edges are steepest, at exactly twice the average slope — while insurance, in the middle, runs dead level. The second is how much horizontal run the diagram was given, which is a fact about the panel it was poured into. Both effects are modest as this one is drawn, and it is worth being exact about how modest, because the waist is the narrowest point a ribbon ever reaches and quoting it as the whole ribbon would be the same trick in the other direction. Averaged along the ribbon, the driver’s bands lose 3.8% of their width and fuel loses 1.0%: driver pay is 44.0% of the money and 43.4% of the ink, fuel 20.6% and 20.9%, so the diagram offers a driver-to-fuel ratio of 2.07 where the data says 2.13. At the waist alone it is 42.2% against 21.4%, a ratio of 1.97. This is not a claim invented here. The effect is the sine illusion, which VanderPlas and Hofmann set out in the Journal of Computational and Graphical Statistics in 2015: readers do discount a sloped band, and only partly, so the direction is settled and the size is not. What makes it worth watching on a Sankey is that the knob is silent and it has a long throw. Squeeze the same nine values in the same order into half the horizontal run — a slide column, a dashboard tile, a figure resized to fill what was left over — and the driver’s waist slope goes from 26.0° to 44.3° and the loss from 10.1% to 28.5%. Nothing on the page names the ordering, and nothing on any chart has ever named the shape of its own box. Underneath both effects sits the plainest problem, and the briefing ticks a box about it in perfect innocence. “No axis truncated, broken or capped” is true. There is no axis. A Sankey prints no scale, draws no gridline and offers no baseline, so a ribbon can only be judged against its neighbours, and the numbers on this one are on labels rather than on the chart. (Both drawings are ours; the briefing, its committee and its read-out are invented, and every figure behind them is real.)
03The fix
Ask what the chart is for, because the answer decides whether a flow diagram belongs at all. Nothing here flowed into anything: this is one number broken out once, so the second stage is decoration, and the only question the committee had was which item is biggest. That question wants a sorted bar chart on one zero-based axis — a straight length per number, with no slope for the layout to charge and no neighbours to have to guess against. Combine what is one thing before you draw it. Wages and benefits are the cost of employing a driver, and drawing them as one bar with the split shown inside it does not even require our arithmetic: $1.028 is the publisher’s own figure, and 2025 is the first year it has averaged above a dollar a mile. The picture reverses on the spot — the single largest lever is not on the vehicle side at all, and it is bigger than fuel and equipment payments combined — and with it the read-out, which had put the driver outside the priority list on the strength of a band that was only ever half the cost. Where a Sankey is the right chart, and it often is, draw it and then give it the three things it is missing. It has no axis, so the scale belongs in the caption the way a unit belongs on one, and every ribbon needs its number printed on it. It has an ordering, so choose one, say what you sorted by, and hold it across every version, since nine items have 362,880 of them and the diagram looks equally finished in all of them — put anything that ought to be read together next to itself while you are there. And it has a shape, which is the part nobody thinks to declare: the same fan in a narrow column charges its outer ribbons several times what it charges them in a wide one, so if the drawing has to live in a fixed tile, say so and let the numbers carry the comparison. The line between the good use and the bad one is worth keeping sharp, because Sankeys are excellent at the thing they are for. They answer where does it go: a quantity that genuinely moves, conserved at every node, with the crossings showing the routing — crude oil into refined product into end use, water through a catchment, a cohort through four stages of a funnel. They do not answer which is biggest, because ranking wants lengths on a common baseline and a fan supplies neither. Where a chart is being asked both questions, the honest layout is usually both charts: the flow for the routing, a short sorted bar chart beside it for the ranking, and the second one is where the decision actually gets made.