telemetry that motivates.
The data is the light and the interface is the dark room: charts that prove the change is real.
the data is the light.
A chart can flatter or it can tell the truth, and most tools quietly choose flattery. The job of a Trainbase chart is the opposite: spend every pixel of ink on the data, root every scale in reality, and let the real change carry the screen. The interface is the dark room. The data is the only thing meant to glow.
Edward Tufte called the worst of it the lie factor: the ratio of the change a chart shows to the change in the numbers. A good chart holds that ratio at one. Trainbase holds it there by default, then makes the honest change impossible to miss.
the same data, two stories.
Here is one week of weigh-ins, drawn twice from the identical seven readings. The only thing that changes is the vertical scale. One panel auto-zooms to the data's own range and turns a 0.4kg wobble into a cliff, complete with real-looking axis numbers. The other starts at zero, where the same week reads as the gentle drift it is. Both are the truth about the numbers. Only one is the truth about the body.
every sin, and its honest practice.
The ways a chart lies are well catalogued, and so are the fixes. These are the four that matter most for body and training data, where a coach is reading a trend they will act on. The honest column is the one Trainbase builds to.
| the sin | the honest practice | |
|---|---|---|
| the baseline | truncated y-axis that starts at the data's min, inflating small moves | a zero-true or genuinely meaningful baseline, so a small move reads small |
| the scale across series | per-series autoscale, so a flat line and a steep one look identical | one shared scale, so a flat series reads flat and the gap reads true |
| the ink | chartjunk: gradients, 3d, shadows, gridlines louder than the line | data-ink: erase everything that is not the data, keep the rest quiet |
| the axes | dual y-axes, where any correlation can be manufactured at will | one axis, one unit; if two things share a chart they share a scale |
The sins are Tufte's lie factor and chartjunk, plus the dual-axis trap every spreadsheet makes one click away. The practice is what a chart looks like when nothing on it is trying to win an argument.
a chart done right.
This is the same body-composition arc the product is built around, drawn the honest way. Three series on one shared scale: lean climbing, fat falling, bodyweight nearly flat between them. A dashed reference line marks the client's lean-mass goal. Nothing here is exaggerated, and nothing true is hidden.
what most improves a chart's honesty.
Not every fix moves the needle equally. Ranked by how much each one raises the chance a reader walks away with the true picture, this is where the leverage is. The first two are nearly the whole game.
Weights are a practitioner's ranking of impact on truthful reading, not a measured statistic. The order is the point: scale and baseline decide whether a chart lies before any styling matters at all.
the honest answers.
Is starting every axis at zero always the right call?
No, and Trainbase does not pretend it is. Zero is right when the quantity's floor is meaningful, like mass or count. For something like body temperature, a zero baseline would waste the whole chart on empty space. The real rule is narrower: never truncate an axis to inflate a change, and never pick a baseline because it makes the line look more dramatic.
Why put a flat line on the same scale as a steep one?
Because that is the truth. If bodyweight barely moves while lean and fat trade places, the only way to show that honestly is one shared scale where the flat line reads flat. Auto-scaling each series to its own range would make a 1kg drift and a 10kg swing look identical, which is exactly the lie this product exists to refuse.
What is wrong with a second y-axis?
A dual axis lets you slide two unrelated scales until any two lines appear to move together. The correlation is manufactured by the chart, not found in the data. When two series genuinely belong together, they belong on one scale; when they do not, they belong on two charts.
How does this handle a noisy reading?
By reading the trend, not the point. Body-fat estimates and daily weigh-ins are noisy, and every estimate is treated as an estimate. The chart is built so one bad reading is a small bump on a long line, never a verdict, and the honest scale keeps that bump the size it actually is.