How do I compare charts with different scales?
Use a fixed order every time: title and population, unit, time span, axis range, baseline, denominator, then the visual pattern. Restating both series on one shared scale takes a couple of minutes and settles most arguments. If you cannot restate them on a common scale, you are comparing impressions rather than data.
The order matters because each step can quietly break the comparison. Two charts can come from one spreadsheet and still measure different things: monthly totals for all users in one, weekly activity for paying users in the other.
The restatement step is the one people skip. If one chart counts people and the other counts visits, convert both to visits per person. If one covers a quarter and the other a year, convert both to a monthly average. The comparison only means something after that conversion.
Write the two restated numbers on one line: 4.1 visits per person per month against 6.8, or 12 percent against 19 percent. If you cannot write that line, the comparison is not ready.
Why can two charts of the same trend look completely different?
Because the eye reads height and slope before it reads labels. A move from 48 to 52 looks flat on an axis running 0 to 100 and looks like a cliff on an axis running 45 to 55. The numbers are identical. The axis range is doing the talking, and no value in either chart is false.
This is why axis inspection comes before pattern recognition in established chart teaching, including the Datawrapper learning guides and the visualisation guide from UCLA's library. Both recommend the same first move: find the axis, note where it starts, note the unit.
Line charts may crop their axes when the detail matters, but the choice has to be stated, because a cropped line emphasises small movements. Bars may not, because a bar is read as a length from zero. A bar that starts above zero is not a smaller version of the truth, it is a different chart.
Which labels and units must match before you compare?
Four things have to match: population, unit, time span, and definition. Users against active users, a monthly average against a year to date total, gross against net, and headcount against full time equivalents are different measures wearing similar labels. Matching the headline word is not enough.
Time spans are the most common mismatch, because a year to date figure ends up beside a quarterly figure and the longer window usually wins the argument. Volumes come next: a total from a bigger group looks superior even when the rate inside that group is worse.
Definitions hide in the footnotes, which is where active and engaged get quietly redefined from one chart to the next. Currency belongs in the same check, since revenue across three years without a word about nominal figures is one number compared to a different number.
How do axis ranges and truncated baselines change the story?
A truncated baseline exaggerates every difference it shows. If the axis starts at 90 instead of 0, a small change can look like a collapse, and it will look that way to a careful reader too, because the exaggeration happens in the geometry rather than in the numbers. Start bars at zero and state the range on line charts.
Overlay charts add a second version of the problem. When two series share one axis that suits the first series, the second is drawn to a scale nobody chose for it, and its movement becomes decoration. Dual axes are worse, because a dual axis chart can be tuned until any two series appear to move together or apart.
The fix is to normalise first: index both series to a shared starting point, or convert both to a rate, then plot them on one axis. Where the comparison still is not meaningful, two small charts side by side beat one overlay chart.
When should you compare percentages instead of raw numbers?
Use percentages when the groups are different sizes, and raw numbers when the question is about scale. Raw counts mostly measure how big a group is. Twice as many complaints from the western region means little if that region is three times larger, and half as many can reverse once you divide by volume.
Percentages bring their own risk: small bases. A move from one case to two is a hundred percent increase and one extra case, and on a large base they hide absolute movement, since one percent of a million users is ten thousand people. The pairing that works is rate plus base.
How can you restate both charts on one common scale?
Pick a shared starting point and express both series relative to it, usually as 100. Then read the lines together. This is indexing, and it answers which series grew faster while setting size aside. The alternative is converting both to a standard rate, such as visits per 1,000 people per month.
Three steps do it: choose the shared start date, divide each series by its own value at that date, then multiply by 100. Indexing hides magnitude, so a doubling of a tiny number looks the same as a doubling of a large one. Keep the original values beside the restated ones.
What is the fastest check for a misleading comparison?
Look at two things: where each axis starts, and what the denominator is. If one axis is truncated, or the two charts divide by different populations, the comparison needs restating before it can support a sentence like twice as many or much faster.
That check is a reading habit you can rehearse. If you want to see what you actually absorbed from a chart or a briefing, Absorb is a comprehension practice on the App Store with short reading, listening, and visual sessions that end with questions you answer from memory.