Common Mistakes When Choosing Graph Scales

You know that moment when you look at a chart and something just feels… off? The line looks like it’s about to launch into orbit, or the bars make a tiny difference look like a chasm? I’ve been staring at graphs for over a decade, and I can tell you—nine times out of ten, the culprit isn’t bad data. It’s bad scale choices. Seriously, picking the right graph scales is one of those unsung skills that separates clear communication from accidental manipulation. And yet, people mess it up constantly. Let’s walk through the trainwrecks I see most often.

The Zero Baseline Fallacy: When Not Starting at Zero Is a Crime (and When It Isn’t)

This is the heavyweight champion of common mistakes when choosing graph scales—the truncated y-axis. Look, I get it. If your sales numbers went from 98.2% to 98.9%, starting your axis at zero makes that tiny bump look like a flat line. It’s boring. So you zoom in. You start the y-axis at 98%. Suddenly, that 0.7% gain looks huge.

Here’s the thing: depending on your audience and message, that might be a lie. Common mistakes when choosing graph scales often involve forgetting that our brains interpret bar heights and line slopes relative to the entire visible range. When you cut off the baseline, you exaggerate the perceived difference. It’s a big deal in political polling or financial reporting.

But—and this is where I see the pendulum swing too far—starting at zero isn’t always mandatory. If you’re showing temperature fluctuations between 68 and 72 degrees, starting at zero just squishes your data into a tiny band. Nobody cares about the absolute zero of temperature here. The trick is knowing your audience. Are they sophisticated analysts who expect a zoomed-in view? Or is this a public-facing chart where zero baseline is the default expectation? Honestly, most people default to bar charts needing a zero baseline, but line charts showing change over time get more leeway. Don’t be the person who blindly applies one rule to every chart.

The “Bars Must Start at Zero” Myth

Let’s get specific. A bar chart’s visual power comes from the bar length. If you chop off the bottom, you break the visual comparison. Period. It’s one of the most egregious common mistakes when choosing graph scales because the viewer’s eye can’t un-see the proportional difference.

I once saw a presentation where a company showed revenue growth over five years. They started the y-axis at $10 million instead of zero. The first year was $10.5 million, and the fifth year was $11 million. The bars looked nearly identical in height even though the actual data barely moved. The presenter was trying to imply steady, strong growth. It was nonsense. The audience, to their credit, called it out.

If you’re using bars to compare values, the baseline must be zero. It’s not a suggestion. It’s a rule of visual integrity. Otherwise, just use a dot plot or a line chart. Don’t trick people with their own eyeballs.

Line Charts and the “Zoomed In” Justification

Line charts are different beasts. Here, common mistakes when choosing graph scales usually involve over-zooming. Starting a line chart at zero for stock prices over a month is a waste of space. The line will be a flat, boring sliver.

But the mistake is going too far the other way. I’ve seen line charts where the y-axis range is so tight that normal volatility looks like a heart attack on paper. The scale should amplify meaningful variation, not noise. A good rule of thumb? Ensure the data occupies about two-thirds of the vertical space. That gives you context without distortion. It’s an art, not a science, but it’s an art you need to practice.

The Logarithmic Scale Trap: When You Should (and Shouldn’t) Use Log Scales

Log scales are a power tool. Like any power tool, they can build a beautiful deck or take off your finger. The biggest common mistakes when choosing graph scales with logs is using them to hide a story, or worse, using them when you don’t understand them yourself.

A log scale compresses the visual distance between large numbers. A jump from 10 to 100 looks the same as a jump from 100 to 1000. This is perfect for data that spans multiple orders of magnitude—think bacteria growth, earthquake magnitudes, or income distribution. But if your data only ranges from 100 to 200, a log scale makes a 50% increase look identical to a 5% increase. That’s a disaster.

I once reviewed a report on website traffic growth. The author used a log scale on the y-axis. Traffic went from 1,000 visitors to 10,000 visitors—a huge win. But on the graph, the line looked like a gentle slope. The CEO saw it and said, “Is that all?” The author was trying to be “sophisticated” but completely killed the impact of the data. Common mistakes when choosing graph scales often stem from trying to look smart instead of trying to be clear.

The “Look, I’m a Data Scientist” Log Scale

You see this one on LinkedIn all the time. Someone posts a chart using a log scale for data that has a perfectly reasonable linear range. Why? Because it makes the line look “exponential” or “trendy.” It’s a vanity move.

Here’s a simple test: if your data’s highest value is less than 10 times its lowest value, a log scale is probably overkill. If you’re comparing rates of change (percentages), not absolute values, a log scale is your friend. If you’re comparing counts or totals, stick with linear. Don’t fall for the trap of thinking log is “more advanced.” It’s just a different tool.

When a Log Scale Is the Only Honest Choice

On the flip side, failing to use a log scale when it’s appropriate is another of those common mistakes when choosing graph scales. Imagine plotting global COVID-19 case counts in early 2020. The first few weeks had dozens of cases. Then it went to thousands, then millions. A linear scale would make the early data invisible—a flat line hugging the bottom axis.

In that case, a log scale is the only way to see the full story of the pandemic’s early spread and eventual plateau. It reveals the exponential growth phase. Without it, you’re just showing a hockey stick that tells you nothing. The key is to label the axis clearly. Don’t assume your audience knows you’re using a log scale. Tell them. Explicitly. “Logarithmic scale” in the axis label isn’t showing off; it’s being honest.

The Inverted and Non-Uniform Scale Catastrophe

Let’s get into the weeds. This section covers mistakes that are subtler but just as damaging. We’re talking about common mistakes when choosing graph scales that mess with our perception of direction and proportion.

I once saw a chart of customer satisfaction scores. The y-axis ran from 100 at the bottom to 0 at the top. Higher satisfaction meant a lower line on the chart. The presenter said, “As you can see, satisfaction is trending down.” The audience nodded. But the line was physically descending because of the inverted axis. It was visual brain candy that actually reversed the truth.

Inverting an axis can be useful for depth or elevation data, or for negative metrics like error rates. But for standard positive metrics? It’s a no-go. Unless you have a very specific reason (and a very clear label), keep your low numbers at the bottom and high numbers at the top. Our brains are wired that way.

The Broken Axis Trick (and Why It Backfires)

A broken axis—where you use a zigzag line or a break symbol to skip over a range of values—is a classic trick. It’s used to show small variations in large numbers. And honestly? It almost always looks like you’re hiding something.

I’m not saying never use it. If you have data with a massive outlier, a broken axis can let you show the bulk of the data without squishing everything into a line. But the moment you use it, you invite scrutiny. People will wonder what you’re hiding. The common mistakes when choosing graph scales here is using breaks for convenience rather than necessity. Ask yourself: can I show this data better with a log scale? Can I split it into two charts? If the answer is no, and you must use a break, label it clearly. And be prepared for someone to call you out.

The “Looks Like a Spiderweb” Non-Linear Scale Problem

Some chart types, like radar charts or polar plots, have their own scale quirks. Common mistakes when choosing graph scales in these formats usually involve using inconsistent intervals between the concentric rings.

Imagine a radar chart where the first ring is 0, the second is 50, the third is 100, but the fourth is 500. The visual area of the shape gets distorted. A value of 400 might look half as significant as it actually is because the scale is non-uniform. If you must use a polar or radar chart, ensure the axis intervals are consistent. Otherwise, you’re just drawing an abstract blob.

Overcrowding and Misleading Axis Intervals

This is the final boss of common mistakes when choosing graph scales. You have your data. You picked your zero or non-zero baseline. You chose linear or log. Now you just need the ticks.

Ever seen a chart with a y-axis that has labels at 0, 10, 20, 30, 40, and then a sudden jump to 100? That’s a rookie move. The uneven intervals make the slope of the line change visually. It’s like changing the ruler mid-measurement. Your audience will feel something is wrong, even if they can’t articulate it.

- Too many ticks: The axis becomes a dense grid of numbers. Nobody reads that. - Too few ticks: The viewer has to guess the value of every data point. - Odd intervals: Avoid 3, 7, 13. Stick to multiples of 5 or 10 for the average audience. It’s cleaner.

The gold standard? Use intervals that make the gridlines useful for quick estimation. If your highest value is 47, label at 0, 10, 20, 30, 40, and 50. Not 0, 5, 10, 15… It’s about balance.

The “Automatic Default” Trap

Software defaults are your enemy. Excel, Google Sheets, and most visualization tools will pick intervals algorithmically. And they often pick wrong. They might choose intervals that emphasize random aspects of your data or create leading gridlines that distract.

One of the most common common mistakes when choosing graph scales is simply accepting the default. Don’t. Manually set your axis minimum and maximum. Adjust the major and minor unit. It takes twenty seconds and will instantly make your chart look custom and professional. Seriously, do this for every single graph you make.

Gridlines That Lie

Gridlines are supposed to help the eye track values. If your gridlines are too thick, they compete with the data. If they are too sparse, they’re useless. But the worst offense is having graph scales that use gridlines that don’t align with your data points.

If your data is at values 12, 24, and 36, but your gridlines are at 0, 20, 40, the lines are nearly useless for reading those points. You’ll have to squint and interpolate. Adjust your graph scales so that major gridlines fall near where the data sits, or use minor gridlines. Your audience’s eyes will thank you.