Use a log scale when your data spans several orders of magnitude or when you care about relative (percentage) change rather than absolute change. A dataset that ranges from 10 to 10,000,000, for example, becomes nearly unreadable on a standard linear axis because the smaller values get crushed into a flat line at the bottom. A logarithmic scale spaces each tenfold increase evenly, giving every part of the range room to breathe.
What a Log Scale Actually Does
On a linear scale, the distance from 0 to 100 is the same as the distance from 100 to 200. Each step represents the same absolute amount. On a log scale, the distance from 1 to 10 is the same as the distance from 10 to 100, and the same as 100 to 1,000. Each step represents the same multiplicative factor, typically a power of 10.
This single change in spacing transforms what you can see in a chart. Patterns that are invisible on a linear axis, like a small dataset growing at 5% per month, suddenly show up as a clean, straight line. Differences between large values that look identical on a linear axis become distinguishable. And data points near zero, which would normally be stacked on top of each other, spread out enough to actually read.
Your Data Spans a Wide Range
The most common reason to reach for a log scale is when your values cover multiple orders of magnitude. If you’re plotting country populations, you might have Iceland at around 370,000 and China at 1.4 billion. On a linear axis, Iceland and dozens of other small countries would be indistinguishable dots at the zero line, while only the top few nations use the rest of the chart. A log scale gives every country a meaningful position.
This comes up constantly in science and engineering: earthquake magnitudes, sound intensity (decibels), acidity (pH), star brightness, and particle sizes all naturally span ranges so wide that linear scales are useless. Income distributions, city populations, website traffic, and disk storage costs over time all share the same property. If you find yourself thinking “most of my data is jammed into one corner of the chart,” that’s the signal.
You Care About Rates of Change
A log scale turns constant percentage growth into a straight line. This makes it the right choice whenever the interesting question is “how fast is this growing?” rather than “how much was added?”
Consider two companies. Company A grows revenue from $1 million to $2 million in a year. Company B grows from $100 million to $150 million. On a linear scale, Company B’s growth looks far more impressive because the absolute dollar amount is larger. On a log scale, Company A’s 100% growth shows a steeper slope than Company B’s 50% growth, which more accurately reflects the growth rate that investors and analysts actually care about.
This is why log scales became so widely discussed during the COVID-19 pandemic. Epidemiologists used them to compare whether the rate of infection was accelerating, decelerating, or holding steady across countries with very different population sizes. A straight line on a log-scale chart of cases meant exponential growth at a constant rate. A curve bending downward meant the rate was slowing. Those distinctions were nearly impossible to see on a linear scale once case counts diverged by orders of magnitude between countries.
Comparing Things of Very Different Sizes
When you need to display multiple datasets on the same chart and they operate at different scales, a log axis keeps them all readable. Stock prices are a good example: if you plot a $5 stock and a $500 stock on the same linear chart, the cheap stock’s movements vanish. On a log scale, a 10% move looks the same size regardless of the starting price, so you can visually compare performance across assets.
The same logic applies to comparing metrics across categories. If you’re charting average income by profession and the range goes from $25,000 to $5,000,000, a log scale prevents the vast majority of professions from being squished together at the bottom while a few outliers dominate the visual.
Your Data Is Exponential or Follows a Power Law
Many natural and human-made phenomena follow exponential or power-law distributions. Bacterial growth, radioactive decay, compound interest, and Moore’s Law are all exponential. Word frequency in language, earthquake frequency by magnitude, and city size distributions follow power laws. Plotting these on a log scale (or a log-log scale for power laws) transforms the curve into a straight line, which is far easier for the human eye to interpret and compare.
If you suspect your data follows one of these patterns, switching to a log scale is both a diagnostic tool and a presentation tool. A straight line on a log-scaled y-axis confirms exponential behavior. A straight line when both axes are log-scaled confirms a power-law relationship. The slope of that line directly tells you the exponent, which is the rate of growth or the steepness of the distribution.
When Not to Use a Log Scale
Log scales are the wrong choice in several common situations. If your data doesn’t span a wide range, a log scale adds complexity without benefit. Plotting monthly temperatures from 40°F to 90°F on a log scale would distort the visual and confuse readers for no reason.
Data that includes zero or negative values can’t be plotted on a standard log scale at all, since the logarithm of zero is undefined and negative numbers have no real logarithm. You’d need to transform the data first (adding a constant, for instance), which introduces its own complications and can mislead viewers.
Audience matters too. Log scales require more effort to read correctly. On a linear scale, most people can glance at a bar chart and estimate that one bar is about twice as tall as another. On a log scale, bars that look similar in height might represent values that differ by a factor of ten. If your audience isn’t comfortable with logarithmic thinking, and most general audiences are not, a log scale can create more confusion than clarity. In those cases, consider alternatives like breaking the data into smaller subsets, using multiple panels, or calling out specific values with annotations on a linear chart.
Choosing Between Log and Linear in Practice
Start with a simple question: what do you want the reader to compare? If the answer is absolute differences (“how many more units did we sell?”), use a linear scale. If the answer is relative differences (“which product line is growing faster?”), use a log scale.
A few practical guidelines help:
- Range test: If the ratio between your largest and smallest values is greater than about 100 to 1, a log scale is almost certainly better.
- Growth comparison: If you’re comparing growth rates across entities of different sizes, log scales make the comparison honest.
- Skewed distributions: If a histogram is piled up on the left with a long right tail (income, city populations, gene expression levels), try logging the x-axis to reveal the structure.
- Time series with compounding: Any metric that compounds over time, like investment returns, GDP, or user growth, is more informatively displayed on a log scale for periods longer than a few years.
One underused option is labeling the log axis with the original values (1, 10, 100, 1,000) rather than the exponents (0, 1, 2, 3). This gives you the spacing benefits of a log scale while keeping the numbers intuitive for readers who aren’t thinking in terms of logarithms. Most charting tools support this by default.