In statistics, the mode is the value that occurs most often in a dataset, or, for a continuous distribution, the value at which the probability density reaches its peak. Also called the statistical mode, it is one of the three classic measures of central tendency, alongside the mean and the median.
The mode is the only measure of central tendency that works for nominal data, such as colors, product categories or answers to a multiple-choice question, because it requires counting but no arithmetic or ordering. A dataset can have one mode, several modes or no meaningful mode at all, which makes the mode both informative and easy to misread.
How the mode is determined
For discrete data, finding the mode means building a frequency table and selecting the value or category with the highest count. In the list 2, 3, 3, 5, 7, 7, 7, 9 the mode is 7, because it appears three times while no other value appears more than twice.
For continuous data, exact repeats are rare: measured heights of 1.7312 and 1.7314 meters are different values. The mode is therefore estimated from grouped data or a smooth density estimate. With a histogram, the class with the highest frequency is the modal class, and a common interpolation formula places the mode inside it, closer to the side of the more frequent neighboring class.
| Data type | Method | Result |
|---|---|---|
| Nominal categories | Count each category | Most frequent category |
| Discrete numbers | Frequency table of values | Most frequent value or values |
| Grouped continuous data | Modal class plus interpolation | Estimated mode |
| Continuous sample | Kernel density estimate | Location of the density peak |
| Theoretical distribution | Maximize the density function | Exact mode |
With kernel density estimation, the estimated mode depends on the bandwidth: a narrow bandwidth produces many small spurious peaks, while a wide bandwidth can merge two real peaks into one. The mode of continuous data is therefore an estimate with real uncertainty, not a counted fact.
Unimodal, bimodal and multimodal distributions
The number of modes is itself a key description of a distribution’s shape. Statisticians use the term mode loosely for any clear local peak, not only the single highest one, so a distribution with two distinct peaks is called bimodal even if one peak is lower.
| Shape | Number of peaks | Typical interpretation |
|---|---|---|
| Unimodal | One | One homogeneous population |
| Bimodal | Two | Often two mixed subgroups |
| Multimodal | Three or more | Several subgroups or rounding effects |
| Uniform | None meaningful | All values roughly equally likely |
A bimodal distribution is an important diagnostic signal. Customer order values that peak at 20 and at 200 euros usually reflect two different customer segments, and summarizing them with a single mean of about 110 euros would describe almost nobody. Detecting such peaks is often the first step toward splitting an analysis by group.
Mode compared with mean and median
The mode, the mean and the median answer different questions: the most typical value, the balance point and the middle value. Choosing among them depends on the measurement level and the shape of the data.
| Property | Mode | Median | Mean |
|---|---|---|---|
| Works for nominal data | Yes | No | No |
| Works for ordinal data | Yes | Yes | Not meaningful |
| Sensitive to outliers | No | Barely | Strongly |
| Always unique | No | Yes | Yes |
| Uses every value | No | Only their order | Yes |
In a symmetric unimodal distribution, such as the normal distribution, the mode, median and mean coincide. In a right-skewed distribution, such as household income, the usual order is mode below median below mean, because a long tail of high values pulls the mean upward. This ordering is a rule of thumb, not a theorem, and it can fail for discrete or multimodal data.
For moderately skewed unimodal distributions, Pearson’s empirical relationship states that the distance between the mean and the mode is roughly three times the distance between the mean and the median. It gives a quick estimate of the mode from the other two measures, but it is an approximation for well-behaved data and should not replace an actual frequency count.
When the mode is the right summary
The mode is the natural choice when the question is about the most common case rather than about totals or balance. Typical situations include the following.
- Categorical survey answers, such as the most chosen reason for canceling a subscription.
- Inventory and sizing decisions, where the most frequently sold shoe or clothing size matters more than an average size that may not exist.
- Discrete counts with few distinct values, such as the most common number of people per household.
- Shape analysis, where the location and number of peaks reveal subgroups or measurement artifacts.
- Classification, where majority voting among classifiers or neighbors selects the modal class.
In software, many spreadsheet functions such as MODE return only the first mode found when several values tie, while functions such as MODE.MULT and statistical libraries can return all of them. This difference silently changes results for multimodal data.
Limitations and common misreadings
The mode is unstable in small samples: adding a single observation can move it to a completely different value. In a sample of 20 values where two values each appear three times, the mode is effectively a coin toss, so reporting it without its frequency is misleading.
Another frequent mistake is reporting a mode for continuous measurements without grouping, where every value appears once and any reported mode is meaningless. The mode is also unsuitable for further arithmetic: unlike the mean, it cannot be combined across groups, because the mode of a merged dataset is not derived from the modes of its parts. Reporting the mode together with the median, the mean and a histogram gives the most honest picture.
Explained without jargon
The mode is simply the most popular answer. If you ask thirty people for their favorite fruit and twelve say apple, apple is the mode.
It works even when the answers are words instead of numbers, which the ordinary average cannot handle. Sometimes two answers tie for first place, and then there are two modes, which often means the group is really made up of two different kinds of people.
A typical use case
A sports shoe retailer reviews one season of sales for a running model: 4,800 pairs across sizes 36 to 47. The mean size is about 41.6, a size that does not exist, and the median is 42. The frequency table, however, shows two peaks, at size 39 and at size 43, with noticeably fewer sales of size 41.
The buying team realizes that the model sells to two groups with different size profiles. For the next season it orders stock around both modes instead of centering the order on size 42. Stockouts in the most requested sizes fall, and fewer unsold pairs in the middle sizes have to be discounted at the end of the season.