---
title: "My classification produced empty buckets or weird breaks"
description: "Fix a classification that shows empty classes or odd breaks. Pick a method that fits how your values are spread, or set the breaks by hand."
seo:
  title: "Fix empty classification buckets"
sidebar:
  label: "Empty or odd class breaks"
---

You styled a layer by a numeric field, but the legend has empty classes or every feature falls in one bucket. The classification method does not fit your data. Pick a method that suits the distribution, or override the breaks per-class to fix it directly.

Classification works the same way on point, line, and polygon layers. A few things differ:

| | Point | Line | Polygon |
| --- | --- | --- | --- |
| Colour driver | `Colour markers by` | `Colour lines by` | `Fill polygons by` |
| Head/tail method | `Head/Tails` | `Head/Tail Breaks` | `Head/Tail Breaks` |
| Show size as well as colour | Yes, with a separate size field | No | No |

## Why this happens

MangoMap offers four classification methods, each generating breaks differently:

| Method | What it optimises for | Fails on |
| ------ | --------------------- | -------- |
| `Equal Interval` | Evenly-sized ranges from min to max | Skewed data — most values cluster in one or two breaks, the rest empty |
| `Quantile` | About the same number of distinct values in each class | Data with extreme outliers — outlier ends up alone or breaks lose meaning. Fields with no more distinct values than classes give a message (Symptom 5) |
| `Natural Breaks (Jenks)` | Statistical clusters in the data | Fields with fewer distinct values than classes. You get a message instead of class breaks (Symptom 5) |
| `Head/Tail Breaks` | Classes that get wider step by step, from the lowest to the highest | Very skewed data — most features can still fall in the lowest class |

Pick the method that fits your data.

For the full reference of how each method computes breaks, see [Classification methods reference](/docs/styling-layers/classification-methods-reference).

## Symptom 1: most features land in one bucket, the rest are empty

Common with `Equal Interval` on skewed data. Income, revenue, population, and other "long tail" fields often have a few very large values that stretch the equal intervals so most features fall into the lowest bucket.

For example, with `Equal Interval` and 5 classes, world country populations put almost every country polygon in the lowest class: China and India stretch the range. On a Singapore roads line layer, road length does the same: a few long roads put almost every short segment in the lowest class.

<img src="/_assets/styling-layers/concepts-choosing-classification-method--equal-interval.png" alt="World Classification Lab map coloured by population with Equal Interval: almost every country in the lowest class, and the legend" width="1600" height="960" />

**Fix**: switch to `Quantile` to spread the distinct values evenly across the classes, or `Natural Breaks (Jenks)` to find statistical clusters. In the `Style` tab, pick the method in the classification method dropdown next to `Style as`, then click `Save`. It has no label.

## Symptom 2: an outlier sits alone in its own bucket

Common with `Quantile` and `Natural Breaks` when one feature is much larger or smaller than the rest. The classification correctly identifies it as different, but the resulting legend has a useless single-feature bucket.

**Fix**:

- **Filter the outlier out** of the dataset, if it's a data-quality issue.
- **Customise the break ranges** to merge the outlier into the next-nearest bucket. Expand `Customise Class Breaks`, edit the relevant range, and click `Save`.
- **Give the outlier's class break its own colour** in `Customise Class Breaks`, so it stands out instead of hiding.

See [How to customise category and class break symbology](/docs/styling-layers/customise-category-and-class-break-symbology) for the per-break edit flow.

## Symptom 3: every feature is the same colour

Two causes. Either the field has very low variation (every value is identical or near-identical), or the classification produced a single break that catches everything.

**Fix**:

- **Confirm the field has variation** — open the dataset's data table and check.
- **Reduce the class count** if you've asked for more classes than the field has distinct values — for example, 9 classes on a field with only 5 distinct values. With `Equal Interval` or `Head/Tail Breaks`, the extra classes stay empty. Open the `Style` tab, pick a lower class count in the `Style as` dropdown, and click `Save`.
- **Style the values as categories** instead — when the field has only a few distinct values, one colour per value reads better. Set `Style as` to `Categories`. See [How to style features by category](/docs/styling-layers/style-by-category).

## Symptom 4: the breaks looked good, then I changed the field and they're broken now

Auto-generated breaks are recomputed every time the field, class count, or classification method changes. If you'd customised individual ranges, those customisations were rebuilt from scratch.

**Fix**: you cannot undo the recalculation. Either redo the per-class customisations on the new breaks, or change back to the original field and method.

See [How to customise category and class break symbology](/docs/styling-layers/customise-category-and-class-break-symbology) for the customisation controls.

## Symptom 5: a message says the field doesn't have enough values

You picked `Natural Breaks (Jenks)` or `Quantile`, and a message like this appears:

```text
Failed to generate breaks: Error: Sorry, lanes doesn't have enough values to support 9 classbreaks. Please reduce the classbreaks.
```

The field has too few distinct values for the class count. Features with the same value always share a class, so `Natural Breaks (Jenks)` can make at most one class per distinct value. `Quantile` needs more distinct values than classes. MangoMap makes no class breaks.

**Fix**:

- **Pick fewer classes** in the `Style as` dropdown. For example, a `lanes` field with the values 1 to 7 supports up to `7 classes` with `Natural Breaks (Jenks)`, and up to `6 classes` with `Quantile`.
- **Style the values as categories** — set `Style as` to `Categories` to give each value its own colour.

## When to switch to a different style approach

Some data isn't well served by numeric-field colouring at all:

- **Few unique values** — set `Style as` to `Categories`. One colour per value reads better than artificial ranges.
- **Density across an area, on a point layer** — switch `Marker type` to `Heatmap`. See [How to make a heatmap from a point layer](/docs/styling-layers/make-a-heatmap).
- **Magnitude as size, on a point layer** — size the markers by the field. See [How to size points by a number](/docs/styling-layers/size-points-by-a-number).

Line and polygon layers show a numeric field by colour only. Line width is the same for the whole layer.

## Related troubleshooting

- [My style change isn't appearing on the map](/docs/styling-layers/troubleshoot-style-not-applying)
- [My colour ramp is the wrong way round](/docs/styling-layers/troubleshoot-colour-ramp-reversed)