How to Read a Heatmap in Employee Survey Results

A survey heatmap is a table that shows employee survey results with categories or items in the rows and units or employee groups in the columns, coloring each cell according to the result. Cells are usually colored by the group's favorable score or by how far that score is from the company average. This lets you read the results of dozens of units across dozens of areas on a single screen and see at a glance which group lags behind in which area. But a heatmap is as easy to misread as it is to read: random differences in small groups can look just as striking as real problems. In this article we explain how a heatmap is built, how to read it and the most common mistakes.

How a Heatmap Is Built

A survey heatmap has four basic components:

• Rows: Survey categories (manager, career and development, recognition, workload and so on), the engagement index and, where needed, individual items. Category level works better for leadership presentations, item level for unit reports.
• Columns: Units or employee groups: department, location, job level, white- or blue-collar, tenure. We explain which breakdowns to use in our article on segmentation in employee surveys.
• Cells: That group's result for that row. In employee surveys the clearest measure is the favorable score (the share answering "Agree" or "Strongly agree").
• Reference column and number of responses: The first column should show the company total, and the top of every column should show that group's number of responses (n). Without the number of responses, it is impossible to judge how reliable the differences are.

Absolute Coloring vs Difference from Average

The most important design decision in a heatmap is what the color represents. There are two common approaches:

In practice, difference coloring is more useful for unit and segment comparisons, but the cell should still display the absolute value. That way the reader sees, in the same cell, both "how different is this group from the average?" and "is this result good in its own right?". An external benchmark can also be used instead of the company average; in that case, state clearly which data the comparison is based on.

Example Heatmap

The table below shows category-level results for an organization where 700 people responded. Cells show the favorable score, with the difference from the company average (in points) in brackets.

The values in the table are illustrative. Results for the Legal unit are not shown because it falls below the 10-person reporting threshold; its responses are included in the company total.

In a real report these cells are colored. For example, five color bands based on the difference could be used: −10 or below dark orange, −5 to −9 light orange, −4 to +4 neutral grey, +5 to +9 light blue, +10 or above dark blue. The bands are illustrative; as explained below, whether a difference is meaningful also depends on the size of the group.

How to Read a Heatmap

Reading a heatmap cell by cell means chasing the most eye-catching color. Instead, read it in three directions.

Down a Column: The Unit's Overall Picture

If a column is almost entirely one color, the problem is not a single topic but the unit as a whole. In the example, the Operations column is below the company average in every category except workload, and in a group of 310, differences of 7–8 points are too large to be explained by chance. This pattern usually points to a broad issue such as unit leadership, communication or working conditions. In that case, rather than an action focused on one category, a general review with the unit's leadership is needed.

Across a Row: Where a Topic Is a Problem

If only a few cells stand out in a row, the topic is a problem in specific units rather than across the organization. In the example, workload and balance is clearly below average in Sales and IT, and above average in the other units. This suggests looking at workload distribution, targets or staffing in these two units rather than launching a company-wide "work-life balance programme".

Clusters: Areas That Move Together

Some categories rise and fall together. In the example, career, recognition and trust in senior leadership are low together in Operations and high together in Sales and Finance. Such clusters usually point to a common cause. Reading the heatmap together with key driver analysis makes it easier to set each unit's priorities: a unit that is below average in an area with a strong impact on engagement needs priority attention.

Which Difference Is Meaningful? Thresholds by Group Size

The most common heatmap mistake is reading differences in small groups with the same seriousness as differences in large groups. In a group of 38, the answers of 2–3 people can move the result by several points. When comparing a group's result with the company average, the following rough thresholds can be used:

Thresholds are approximate and calculated for a favorable score of around 50%; as the score approaches 80% or 20%, the threshold gets slightly smaller. For a precise assessment, use a statistical significance test.

Applying these thresholds to the example:

• Operations (n=310): Differences of 7–8 points are above the threshold; the pattern points to a real difference.
• Relationship with manager in IT (n=60, −10): Just below the threshold, a borderline difference. Rather than drawing a conclusion straight away, look at item-level results and open-ended comments.
• Career in Marketing (n=38, +10): The eye-catching dark blue cell may be chance at this group size. It is too early to present it as "best practice".

Because the group is compared here with the company average, which rests on many responses, the thresholds are somewhat lower than those used when comparing two years of the same group. We explain which thresholds apply when comparing two measurements of the same group in our article on comparing engagement survey results year over year. We cover the relationship between group size and margin of error in our article on sample size and margin of error.

A good heatmap does not leave this to the reader: applying color only to statistically significant differences, or marking significant differences with a symbol, keeps random colors in small groups from becoming a distraction.

Small Groups: The Suppression Rule

Results for groups with fewer responses than the minimum reporting group, that is, 10 people, are not shown in the heatmap. That is why the Legal column in the example is empty. A few points to watch:

• A suppressed group should be suppressed in every row: If a group passes 10 responses on some items but not on others (for example, because of partially completed surveys), handle the whole column with one consistent rule rather than suppressing row by row.
• Prevent identification by inference: If the result of a parent unit and the results of its other sub-units are known, the suppressed small unit's result can be calculated. If this risk exists, show the small unit combined with a sibling unit.
• Do not lose suppressed groups: Their responses are included in the company total and the parent unit's result; they are simply not shown as a separate column.

You can find the details of anonymity rules in our article on anonymity in employee surveys.

Color Choice: Color-Blind Friendly Palettes

The red-green palette most often used in heatmaps is the hardest combination to tell apart for readers with color vision deficiency. Red-green color vision deficiency is fairly common among men, which means the chance that someone on your leadership team cannot read this palette is not small. Our recommendations:

• Use a diverging palette such as blue-orange: Shades of orange for negative differences and shades of blue for positive ones remain distinguishable for readers with color vision deficiency.
• Leave a neutral band in the middle: Cells close to the average should stay grey or white. A map in which every cell is colored highlights nothing.
• Do not let color carry the information alone: Write the value in the cell and, if needed, show the difference with "+" and "−" signs. The table should remain readable when printed in black and white.
• Limit the number of color bands: More than five shades leads readers to try to interpret the difference between light blue and medium blue.
• Always show the legend: What the color represents (absolute score or difference) and where the bands start and end should appear next to every heatmap.

Common Mistakes

• Ranking managers publicly: Showing a manager-level heatmap to the whole leadership team or the whole organization turns the survey from a development tool into a performance ranking. Managers should see their own results compared with the company average; other managers' named results should not be shared. Otherwise, by the next survey managers start "managing" the result.
• Over-reading small differences: Writing an action plan for a 6-point difference in a group of 40 means investing in chance.
• Too many columns: 40 units and 60 items on one screen means 2,400 cells that nobody can read. Limit columns to 8–12 per screen; if there are more units, show them hierarchically (first divisions, then the units under each division).
• Putting different levels in the same table: Showing a division alongside a team within that division counts the same responses twice and blurs the comparison.
• Not showing the number of responses: Without n, it is impossible to know which differences to take seriously.
• Reading only the color: With difference coloring, a unit slightly above average in an area that is low company-wide looks blue, yet its absolute result may still be weak. Read the value in the cell too.
• Stopping at the heatmap: A heatmap answers "where", not "why". Dig into the cells that stand out with item-level results, open-ended comments and conversations with the unit's manager.

Where the Heatmap Fits in the Analysis Process

The heatmap belongs in the middle of the analysis process, not at the start: first the company-wide picture and the areas that drive engagement are identified, then the heatmap shows which units differ in those areas, and finally you drill down to item and comment level in the units that stand out. We explain the whole process step by step in our guide on how to analyze employee survey results.

Frequently asked questions