Online Tool Store
· 4 min read
How to Chart Likert Survey Responses
Heshan Fernando
Co-founder & COO
Twelve survey questions, each answered on a five-point agreement scale, and you need one chart that shows which ones people agreed with. Stack the responses as bars and every segment starts at a different place, so comparing question 3 with question 9 means measuring with your eye across a gap.
The diverging stacked bar solves that with one change: a shared centre line.
Why the centre line matters
In a standard stacked bar, each question’s bar runs from 0 to 100%. The “agree” segment might start at 31% on one question and 58% on another, so their left edges don’t line up and the eye can’t compare them.
A diverging chart puts the neutral point at a fixed horizontal position for every question. Disagreement extends left, agreement extends right, and every bar pivots around the same line.
Now comparison is trivial: the further right a bar extends, the more agreement, and you can rank twelve questions at a glance without reading a single number.
Where to put neutral
Two conventions, both defensible.
Split neutral across the centre — half on each side. This centres the chart on true neutrality, and it makes the agree and disagree extents directly comparable. The cost is that the neutral segment is visually divided.
Place neutral entirely on one side, usually the disagree side. Keeps the segment intact and is slightly easier to read, at the cost of shifting the pivot.
Pick one, apply it to every chart in the report, and say which in the caption. Mixing them within a document is worse than either choice.
The averaging argument
Somebody will want a mean score per question. It’s tempting — one number per question, easy to rank.
The objection is that a Likert scale is ordinal. The responses are ordered, but the distances between them aren’t known to be equal: the gap between “agree” and “strongly agree” may not equal the gap between “neutral” and “agree”. Averaging assumes those gaps are equal, which the scale doesn’t establish.
In practice, means of Likert items are reported constantly, and the argument about whether that’s legitimate is long-running and unresolved. The pragmatic position: report the distribution as the primary result, since it’s unambiguous, and if you also report a mean, be clear it’s a summary rather than a measurement.
| Approach | Strength | Weakness |
|---|---|---|
| Diverging bar | Compare across questions instantly | Needs a consistent neutral rule |
| Stacked bar | Shows full distribution | Hard to compare between questions |
| Mean score | One number, easy to rank | Assumes equal intervals |
Common mistakes to avoid
- Using a stacked bar and asking readers to compare segments that don’t share a baseline.
- Changing the neutral convention between charts in one report.
- Ranking questions by mean without showing the distribution, which hides polarisation — a question with everyone neutral and one with half strongly agreeing and half strongly disagreeing can have identical means.
- Colouring agreement and disagreement in hues that a colourblind reader can’t separate.
- Reporting percentages without the response count, so a question answered by nine people looks as solid as one answered by nine hundred.
How to do it with Likert Chart Maker
The Likert Chart Maker builds the diverging chart in your browser.
- Enter the response counts for each question, in scale order.
- Choose where the bars centre and apply it consistently.
- Read across questions at the same point on the scale — that’s what the shared centre enables.
- Include the response count per question alongside the chart.
Other chart builders are in the tools directory.
Frequently asked questions
Why a diverging bar rather than a stacked one?
Because a shared centre line lets you compare agreement across questions at a glance. In a normal stacked bar every segment starts at a different position.
Should neutral be split across the centre?
Both conventions exist. Splitting centres on true neutrality; placing it on one side keeps the segment intact. Pick one and use it consistently.
Can I average Likert responses?
It’s common and contested. The scale is ordinal, so the distance between points isn’t known to be equal. Reporting the distribution avoids the argument.
Final thought
Put the centre line in the same place on every question and the chart answers the question by itself. Everything else about Likert charting is a caption.