Likert Scale Summary
Paste your survey answers and get per-item means, SDs, full response distributions, top-2-box and bottom-2-box percentages, and stacked bars. Runs in your browser; nothing is uploaded.
About this tool
Likert Scale Summary turns raw agreement/satisfaction/frequency answers into the numbers survey reports actually quote: a mean per item, its standard deviation, median and mode, the full response distribution, and the top-2-box / bottom-2-box percentages — plus text stacked bars you can paste straight into a doc or ticket.
Feed it either shape of data. Responses is the usual export: the first row
holds the item (question) headers and every later row is one respondent, with
answers as codes 1–N or as the labels themselves (Agree, Strongly agree,
…). Counts is the tally you often get from a summary table: one row per item,
then how many respondents chose each category. Everything runs locally in your
browser — no upload, no account.
Worked example
Seven respondents rated three items on a 5-point agreement scale:
Ease of use,Support,Value for money
5,2,4
4,3,4
5,1,3
4,2,5
3,4,4
5,3,4
4,2,5
With Item order set to Highest mean first, the item table reads:
Item n miss mean sd median mode Bottom 2 Neutral Top 2
-------------------------------------------------------------------------------------------------
Ease of use 7 0 4.29 0.76 4.00 4,5 0.0% 14.3% 85.7%
Value for money 7 0 4.14 0.69 4.00 4 0.0% 14.3% 85.7%
Support 7 0 2.43 0.98 2.00 2 57.1% 28.6% 14.3%
Overall mean of item means: 3.62 (21 valid answers, 0 missing)
The distribution and the stacked bars follow, one row per item, with a key mapping each character back to its category:
Ease of use 3333334444444444444444455555555555555555
Value for money 3333334444444444444444444444455555555555
Support 1111112222222222222222233333333333444444
Key: 1=Strongly disagree 2=Disagree 3=Neutral 4=Agree 5=Strongly agree
Support is clearly the weak item: 57.1% in the bottom two categories against
85.7% top-2-box for the other two.
Options
- Data shape — responses (one row per respondent) or counts (one row per item holding a tally per category).
- Item columns — name just the Likert columns when your export also carries respondent IDs, timestamps, or free-text comments.
- Scale points — 2 to 11 categories; 4, 5 and 7 are the common ones.
- Category labels — agreement, satisfaction, frequency, quality, plain numeric, or your own comma-separated list.
- Reverse-scored items — flip negatively worded items with
new = points + 1 − answerso every item points the same way. - Box size — 2 gives the usual top-2-box / bottom-2-box; 1 gives top-box only.
- Missing answers — drop them item by item, or drop any respondent who skipped an item (listwise).
- Stacked bars / diverging — plain left-to-right bars, or bars centred on the neutral midpoint so negative and positive halves are easy to compare.
- Cronbach's alpha — internal-consistency reliability for the items taken as one scale.
Limits and edge cases
- Answers must be whole codes from
1to your scale-point count, or one of the category labels (matched case-insensitively, and by unique prefix). Anything else is an error naming the item and the value — so a stray ID column fails loudly rather than being scored as data. - Blanks and the markers
NA,N/A,-,.,none,null,missingand?count as missing; every other value is data. - Means and SDs treat the scale as interval data. That is standard practice for reporting, but Likert answers are strictly ordinal — the median, mode and box percentages carry no such assumption.
- The SD is the sample SD (n − 1) and needs at least 2 answers; the median of an even number of answers is the average of the two middle values, so it can land on a half-point.
- Cronbach's alpha needs at least 2 items and at least 2 respondents who answered every item, and it is not available from counts input (it needs respondent-level rows). A negative alpha usually means an item runs the other way — reverse-score it and re-run.
- Percentages are always shown to 1 decimal place; the decimals setting applies to means, SDs, medians and alpha.
- Bars are exactly 40 characters, allocated by largest remainder, so a category under about 1.25% of an item can round to zero characters while still appearing in the distribution table.
FAQ
What is a top-2-box score, and why use it?
Top-2-box is the share of respondents who picked either of the two most positive categories — on a 5-point agreement scale, Agree plus Strongly agree. Bottom-2-box is the mirror image at the negative end. They are quoted alongside the mean because a mean of 3.5 can come from a well-liked item or from a badly split one, while the box percentages show the split directly. Set Box size to 1 if you want top-box (top category only) instead.
How do I handle reverse-worded items?
List them under Reverse-scored items, by header name or 1-based index. Each
answer becomes points + 1 − answer, so on a 5-point scale a 1 becomes a 5. Do
this before reading the item means or Cronbach's alpha: an un-reversed negative item
drags the scale mean down and can push alpha below zero.
Can I paste a summary table instead of raw responses?
Yes — switch Data shape to Counts and give one row per item: the item name, then how many respondents chose each category, lowest category first. A header row of category names is detected and skipped. Everything except Cronbach's alpha is computed the same way, because means, SDs, medians and box percentages all follow from the frequency counts.
What do the diverging bars show?
The plain bars stack every category left to right across 40 characters. The diverging option splits each bar at the scale midpoint: negative categories extend left of the centre line and positive ones right, with the neutral category split across it. That lines the items up on a common centre, which makes "which items lean negative" readable at a glance rather than requiring you to compare segment widths.
Do means make sense for Likert data?
They are contested. Likert answers are ordinal — the gap between Disagree and Neutral is not guaranteed to equal the gap between Agree and Strongly agree — so a mean assumes something the data does not strictly provide. In practice means are the standard way to rank items and track them over time, which is why this tool reports them. The median, mode, distribution and box percentages are reported next to every mean precisely so you can check any claim against a measure that makes no interval assumption.
What counts as a missing answer?
An empty cell, or a cell holding NA, N/A, -, ., none, null, missing
or ? (any capitalization). With Missing answers set to Exclude, each item
keeps every answer it actually received and reports its own n and miss counts.
With Listwise, any respondent who skipped at least one item is dropped entirely,
which keeps every item based on the same people — the report says how many rows were
dropped.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool likert-summary "Ease of use,Support,Value for money
5,2,4
4,3,4
5,1,3"New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/likert-summary/?data=Ease%20of%20use%2CSupport%2CValue%20for%20money%0A5%2C2%2C4%0A4%2C3%2C4%0A5%2C1%2C3&input=responses&items=Ease%20of%20use%2CSupport&points=5&scale=agreement&labels=Never%2CRarely%2CSometimes%2COften%2CAlways&reverse=Support&box_size=2&missing=exclude&sort=input&decimals=2&chart=true&diverging=true&alpha=true&delimiter=%2CMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.