Every closed-ended question in your survey results comes with a chart you can reshape, read statistically, compare against other surveys or tests, and export as an image or a spreadsheet. This article covers all of those controls in one place, from switching between counts and percentages to running an A/B comparison across the concepts in an activity group.
Reach for it when you are preparing a readout and need a chart that says exactly what you mean, when you want to know whether a 62% result is solid or just small-sample noise, when you are checking whether a rating improved since the last round of testing, or when three concepts were shown to different testers and you need to see them side by side.
The chart controls are the same on every plan and every access level. What changes is the question: rating scales get averages and confidence intervals, category questions get shares and a most common answer, and ranking questions get scores. The Response Summary under each chart works out which applies.
Changing the chart type and number format
Open your test's Survey Results and find the question. The controls sit in the toolbar at the top right of its chart.
Click the bar chart icon to choose Column Chart, Bar Chart, or Pie Chart. The current one is marked (Current). Every question starts as a column chart.
Click the # or % button to switch the bars and labels between response counts and the percentage of testers. The tooltip reads Change to Counts or Change to Percent of Testers. Scale questions start in percent; everything else starts in counts.
Counts are best for understanding how many people are behind a bar; percentages are best for comparing questions with different response counts or reporting to stakeholders. The statistics underneath do not change either way, only how the distribution is drawn. On a multi-select question the note under the chart reminds you that each percentage is the share of testers who chose that option, so the values can total more than 100%. Ranking questions always show a score, so the toggle does not appear on them.
Chart type and number format are not saved. Reloading the page returns every chart to a column chart in its default format.
Chart statistics
Under every closed-ended chart is a Response Summary, a strip of headline statistics with a chip showing the measurement type BetaTesting detected for the question, such as Likert, Numeric Rating, Nominal · single-select, Binary, or Ranking. Hover the chip for a one line description. Click Show more to expand the full panel and Show less to collapse it.
The collapsed strip shows Responses and the statistics that suit the question: Average, Median, Mode (most common), and Range for numeric scales; a signed Average with Share (+) and Share (-) for agreement scales; Mode and a Binary verdict for yes or no questions; Top ranked for rankings. A 0 to 10 recommendation question adds Net Promoter Score.
Expanding adds the spread and certainty measures: Confidence Interval for Average, Interquartile range, Standard Deviation, and Variance, plus a table with each option's Count, Share, and the confidence interval for both.
When expanded, use the α menu in the strip header to switch between 90% confidence, 95% confidence, and 99% confidence. Every interval in the panel, on the chart, and in the chart's spreadsheet export follows the level you pick. The default is 95%.
Click any statistic to draw it on the chart as a violet overlay: a dashed line for the average, a shaded band for its confidence interval or the interquartile range, markers for the range, and an outline around the median or most common bar. Click it again to clear it. Hover a cell in the option table to preview that option's value and interval on the chart.
Two details are worth knowing when you read the numbers. On agreement scales the average is signed and centered on neutral, so +36% reads as net agreement and a confidence interval that straddles zero flags a divided result. And when an option such as "Not applicable" is not a point on the scale, an info icon next to the type chip explains that it still counts as a response but is left out of the average, median, and spread.
For a full walkthrough of each statistic, how the confidence intervals are calculated, and how Likert averages are scored, see Chart statistics and overlays.
Comparing charts across surveys, tests, or time
Any multiple choice, checkbox, or scale question can be compared with the same question elsewhere. A match is a question with exactly the same text and type in a live survey that has responses, anywhere in your company's tests.
Click the three dot menu at the top right of the chart and choose Compare. BetaTesting searches for comparable items; if there are none, it tells you no past surveys could be compared with this question.
In the Compare Data dialog, pick the surveys under Results to compare, or leave it on All Results. Each option shows the test, the survey, and the test's start date.
If you kept All Results, choose a Time period, from Last month to Last 5 years. The default is the last three months.
Choose the X-Axis groupings: By test to see each test as its own column, or By time period to combine the tests that ran in the same week, month, or year and compare periods.
Choose the Chart Type: Stacked Column or Stacked Area to compare the full distribution, or Line (w/ stick intervals) to plot each group's average with a stick showing its standard deviation. Click Compare.
The comparison replaces the single chart. Each compared group becomes a series named after its test, or after the test and survey when one test contributes more than one. You can still switch chart types from the bar chart icon, and the footer reads Comparing past results with an Exit comparison link to return to this survey alone.
On long scales the compared bars are shaded from one end of the scale to the other. For scales with eight or more points a switch above the chart lets you view the Full scale or collapse it into Thirds, and eleven point scales add Quartiles, which makes a shift from one end to the other easy to see. Exports always contain the full scale.
The filters at the top of the page apply to comparisons too. Each compared survey is filtered the same way, so every series counts the same kind of tester. Two kinds of filter cannot travel: filters on answers to this survey's questions, and filters that belong to this test such as test devices. When one is dropped, a note above the chart says so.
See Comparing charts for worked examples of each comparison chart type.
Using comparison for A/B and concept tests with activity groups
An A/B or concept test on BetaTesting is one test with one survey per variant, bundled in a randomized activity group so each tester receives one concept, or all of them in a balanced order. Because every variant survey asks the same questions, chart comparison is how you read the result.
Build one survey per concept with identical question text, then group them and turn on randomization. See Using activity groups.
In Survey Results, use the survey dropdown in the page heading to open any one of the variant surveys.
On the question you want to compare, choose Compare from the chart's three dot menu and select the sibling surveys under Results to compare. Each series is labeled with the test and survey name, for example "Concept test — Concept B".
Use Stacked Column to compare the distribution of answers across concepts, or Line (w/ stick intervals) to compare their averages with a spread stick. Apply the page filters to check whether a concept wins only with a particular segment.
Matrix questions have their own version of this. Click Compare on the matrix toolbar, choose Surveys to compare or All surveys in this test, and the summary grid becomes a comparison: each cell is shaded by its rank within the statement, and clicking a statement expands its comparison chart. Only surveys in the same test with a matching matrix are offered, and the dialog tells you when a survey is missing a statement or reads its scale differently. See Ask several ratings at once with a matrix question.
Comparison charts do not run significance tests. To judge whether a difference between concepts is real, open each concept's survey and read the confidence interval on its Response Summary at the same confidence level; if the intervals do not overlap, the difference is unlikely to be chance.
Exporting a chart as an image
The three dot menu at the top right of every chart has three ways to get a picture of it.
Save as Image downloads the chart as a PNG named after the question.
Print opens a print view with the question text and the chart.
Save as PDF captures the chart together with its Response Summary and any overlay you have drawn, and saves it as a PDF named after the question.
Save as Image and Print render the chart alone, so overlays and the statistics strip are not included; use Save as PDF when you want them in the picture. The Print button at the top of the page does the same for every chart at once, with Print, Print (charts only), Save as PDF, and Save as PDF (charts only).
Exporting a chart to a spreadsheet
From the same three dot menu, choose Export the data to Excel to download an .xlsx workbook for that one question. It contains:
The chart itself, as a native Excel chart you can restyle.
Response Data: the count for each answer option.
Response Summary: the number of responses, the measurement type, the average, shares, median, most common answer, standard deviation, range, Net Promoter Score where it applies, and the confidence intervals at whatever level the α menu was set to when you exported.
Detailed Response Data: one row per tester with their tester ID and answer.
The export respects the filters applied to the page, so the detailed rows always match the chart above them. Exporting a comparison chart produces a combined workbook with the aggregated counts across every compared survey and detailed rows that add the test name, test ID, and test date, without the statistics block.
This is different from the Export button at the top of the survey results page, which downloads the testers grid: one row per tester across every question, with whichever demographic and tag columns you tick, and no statistics. Use the chart export when you want one question with its statistics; use the page export when you want everyone's answers to everything.
Things to know
Filters apply everywhere. Charts, the Response Summary, comparisons, and exports all recalculate against the filters at the top of the page, which is how you compare segments such as new versus returning testers.
Nothing about a chart's view is saved. Chart type, count or percent, the scale grouping, overlays, and comparisons all reset when you reload the page. Export the image or spreadsheet if you need to keep a particular view.
The Response Summary appears on closed-ended questions once a test has enough responses to classify the question. Open text and file upload questions do not get one, and a question that is genuinely ambiguous is left without a summary rather than mislabeled.
Comparison matches on exact question text. If you edit a question's wording between rounds of testing, the two versions will not be offered for comparison.
Comparisons show distributions and averages, not significance. Use the confidence intervals in each survey's Response Summary to judge how certain a difference is.