Guide · Survey analysis
Survey analysis: how to analyse your survey results step by step
To analyse survey results, you first clean the data, then read the frequencies for each question, compare groups in crosstabs and with filters, sort the open answers and check which differences hold up. Finally, you summarise the key findings in a report. This guide walks through all seven steps and shows what 121surveys does for you along the way.
Building and testing is free. To publish, you choose a paid plan.
Analysing a survey in seven steps
1. Clean the data
Before you calculate anything, check which answers belong in the analysis. Incomplete responses, test runs and suspicious patterns distort results: anyone who clicks through a questionnaire in a fraction of the usual time or picks the same point all the way down a grid probably did not read it. Decide beforehand which rules you will use to exclude answers, and stick to them – otherwise you unconsciously bend the data towards the result you hoped for.
2. Read the frequencies
Start by looking at each question on its own: how are the answers distributed, and how many people answered at all? Always state the base, meaning the number of answers a percentage refers to. For scales, the mean and median help; for multiple choice, the share per option – which then adds up to more than one hundred per cent, because one person can give several answers.
3. Build crosstabs
A crosstab sets two questions against each other, such as satisfaction by age group or likelihood to recommend by product. That shows you whether one group answers differently from another. Watch out for small cells: five answers in a group are not enough to draw conclusions, however striking the percentage looks.
4. Filters and segments
Filters narrow down the analysis – to a period, a distribution channel, an answer to another question or a piece of information from the link. Segments put several groups side by side. That lets you answer questions such as: has satisfaction changed since the last survey? Do new customers answer differently from long-standing ones?
5. Analyse open answers
Free-text answers often contain the most valuable hints, but they take effort to read. Create categories, assign each answer and count how often a topic comes up. With many answers, AI can do this; spot-check its assignments before you pass on the figures, and label AI-generated results as such.
6. Treat significance with care
A significance test tells you whether a difference is probably more than a chance result of your sample. It does not tell you whether the difference is large or important. With very many answers, even tiny differences become significant; with few, large ones go unnoticed. So always look at the strength of a relationship and the number of answers too, and do not keep testing groups against each other until something turns up by chance.
7. Report and presentation
A good report starts with the three to five most important findings, not with question 1. Every chart has a title that states the message, plus the base. Keep results separate from interpretation, name the limits of the survey and close with recommendations. For further work, include the raw data in an open format.
What 121surveys handles in the analysis
Quality check
A quality score flags responses that were too fast and straight-lining in grids; a filter lets you hide weak responses. Test runs are left out of every statistic.
Frequencies and key figures
Every question with its distribution, mean and metrics such as NPS, CSAT or CES – calculated in the app, without exporting.
Crosstab with a test
Two questions, or a question and a link variable, against each other, with a chi-square test, p-value and Cramér's V for the strength of the relationship.
Filters, segments, weighting
Filter by period, distribution channel, answers or link variables, compare segments side by side and weight the sample to known proportions.
AI categories for free text
AI assigns open answers to categories and a sentiment in the background. 121surveys removes email addresses and phone numbers first, and results are labelled as AI-generated.
Export, report, presentation
Raw data as CSV, Excel, SPSS, ODS or JSON; reports as PDF or PowerPoint and by link; presentations for PowerPoint or Google Slides.
Frequently asked questions
How do I analyse a survey in Excel?
Export the raw data: each row is one response, each column one question. With pivot tables you count frequencies and build crosstabs. Much of this, including significance tests and filters, 121surveys already calculates in the app – the export is mainly meant for your own further processing.
How do I analyse open-ended questions?
First read a sample, derive categories from it and assign all answers. Count how often each category occurs and pick typical quotes. In 121surveys, AI does the assigning in the background; you check the result and decide what to do with it.
How many responses make a survey meaningful?
It depends on how precise you need to be. With 100 responses, the margin of error for a proportion is up to about ±10 percentage points; with 400 responses, about ±5. For group comparisons you need enough responses in every group. The sample size calculator works it out for your figures.
What does significant mean in a survey?
A difference is significant if, at the chosen error probability – usually five per cent – it is unlikely to have arisen by chance in the sample. That is a statement about chance, not about importance: whether the difference matters to you depends on its size.
Can I share results with my team?
Yes. You share reports by link, including with people who have no account, or export them as PDF, PowerPoint or Google Slides. Within the team you work on the same survey together, with read or edit rights depending on each person's role.
From raw data to report
Create your survey in 121surveys, try it for free via the test link and start analysing as soon as the first real responses come in.
