Building the survey is the half everyone expects to be hard, and it is the half that takes twenty minutes. The part that causes trouble later is the set of decisions made along the way that are difficult or impossible to reverse: which collector was used, whether the data can be exported at all on the current plan, and what happens to collected answers when a question is edited after the fact.
This walks the whole path in order, from the Create Survey button to a file on a disk, and names the points where a choice locks something in.
Four ways to start, and the one that decides the rest
Create Survey sits in the upper right of the account. It offers four routes: start from scratch, copy an existing survey, search surveys and templates, or have it built for you. The build option is an AI assisted draft, powered by OpenAI, that turns a description into a set of questions.
There is also a Question Bank, a library of pre written questions, and a paste option that takes questions and answer choices written elsewhere and turns them into a survey all at once rather than one question at a time. For anyone who already has the questions in a document, that paste route is usually the fastest way in and it is the one people most often miss.
The route to prefer depends on whether this survey will run again. If it will, copy an existing survey rather than starting fresh, because a copy keeps the question wording identical and identical wording is what makes two rounds comparable later. Templates are the right starting point for a genuinely new topic, and the template library is large, but a template rewritten heavily is no longer a template and the comparability argument no longer applies.
Building the questions without painting yourself into a corner
Question types divide into open ended and closed ended, and the settings inside each one control the shape of the data it collects. The rule that saves the most time is to choose the question type for the report you want rather than for the way it looks on screen. A free text box where a rating scale belonged cannot be turned into a chart afterwards.
Three things about building are worth knowing before rather than after.
Deleting a question deletes the responses to it. This is stated plainly in the help documentation, and it is the single most expensive mistake available during a live survey. If a question is wrong halfway through collection, the safe move is to stop showing it rather than to delete it.
On the free plan, everything counts towards the ten question limit. Text and image elements are included in that count, so a survey with a welcome note, a logo and eight real questions is already at the ceiling. Teams that plan around ten questions and then add explanatory text find the limit earlier than expected.
Very long single pages cause loading problems. A design with a large number of questions, answer choices or images on one page can be slow enough to time out. Splitting across pages is not only a readability decision.
Two further options are worth knowing about while building. A consent statement or privacy notice can be placed in front of the questions, with skip logic used to disqualify anyone who does not agree, which is how a lawful basis gets recorded rather than assumed. And accessibility is supported to Section 508 and WCAG 2 standards, which matters if the survey is going to a public sector audience or to anyone using a screen reader.
Two things a survey can do that people often build elsewhere
Before reaching for a second tool, two capabilities are worth checking against the requirement.
Payments can be taken at the end of a survey through a Stripe integration, by adding a payment page after the questions. For a paid workshop booking or a membership renewal, that removes the usual awkward step of sending someone off to a separate checkout and hoping they arrive.
Any survey can also be turned into a quiz by adding quiz questions. Respondents can be told whether an answer was right, or the scoring can be kept private, and the analysis then includes overall score statistics as well as an individual score for each person. Training teams building a comprehension check after a course session frequently write this by hand in a spreadsheet when it already exists.
Neither is a reason to choose the tool on its own. Both are reasons to check the feature list before commissioning something custom.
Preview it as a respondent, on a phone
The preview is not a formality. Open it on a phone, answer every question honestly, and time how long it takes, because the length that feels reasonable while building rarely feels reasonable while answering. Check that required questions are the ones you would still require if a stranger were filling it in at a bus stop, and check that the skip logic actually skips. Logic errors are invisible in the builder and obvious the moment a real path is walked through, and they are far cheaper to find before the collector goes out than after the first hundred answers arrive down the wrong branch.
Choosing a collector, which is a bigger decision than it looks
Once the survey is designed, responses are gathered through a collector, and the collector determines what you will be able to say about the results.
| Collector | What it gives you | What it costs you |
|---|---|---|
| Web Link | A short URL and a QR code, postable anywhere, with a custom URL ending available | No knowledge of who received it |
| Email Invitation | Tracking of who opened, clicked through and responded, plus reminders and thank you messages | Requires the address list and a plan that includes it |
| Embed | The survey inside a page of your own | The same anonymity as a web link |
| Bought responses | An audience you do not have | Sold separately from the plan allowance |
Email Invitation is the only route that answers the question everybody eventually asks, which is who has not responded yet. It tracks opens, click throughs and responses per recipient, and it supports reminder and thank you messages and scheduled sending. A web link cannot do this, and no amount of analysis afterwards recovers the information.
The collector options screen is where several useful settings hide. A survey can require a password, close automatically on a date or after a set number of responses, and be kept anonymous. Multiple responses from the same browser can be allowed, which is the setting for a shared device or a kiosk. And response editing can be switched on, which has a second effect worth knowing: it is also the mechanism by which a respondent can complete part of the survey and come back later to finish, because without it there is no way to save and resume.
There is also a recurrence option that generates fresh web links on a schedule and emails them to you when they are ready, which is how a quarterly survey stops being a manual chore.
Getting the answers back out
Responses land in the analyse results section, and this is where the plan matters more than anywhere else, because exporting is a paid feature. On a free account the answers can be read in the browser and not taken out of it.
The export path itself is short. Go to analyse results, select Export, then New export. Either question summaries or individual responses will already be selected depending on where the export was started from. Then choose the file format from PDF, PPT, XLS, CSV or SPSS, adjust the options that apply to the chosen format, optionally rename the file, and export. A single individual response can also be exported as a PDF from the individual responses tab in single view.
Two limits apply to the output and both are documented. Finished exports are kept for 14 days and can be retrieved from Previous exports, so an export is not a permanent archive. And when a survey passes 100,000 responses, or when respondents multiplied by questions passes 100,000, only CSV is available. Anyone planning to hand a formatted PPT deck to a committee after a very large survey should know that before promising it.
One more thing to note in the other direction: responses can be entered manually, one at a time, for answers collected offline, but there is no way to upload responses in bulk. Data does not flow into SurveyMonkey from a spreadsheet.
The three decisions that are hard to undo
Read back over the path and three points stand out as the ones that cause real damage.
The first is deleting a question mid survey, which takes its responses with it. The second is choosing a web link when the question that will be asked later is who has not replied, because that information was never collected. The third is building substantial work on a free account, since exporting is paid and the answers stay inside the browser until the plan changes.
None of these are obscure. They are all visible in the documentation before the survey goes out, and all three are usually discovered afterwards.
When the answers are one to one rather than a sample
There is a category of form that runs through all of this awkwardly, and it is worth naming because a great many people reach for a survey tool to do it.
A survey is built to measure a population. Aggregate the ratings, cross tabulate them, export the deck. The unit of value is the chart. An application, an enquiry, a repair request or a course booking is the opposite: each response is a person waiting for an answer, and no chart helps. What is needed there is to know who is handling each one, what stage it has reached, and whether a reply has gone out.
Survey platforms do not hold that state, which is why teams using them for applications end up maintaining a spreadsheet beside the tool, adding a status column and an owner column by hand. A form tool with response management treats exactly that as the job: every response carries an owner and a stage, the reply is written on the same screen as the answers, and the history of what was sent stays attached to the person. It is worth comparing what the tool does after the submit rather than how the builder looks, and the patterns where one to one handling is the whole point are easy to recognise once separated from survey work.
What to set up before sending anything
Decide the collector before writing a single question, because that choice determines whether you will ever know who has not answered. Then check that the current plan includes exports, since that is the difference between a result you can hand to someone and one you can only look at. If the answers turn out to be one to one rather than a sample, the tool to try is one built for replying rather than for analysing.
Q1. Can questions be edited after the survey has started collecting?
Editing is possible, but deleting a question also deletes the responses already collected for it, which the documentation states directly. For a live survey the safer approach is to add a corrected question and stop using the old one rather than removing it, so that the answers already given stay in the data.
Q2. Does the free plan really only allow ten questions?
Yes, and the count includes text and image elements as well as questions, so a welcome message and a logo consume two of the ten. Exports are also a paid feature, which means work done on a free account can be read in the browser but not taken out of it.
Q3. How do you tell who has not responded yet?
Only an email invitation collector can answer that. It tracks who opened the invitation, who clicked through and who responded, and it supports reminders. A web link or QR code collects responses without identifying recipients, and that gap cannot be filled in afterwards.
Q4. What formats can results be exported in?
PDF, PPT, XLS, CSV and SPSS, chosen at the point of export, with different options available depending on the format. Above 100,000 responses, or when respondents multiplied by questions exceeds 100,000, only CSV is offered. Completed exports remain available for 14 days under Previous exports.
Q5. Can existing responses be imported into a survey?
No. Responses can be entered manually one at a time, which covers answers collected on paper or by phone, but there is no bulk upload. Anyone consolidating data from another tool should plan to keep the old records where they are, or to combine the exported files outside the platform.