# Playtest survey questions: what to ask your playtesters

> The playtest survey questions that work for most games, with questions by goal, questions to avoid, scales, sample sizes and a free survey builder.

By Yos Riady. Updated 2026-11-19. Published 2026-11-19. Canonical URL: https://manalens.dev/blog/playtest-survey-questions

Ask 11 questions right after the session: one overall rating, ratings for clarity, difficulty and controls, and open questions about the best and the worst moment. Keep both ends in each rating question, such as "easy or hard". Five testers find about 85% of usability problems (Jakob Nielsen, 2000). Ratings need 20 testers or more.

## Key points

- Ask 11 questions or fewer right after the session: one overall rating, ratings for clarity, difficulty and controls, and open questions about the best and worst moment.
- Do not ask testers if they will buy the game; ask which similar game they bought most recently and what they did after the test.
- Five testers find about 85% of usability problems (Jakob Nielsen, 2000), but average ratings need 20 testers or more.

## Which playtest survey questions work for most playtests?

These 11 questions work for most playtests. Copy the list, or open the playtest survey builder to change it for your playtest type, genre, session length and goals. The answer type of each question is in brackets.

The list has 6 closed questions and 5 open questions. By our estimate of 20 seconds for a closed question and 60 seconds for an open one, it takes about 7 minutes. The order goes from broad to specific, as games user researcher Steve Bromley advises: the overall rating comes before the questions about parts of the game.

Each rating is paired with an open question about the same topic. Bromley calls a rating question plus an open "why" question a powerful combination. He also says it does not replace an interview, because a survey cannot ask a follow-up question.

We wrote these questions for this guide and for the tool. They are not a validated questionnaire. Change the words to match the words your players use.

- How often do you play games like this one? (Single choice: Most weeks; A few times a year; I tried one or two; Never before this test)
- Overall, how was the game? (Scale 1 to 5 (1 = very bad, 5 = very good))
- Which moment do you remember best? Tell us what happened. (Open text)
- Where were you bored or annoyed? Tell us what happened. (Open text)
- How easy or hard was it to understand what to do in the first minutes? (Scale 1 to 5 (1 = very hard, 5 = very easy))
- Describe a moment when you did not know what to do next, if there was one. (Open text)
- How easy or hard was the game for you? (Scale 1 to 5 (1 = much too easy, 3 = about right, 5 = much too hard))
- How easy or hard were the controls to use? (Scale 1 to 5 (1 = very hard, 5 = very easy))
- What broke, froze or looked wrong? Tell us where it happened. (Open text)
- Why did you stop playing? (Single choice: I reached the end of the build; I ran out of time; I got stuck; I got bored; A bug or crash stopped me; Other)
- What else do you want to tell us? (Open text)

- [Playtest survey builder](https://manalens.dev/tools/playtest-survey-builder)

Sources: Steve Bromley, Games User Research (12 January 2022).

## How many questions should a playtest survey have?

Our practice: 6 to 12 questions, and fewer after a short session. The table shows the targets that the playtest survey builder uses. These numbers are our own practice, not a research result.

Short is better for three reasons. First, the tester has just played and has limited patience, and you want that patience for the open questions. Second, each question must lead to a decision: Bromley advises to start from what you want to learn and to keep the survey short. Third, a fixed short list is easy to repeat with every build.

Schell Games is an example of a short fixed list. Project director Marc Tattersall asked testers of I Expect You to Die the same five questions after they played. Jesse Schell said that a fixed number of questions gives consistency between playtests, so the team can see if it makes progress.

Published questionnaires show the same pressure to stay short. The Player Experience Inventory has 30 items, and its website also offers a mini version with 11 items. The Game Experience Questionnaire has a concise in-game version next to its core module. If you need more than 12 questions, split the survey: ask up to 2 things during an in-person session and up to 2 things in a message a day later.

- [In-game feedback surveys: join survey answers to player data](https://manalens.dev/blog/in-game-feedback-survey-player-data)

*Target question count by session length in the playtest survey builder (our practice)*

| Session length | Target question count |
| --- | --- |
| Under 15 minutes | 6 to 8 |
| 15 to 45 minutes | 6 to 10 |
| 45 minutes to 2 hours | 6 to 12 |
| Several sessions over days | 6 to 12 |

Sources: Steve Bromley, Games User Research (12 January 2022); Alissa McAloon, Game Developer (10 October 2016); PXI Bench (read 19 November 2026); IJsselsteijn, de Kort and Poels (2013).

## Which questions test onboarding and clarity?

Ask for a rating of the first minutes, then ask the tester to describe a moment of confusion. The two questions to start with are "How easy or hard was it to understand what to do in the first minutes?" and "Describe a moment when you did not know what to do next, if there was one."

The strongest clarity question is not a rating. Ask "In your own words, what is the goal of the game?" A tester can rate the tutorial as easy and still give a wrong goal. The wrong answer is the finding.

A survey tells you that a tester was lost. It does not tell you where every tester stopped. For that, read the first session funnel of the same build, and compare the place of the largest drop with the moments that testers describe.

- [First session funnel and FTUE analysis](https://manalens.dev/blog/first-session-funnel-ftue)

*Questions for onboarding and clarity, from the question bank of the playtest survey builder*

| Question | Answer type | When to ask | What the answer tells you |
| --- | --- | --- | --- |
| How easy or hard was it to understand what to do in the first minutes? | Scale 1 to 5 (1 = very hard, 5 = very easy) | Right after the session | Rates the first minutes with both ends of the scale named in the question. |
| Describe a moment when you did not know what to do next, if there was one. | Open text | Right after the session | Finds the exact place where the game did not explain itself. |
| In your own words, what is the goal of the game? | Open text | Right after the session | A wrong answer shows a clarity problem that a rating hides. |
| How was the amount of explanation at the start? | Single choice: Too little; About right; Too much | Right after the session | Tells you which way to move the tutorial: more help or fewer words. |
| Which rule or feature did you understand late, or still not understand? | Open text | Right after the session | Lists the things the tutorial must teach earlier. |
| How easy or hard was the story to follow? Only for: RPG or adventure (quests). | Scale 1 to 5 (1 = very hard, 5 = very easy) | Right after the session | A quest means little when the tester lost the story. |
| How clear or unclear was it why you won or lost a match? Only for: Strategy or tactics (matches); Multiplayer (matches). | Scale 1 to 5 (1 = very unclear, 5 = very clear) | Right after the session | Players learn a match game only when they can read the result. |
| How often did you know what you wanted to build or do next? Only for: Simulation or management (days). | Scale 1 to 5 (1 = almost never, 5 = almost always) | Right after the session | A management game needs a next goal in the head of the player at all times. |
| What are you trying to do right now? Only for: In-person session. | Open text | During the session | Shows the goal the tester has in mind, without a hint from you. |

Sources: Manalens question bank (our own questions).

## Which questions test fun and the core loop?

Ask for the best moment and for the worst moment in two open questions: "Which moment do you remember best? Tell us what happened." and "Where were you bored or annoyed? Tell us what happened." Concrete moments are more useful than a fun score, because you can find them in the build.

Schell Games used the same idea in its five questions for I Expect You to Die: the list starts with the favorite and the least favorite moment or interaction of the tester. Our questions use our own words.

To test the pull of the core loop, ask about a feeling the tester has now: "When the session ended, how much did you want to keep playing?" Then check the behaviour: one day later, ask how often the tester played the build again.

- [Demo analytics: playtime and demo to wishlist benchmarks](https://manalens.dev/blog/steam-demo-analytics-benchmarks)

*Questions for fun and the core loop, from the question bank of the playtest survey builder*

| Question | Answer type | When to ask | What the answer tells you |
| --- | --- | --- | --- |
| Which moment do you remember best? Tell us what happened. | Open text | Right after the session | Names the part of the game to protect and to show in your trailer. |
| Where were you bored or annoyed? Tell us what happened. | Open text | Right after the session | Points to the moment to fix, in the words of the tester. |
| When the session ended, how much did you want to keep playing? | Scale 1 to 5 (1 = not at all, 5 = very much) | Right after the session | Asks about a feeling the tester has now, not about a plan for the future. |
| How would you describe this game to a friend in one sentence? | Open text | Right after the session | Shows what testers think the game is. Compare it with your own pitch. |
| How fresh or repetitive did each new level feel? | Scale 1 to 5 (1 = very repetitive, 5 = very fresh) | Right after the session | Tests if the core loop still gives something new after the first minutes. |
| Since the test, how often have you played the build again? | Single choice: More than once; Once; Not yet, but I want to; Not at all | A day later | A report of what the tester did, which is stronger than what the tester promised. |
| One day later, what do you remember most about the game? | Open text | A day later | What stays after a night is what players will tell their friends. |
| After a failed run, how much did you want to start a new one? Only for: Roguelike or roguelite (runs). | Scale 1 to 5 (1 = not at all, 5 = very much) | Right after the session | The pull to start again is the core loop of a run-based game. |
| How did it feel to solve a puzzle? Only for: Puzzle (levels). | Scale 1 to 5 (1 = I felt nothing, 5 = very satisfying) | Right after the session | A puzzle game lives on the moment of the solution. |
| What are you thinking right now? Only for: In-person session. | Open text | During the session | A neutral prompt when a tester goes quiet. It suggests no answer. |

Sources: Alissa McAloon, Game Developer (10 October 2016).

## Which questions test difficulty and pacing?

Use a scale where the middle is the target: "How easy or hard was the game for you?" The answer 3 means about right. Both ends of the scale are bad, so do not read a high average as a good result.

Mike Ambinder of Valve showed a similar survey question as an example in his 2009 talk on playtesting: how challenging was each enemy, on a scale from 1 (very easy) to 7 (very hard). A rating for each enemy, level or boss tells you more than one rating for the whole game.

Read the ratings next to the data. A tester who calls the game too easy but failed a level five times had another problem, for example unclear rules. Fail rates and retries per level come from your events, not from the survey.

- [Difficulty tuning with data: fail rates and retries](https://manalens.dev/blog/game-difficulty-tuning-analytics)

*Questions for difficulty and pacing, from the question bank of the playtest survey builder*

| Question | Answer type | When to ask | What the answer tells you |
| --- | --- | --- | --- |
| How easy or hard was the game for you? | Scale 1 to 5 (1 = much too easy, 3 = about right, 5 = much too hard) | Right after the session | The middle of the scale is the target. Read it next to how far each tester got. |
| Which level or moment was the hardest, and what made it hard? | Open text | Right after the session | Gives you a place to look for in your fail and quit data. |
| How did the pace of the game feel? | Single choice: Too slow; About right; Too fast; It changed a lot | Right after the session | Separates a pacing problem from a difficulty problem. |
| When you failed, what did you mostly feel? | Single choice: I knew what I did wrong; It felt like bad luck; It felt unfair; I did not fail | Right after the session | Shows if a failure teaches the player or only blocks the player. |
| How did the length of the session feel? | Single choice: Too short; About right; Too long | Right after the session | Useful for a demo: it tells you if the slice is the right size. |
| When you failed, how fair or unfair did it feel? Only for: Action or platformer (levels); Roguelike or roguelite (runs). | Scale 1 to 5 (1 = very unfair, 5 = very fair) | Right after the session | A hard game can keep players. A game that feels unfair does not. |
| How fair or unfair did the matches feel? Only for: Multiplayer (matches). | Scale 1 to 5 (1 = very unfair, 5 = very fair) | Right after the session | Checks matchmaking and balance from the side of the player. |
| Right now, how easy or hard is this part? Only for: In-person session. | Scale 1 to 5 (1 = very easy, 5 = very hard) | During the session | A rating in the moment, before memory smooths the session. |

Sources: Mike Ambinder, Valve's Approach to Playtesting (GDC 2009).

## Which questions test controls and UI?

Ask one rating for the controls and one question about a concrete event: "How easy or hard were the controls to use?" and "Which action did not do what you expected? Tell us what you pressed and what happened."

Always ask what the tester played with. Scores for keyboard and mouse, controller and handheld are often different, and an average of all three hides the problem.

Controls and interface problems are usability problems, and Jakob Nielsen found that a test with 5 users finds most of those. Our practice: watch 5 testers play first, then use the survey to count how many testers had the problem that you saw.

*Questions for controls and UI, from the question bank of the playtest survey builder*

| Question | Answer type | When to ask | What the answer tells you |
| --- | --- | --- | --- |
| How easy or hard were the controls to use? | Scale 1 to 5 (1 = very hard, 5 = very easy) | Right after the session | A low score here can pull every other score down. |
| What did you play with? | Single choice: Keyboard and mouse; Controller; Touch screen; Steam Deck or other handheld; Other | Right after the session | Lets you split the control scores by input device. |
| Which action did not do what you expected? Tell us what you pressed and what happened. | Open text | Right after the session | Asks for a concrete event, not for an opinion about the controls. |
| How easy or hard was it to find information on the screen, such as health, goals and menus? | Scale 1 to 5 (1 = very hard, 5 = very easy) | Right after the session | Rates the interface apart from the controls. |
| How was the size of the text? | Single choice: Too small; About right; Too large | Right after the session | Text size problems are common on handhelds and TVs. |
| Which settings did you look for or change? | Multi choice: Key or button remapping; Audio volume; Graphics quality; Text size or subtitles; Difficulty; Language; None | Right after the session | Shows which settings testers expect, including the ones you do not have yet. |

Sources: Jakob Nielsen, Nielsen Norman Group (18 March 2000).

## Which questions find bugs and performance problems?

Ask one open question and one count: "What broke, froze or looked wrong? Tell us where it happened." and "How many times did the game crash or freeze?" Then ask for the device, because a bug report without the device is hard to reproduce.

A survey is a weak bug tracker. Testers forget what happened, and they do not know the build number. Our practice: use the survey to learn how much the bugs hurt the session, and use error and crash events from the build to count them.

Steam does not collect this for you. The Steamworks documentation for Steam Playtest says that collecting feedback from players is up to you, and that some developers put a feedback button on the main menu of the game.

- [Playtest analytics: what to measure](https://manalens.dev/blog/playtest-analytics)

*Questions for bugs and performance, from the question bank of the playtest survey builder*

| Question | Answer type | When to ask | What the answer tells you |
| --- | --- | --- | --- |
| What broke, froze or looked wrong? Tell us where it happened. | Open text | Right after the session | An open question finds the bugs that you did not know to list. |
| How many times did the game crash or freeze? | Single choice: Never; Once; 2 or 3 times; More than 3 times | Right after the session | A count from the tester to compare with your crash data. |
| What did you play on? For example: a Windows laptop with a GTX 1650, or a Steam Deck. | Open text | Right after the session | A bug report without the device is hard to reproduce. |
| How smooth or choppy did the game run? | Scale 1 to 5 (1 = very choppy, 5 = very smooth) | Right after the session | Rates performance as the player felt it, not as a frame counter saw it. |
| How much did bugs get in the way? | Single choice: Not at all; A little; A lot; I could not continue | Right after the session | Tells you if the other answers describe the game or describe the bugs. |
| Which new problems did you find after the first survey? Tell us where they happened. | Open text | A day later | Later sessions reach content that the first session did not. |

Sources: Steamworks documentation: Steam Playtest (Valve, read 19 November 2026).

## Which questions test if players want to wishlist or buy?

Do not ask if the tester will buy the game. Ask what the tester did: "Why did you stop playing?" and "Which game like this one did you buy most recently? What made you buy it?"

Steve Bromley gives the reason. A question such as "would you buy this game?" asks players to predict their future behaviour, and that is not reliable data. He advises to ask about what players did before, for example when they bought a similar game and what made them buy it.

The best intent question is a follow-up one day later: "Since the test, which of these have you done?" The options are actions, such as a wishlist add or a second session. For a demo, compare the answers with your demo conversion rate. For a Steam Playtest, track the Wishlist button in the build.

- [Playtest to wishlist: track which playtesters wishlist and buy](https://manalens.dev/blog/steam-playtest-to-wishlist)
- [Demo to full game conversion: how to measure it on Steam](https://manalens.dev/blog/steam-demo-to-full-game-conversion)
- [What is demo conversion rate?](https://manalens.dev/glossary/demo-conversion-rate)

*Questions for willingness to wishlist or buy, from the question bank of the playtest survey builder*

| Question | Answer type | When to ask | What the answer tells you |
| --- | --- | --- | --- |
| Why did you stop playing? | Single choice: I reached the end of the build; I ran out of time; I got stuck; I got bored; A bug or crash stopped me; Other | Right after the session | A report of what happened. It separates lost interest from a lack of time. |
| Which game like this one did you buy most recently? What made you buy it? | Open text | Right after the session | Past purchases predict better than a promise to buy. |
| Which game does this one remind you of? What does that game do better or worse? | Open text | Right after the session | Names the games you compete with in the head of the player. |
| Think of one person you know who would enjoy this game. What do they like to play? Please give no names. | Open text | Right after the session | A recommend question that gives you an audience, not only a score. |
| Since the test, which of these have you done? | Multi choice: Added the game to my wishlist; Followed the game or the studio; Told a friend about it; Played the build again; Looked for videos or news about it; None of these | A day later | Actions after the test are the best intent signal a survey can give. |
| How did the end of the demo feel? Only for: Demo before Next Fest. | Single choice: It ended too early; It ended at a good point; It went on too long; I did not reach the end | Right after the session | A demo sells the full game at its last screen. |
| Is this game on your Steam wishlist now? Only for: Demo before Next Fest; Closed alpha or beta; Steam Playtest. | Single choice: Yes, from before the test; Yes, I added it after I played; No; I do not use wishlists | Right after the session | Asks about a fact of today, not about a plan. |

Sources: Steve Bromley, Games User Research (1 May 2023); Steve Bromley, Games User Research (12 January 2022).

## Which questions test price sensitivity?

Start with a price that the tester really paid: "What did you pay for the last game like this that you bought? Name the game and the price." Then ask for a range with two questions: "At what price would this game feel like a great deal?" and "At what price would this game start to feel expensive, so that you would need to think about it?"

The two range questions follow the idea of the Van Westendorp Price Sensitivity Meter from 1976. The full method asks four price questions and plots the answers of many people. Our two questions are a short form that we wrote, not the full method.

Treat every price answer as an opinion. A tester of a free build has not paid, and a price guess is a prediction of future behaviour. Our practice: use the answers to see if your planned price is inside or outside the range that testers name, and set the price from comparable games.

*Questions for price sensitivity, from the question bank of the playtest survey builder*

| Question | Answer type | When to ask | What the answer tells you |
| --- | --- | --- | --- |
| What did you pay for the last game like this that you bought? Name the game and the price. | Open text | Right after the session | A real price that the tester paid anchors the other price answers. |
| At what price would this game feel like a great deal? | Open text | Right after the session | The low end of the range that testers accept. |
| At what price would this game start to feel expensive, so that you would need to think about it? | Open text | Right after the session | The high end of the range that testers accept. |
| How do you usually buy games like this? | Single choice: At full price near launch; At the first discount; Only at a deep discount; In a bundle or a subscription; I rarely buy them | Right after the session | A habit, so it is more reliable than a price guess. |
| How many hours do you expect a full game like this to last? | Single choice: Under 5 hours; 5 to 10 hours; 10 to 30 hours; More than 30 hours | Right after the session | Players judge a price against the hours they expect. |

Sources: Wikipedia: Van Westendorp's Price Sensitivity Meter (read 19 November 2026); Steve Bromley, Games User Research (12 January 2022).

## Which playtest questions should I avoid?

Avoid questions that suggest the answer, questions about the future, and questions that ask the tester to design the game. The table gives a better version of each one from our question bank.

A leading question includes or implies the answer that you want to hear. Amy Schade of Nielsen Norman Group writes that leading questions give biased or false answers, because people tend to repeat the words of the person who asks. Bromley gives a simple fix for ratings: name both ends, and ask "how easy or difficult" in place of "how difficult".

Asking players for solutions has two problems. Mike Ambinder of Valve listed "People don't know why they do what they do" as a weakness of question and answer sessions. Jesse Schell calls "what do you think we should change" a very abstract question. His team asks a concrete one about a magic wand. Our practice: ask what happened, and keep the solution as your job.

*Questions to avoid, with a better version of each*

| Do not ask | What is wrong | Ask this |
| --- | --- | --- |
| Did you enjoy the exciting combat? | Leading: the question tells the tester what to feel. | Where were you bored or annoyed? Tell us what happened. |
| How difficult was the game? | Leading: it names only one end of the scale. | How easy or hard was the game for you? |
| Would you buy this game? | It asks the tester to predict future behaviour. | Which game like this one did you buy most recently? What made you buy it? |
| Will you play again next week? | It is also a prediction. Ask later what the tester did. | Since the test, how often have you played the build again? |
| How should we fix the boss fight? | It asks for a solution. The tester knows the problem, not the fix. | Which level or moment was the hardest, and what made it hard? |
| Did you understand the tutorial? | A yes or no question. Most testers say yes. | In your own words, what is the goal of the game? |
| How were the controls and the camera? | Two topics in one question. You cannot tell which one the score is for. | How easy or hard were the controls to use? |
| Was the time to kill of the ranged class balanced? | Jargon. Use the words that players use. | When you failed, what did you mostly feel? |

Sources: Amy Schade, Nielsen Norman Group (17 December 2017); Steve Bromley, Games User Research (12 January 2022); Steve Bromley, Games User Research (1 May 2023); Mike Ambinder, Valve's Approach to Playtesting (GDC 2009); Alissa McAloon, Game Developer (10 October 2016).

## When should I ask: during play, right after, or a day later?

Ask most questions right after the session. In our question bank, 47 of the 54 questions are for right after the session, 3 are for during play and 4 are for a day later.

Right after the session, the memory is fresh. The Game Experience Questionnaire is meant to be given immediately after the game session has finished. The guide of the Player Experience Inventory says to measure when the experience is still salient. Microsoft Game Studios asked its playtest questions immediately after play, with a focus on the first hour of the game.

During play, ask little. Mike Ambinder of Valve noted that verbal reports give a view into the thoughts of the player in real time, but that they interfere with the gameplay and create an artificial experience. Our practice: in an in-person session, use neutral prompts only, such as "What are you trying to do right now?", and only when the tester is stuck or silent.

A day later, ask about behaviour. A tester cannot tell you today if the game will stay in memory or if they will come back. One day later, both are facts. Our practice: send two questions at most, in the same message as your thank-you note.

*When to ask, and what each moment is good for*

| When | How | Good for | Questions in our bank |
| --- | --- | --- | --- |
| During the session | Aloud, in an in-person session | What the tester tries to do and thinks at that moment | 3 |
| Right after the session | A form that opens when the session ends | Ratings, best and worst moments, confusion, bugs | 47 |
| A day later | A short message or email | What the tester remembers and did after the test | 4 |

Sources: IJsselsteijn, de Kort and Poels (2013); PXI Bench (read 19 November 2026); Davis, Steury and Pagulayan, Game Studies (October 2005); Mike Ambinder, Valve's Approach to Playtesting (GDC 2009).

## Should I use a 1 to 5 or a 1 to 7 scale?

Either works. Use one scale for the whole survey and keep it the same for every build. The questions in this guide use 1 to 5 with a label on both ends.

The research gives a small edge to 7 points. Jeff Sauro of MeasuringU wrote that 7-point scales are a little better than 5-point scales, but not by much, and that more points help most when a single question measures a topic. Maria Rosala of Nielsen Norman Group notes that agreement scales typically have 5 points and that scales between two opposite words typically have 7.

Game research uses both. The Player Experience Inventory uses a 7-point agreement scale from -3 to +3. The example survey question in the 2009 Valve talk used 1 to 7. Our practice for a small playtest: 5 points. With 10 testers, you will not see a real difference between a 5 and a 6 out of 7.

Two details matter more than the number of points. Label both ends in the question and in the scale, so that 1 and 5 mean the same to every tester. And do not change the scale between builds: Ambinder listed comparisons over time as a strength of ratings, and they only work when the question stays the same.

Sources: Jeff Sauro, MeasuringU (25 August 2010); Maria Rosala, Nielsen Norman Group (7 June 2020); PXI Bench (read 19 November 2026); Mike Ambinder, Valve's Approach to Playtesting (GDC 2009).

## Should I ask an NPS style recommend question?

Not as a score in a small playtest. The Net Promoter Score asks how likely a person is to recommend a product, on a scale from 0 to 10. Answers of 9 or 10 are promoters and answers of 0 to 6 are detractors. The score is the percentage of promoters minus the percentage of detractors.

Nielsen Norman Group lists the limits. The score is rarely relevant with a small sample: a test with 5 users and the question at the end is unlikely to give valid data. The three groups also throw away information. And the score tells you how people see the product, not why.

Our view for games: the limits are larger in a playtest. A tester of a free and unfinished build is not a customer. "Would you recommend it" is a prediction, and testers who signed up for a playtest are kinder than buyers. If you ask it, report the average of the raw 0 to 10 answers and compare builds, not games.

Our practice is an open question in its place: "Think of one person you know who would enjoy this game. What do they like to play? Please give no names." The answers describe an audience in the words of your testers, which is useful for your store page and your pitch to creators.

Sources: Therese Fessenden, Nielsen Norman Group (21 June 2024).

## Can I use a standard questionnaire such as the GEQ or the PXI?

Yes, when you need scores that you can compare with other studies. For a small playtest that must find problems, a short survey of your own is enough. We did not copy items from these questionnaires. Get the items from the authors.

The Game Experience Questionnaire (GEQ) comes from IJsselsteijn, de Kort and Poels at Eindhoven University of Technology (2013). It has a core module, a social presence module and a post-game module, plus a concise in-game version. The core module scores seven components: immersion, flow, competence, positive affect, negative affect, tension and challenge. Know its limits: a 2018 review by Law, Brühlmann and Mekler notes that the GEQ has no formal peer-reviewed publication, and their validation study with 633 participants found no evidence for the seven-factor structure.

GEQ licence terms: the document is on the research portal of the university. The portal says that the authors or other owners keep the copyright, that you may download and print one copy for private study or research, and that you may not distribute the material further or use it for a profit-making activity or commercial gain. Ask the authors before you use it in a commercial project.

The Player Experience Inventory (PXI) comes from Vanden Abeele, Spiel, Nacke, Johnson and Gerling. It has 10 constructs with 3 items each. Five are about the functional level, such as ease of control and clarity of goals, and five are about the psychosocial level, such as mastery and immersion. The authors developed the items with 64 games user research experts and validated the scale with 529 participants.

PXI licence terms: its website calls it an open-access scale that is free to use. The user guide adds a condition for the name: the scale is validated as it is. If you remove constructs, change the words or use another answer scale, you have made your own variant, and you can no longer compare your data with the benchmark data. For more on survey design in games, see the chapter "Surveys in Games User Research" by Brühlmann and Mekler in the book Games User Research (Oxford University Press, 2018). We name it as further reading and did not take claims from it.

Sources: IJsselsteijn, de Kort and Poels (2013); Law, Brühlmann and Mekler, CHI PLAY 2018 (23 October 2018); Vanden Abeele and others, International Journal of Human-Computer Studies (online 19 October 2019); PXI Bench (read 19 November 2026); Brühlmann and Mekler (2018).

## How many playtesters do I need?

It depends on what you want from the test. To find problems, 5 testers for each round are enough. To trust an average rating, you need 20 or more.

For problems: Jakob Nielsen wrote in 2000 that one test user finds almost a third of the usability problems, and that a test with 5 users finds about 85%. His model uses a typical value of 31% of problems found by a single user. He advises to spend a budget of 15 users on 3 tests with 5 users each, and to fix the design between the tests. This is about usability problems that you observe, not about survey ratings.

For numbers: Nielsen wrote in 2012 to test at least 20 users to get statistically significant numbers. In 2021, Raluca Budiu and Kate Moran of the same group recommended 40 participants for most quantitative studies, which gives a 15% margin of error at a 95% confidence level. In games, Microsoft Game Studios described its playtest method in 2005: a sample of 25 to 35 participants, based on power analyses of its statistical tests, so that it can compare one game or one build with another.

Plan for testers who do not answer. Not every tester who gets a build will play, and not every player fills in the form. Our guide to in-game feedback surveys lists the response rates that developers published.

- [In-game feedback surveys: join survey answers to player data](https://manalens.dev/blog/in-game-feedback-survey-player-data)
- [itch.io for playtesting: restricted pages and keys](https://manalens.dev/blog/itch-io-playtesting)

*How many testers, by what you want to learn*

| What you want | Testers | Source |
| --- | --- | --- |
| Find the main usability problems | 5 for each round, in several rounds | Jakob Nielsen, 2000 |
| Statistically significant numbers | At least 20 | Jakob Nielsen, 2012 |
| Ratings to compare builds or games | 25 to 35 | Davis, Steury and Pagulayan, Microsoft Game Studios, 2005 |
| A rate with a 15% margin of error at 95% confidence | 40 | Budiu and Moran, 2021 |

Sources: Jakob Nielsen, Nielsen Norman Group (18 March 2000); Jakob Nielsen, Nielsen Norman Group (3 June 2012); Raluca Budiu and Kate Moran, Nielsen Norman Group (25 July 2021); Davis, Steury and Pagulayan, Game Studies (October 2005).

## How do I read the answers?

Look for patterns, and trust what testers did more than what they said. One loud answer is a story. The same moment named by several testers is a finding.

Behaviour first. Jakob Nielsen gave the rule in 2001: pay attention to what users do, not what they say. He found a correlation of only 0.44 between the measured performance of users and the design they said they preferred. Mike Ambinder of Valve wrote the same about watching playtests: the importance is in what people do, not in what they say.

A rating needs a comparison. The Microsoft Game Studios authors give an example: if 70% of the testers liked the character animations, is that good or bad? You can only tell when you compare it with another game or an earlier build. So keep the questions fixed and compare builds.

- Clean the data first. Bromley advises to remove testers who clicked through as fast as they could or gave every question the same rating.
- Split the answers by the first question. Answers from people who play games like yours every week count more than answers from people who never do.
- Group the open answers by moment or topic, and count how many testers name each one.
- Read each rating next to its open question. The rating says how many testers, and the open answer says why.
- Check each claim against the play data of the same build. If testers say the game is too hard, look at the fail rates.
- Write down one decision for each finding. A question that leads to no decision should leave the next survey.

- [Playtest analytics: what to measure](https://manalens.dev/blog/playtest-analytics)

Sources: Jakob Nielsen, Nielsen Norman Group (4 August 2001); Mike Ambinder, Valve's Approach to Playtesting (GDC 2009); Davis, Steury and Pagulayan, Game Studies (October 2005); Steve Bromley, Games User Research (12 January 2022).

## What can survey answers tell me that analytics data cannot?

A survey tells you why and how it felt. Analytics data tells you how many and where. You need both, and you get the most from them when you can put the answers of a tester next to the play data of the same tester.

The join needs one ID in both places. Give each tester a short code, not a name or an email. In the build, call identify() with that code as the user ID. The SDK then adds the user ID to the events that follow. In the survey, ask for the same code, or put it in the link: Google Forms can make a link with pre-filled answers, and Tally can read a hidden field from the form URL.

In Manalens, the Users page has a search by user ID. The profile of a tester shows the history of their events, so you can read what a tester did before a low rating. You can also send each answer as an event from the game with track(), or open a form on your own website with link_url(). Our guide to in-game feedback surveys shows both ways.

Tell testers about the join. Say in the consent line that the tester ID links the answers to the play data. If your playtest needs consent before tracking, start the SDK in consent mode.

- [In-game feedback surveys: join survey answers to player data](https://manalens.dev/blog/in-game-feedback-survey-player-data)
- [Game tracking plan generator](https://manalens.dev/tools/game-tracking-plan-generator)
- [Privacy friendly game analytics and GDPR](https://manalens.dev/blog/privacy-friendly-game-analytics-gdpr)

*What each source can tell you about a playtest*

| Question | Survey answers | Analytics data |
| --- | --- | --- |
| Where do testers stop? | Only what they remember | Yes: the funnel of the first session |
| Why did they stop? | Yes: the open questions | No, only what they did before |
| Is the game too hard? | How hard it felt | Fail rates and retries for each level |
| How long did they play? | A rough guess | Yes: session length |
| Did the build crash? | What the tester noticed | Yes: error and crash events for each build |
| What did they like most? | Yes | No |
| Did they come back? | What they report a day later | Yes: sessions on later days |
| Do they want the game? | Past purchases and actions they report | Clicks on your Wishlist button |

``` Godot: one tester ID in the build and in the survey
# Godot: use the code from the invite as the user ID of this tester.
# A code, not a name or an email.
Manalens.identify("tester_017", {"playtest_wave": "wave_1"})

# Later events of this tester carry the same user ID.
Manalens.track("level_complete", {"level": 3, "attempt": 2})
```

Sources: Manalens Godot SDK, identify users and Users documentation; Google Docs Editors Help (read 19 November 2026); Tally Help (read 19 November 2026).

## What should the invite and the thank-you message say?

The invite says what the tester gets, how long it takes and what you will ask. The thank-you message says what you changed. Both are short. Change the parts in square brackets.

Our practice: tell testers in the invite that there is a survey and how long it is, so that the form is not a surprise at the end. Put the survey link in the game, on the last screen or on the main menu. Valve notes that some developers put a feedback button on the main menu of the game for a Steam Playtest.

Send the thank-you message one day later, with your two follow-up questions in it. Tell testers one thing that you will change because of the test. Testers who see a result are more willing to test the next build. That last sentence is our experience, not a research result.

Not sure if you need a closed test or a public demo? The Steam Playtest or demo tool gives a suggestion from seven questions. The guide on how to market an indie game shows where a playtest and a demo fit in the plan.

- [Steam Playtest or demo?](https://manalens.dev/tools/steam-playtest-or-demo)
- [How to market an indie game](https://manalens.dev/blog/how-to-market-an-indie-game)
- [What is a Steam Playtest?](https://manalens.dev/glossary/steam-playtest)
- [Do Steam demos increase sales? Demo to wishlist data](https://manalens.dev/blog/do-steam-demos-increase-sales)
- [Test your game's hook before you build: ads and pages](https://manalens.dev/blog/test-game-hook-before-you-build)
- [How to run a Steam Playtest](https://manalens.dev/blog/how-to-run-a-steam-playtest)

``` Templates: the invite and the thank-you message
INVITE

Subject: Playtest [game name]: 30 minutes of play, 5 minutes of questions

Hi [name],

You signed up to test [game name]. Your build is ready.

What to do:
1. Download the build here: [link]. Your tester ID is [tester ID].
2. Play for about [30] minutes, at any time before [date].
3. Answer [10] questions at the end. It takes about [5] minutes.

The build is not finished. You will see bugs and missing art.
We test the game, not you: there are no wrong answers.
We collect anonymous play data in this build, and we link it
to your survey answers with your tester ID. We do not ask for your name.

Thank you,
[your name], [studio]

THANK YOU (one day later)

Subject: Thank you for testing [game name]

Hi [name],

Thank you for playing [game name]. [Number] people tested this build.
The first thing we will change because of your answers: [one change].

Two last questions (1 minute): [link]

We will write to you when the next build is ready.
Reply with "stop" if you do not want more emails.

[your name], [studio]
```

Sources: Steamworks documentation: Steam Playtest (Valve, read 19 November 2026).

## How do I build my survey and see what testers did?

Manalens is the analytics and growth engine built for games. Build the survey with the free playtest survey builder, and put the same tester ID in the build with identify(). Then you can read each answer next to the first session funnel, the fail rates and the returns of the same tester and the same build.

- [Playtest survey builder](https://manalens.dev/tools/playtest-survey-builder)

Sources: Manalens product documentation.

## Frequently asked questions

### What questions should I ask after a playtest?

Ask 11 or fewer: how often the tester plays games like yours, an overall rating, the best and the worst moment, a rating for clarity, difficulty and controls, what broke, why they stopped, and one open question at the end. Ask right after the session. The playtest survey builder gives you the list as a form.

### How long should a playtest survey be?

Our practice is 6 to 12 questions, which takes about 5 to 8 minutes by our estimate. After a session under 15 minutes, use 8 questions or fewer. Steve Bromley, a games user researcher, also advises to keep the survey short. Each question must lead to a decision, or it must go.

### Is there a playtest survey template for Google Forms?

Our playtest survey builder gives you the questions as plain text, Markdown and a CSV file. It does not create the form. Paste the questions into Google Forms or Tally: use a linear scale from 1 to 5 for ratings, multiple choice for single answers, checkboxes for multiple answers and a paragraph field for open text.

### Should I ask playtesters if they would buy the game?

No. Steve Bromley writes that the question asks players to predict their future behaviour, and that this is not reliable data. Ask about the past: which similar game the tester bought most recently and why. One day later, ask what the tester did, such as a wishlist add or a second session.

### How many playtesters do I need for a survey?

To find problems, 5 testers for each round: Jakob Nielsen found that 5 users find about 85% of usability problems. For ratings that you compare between builds, use more. Nielsen advises at least 20 users for numbers, and Microsoft Game Studios used 25 to 35 participants for each playtest in its 2005 method.

### Can I use Steam to collect playtest feedback?

Steam Playtest gives testers access to your build, but it has no survey. The Steamworks documentation says that collecting feedback is up to you. Valve notes that a number of developers put a feedback button on the main menu of the game. It can lead to a Steam forum, a Discord channel, a bug tracker or your own website.

## Sources

- [Steve Bromley, Games User Research: How to write a playtest survey (12 January 2022)](https://gamesuserresearch.com/how-to-write-a-playtest-survey/)
- [Steve Bromley, Games User Research: Discover if players will buy your game (1 May 2023)](https://gamesuserresearch.com/discover-if-players-will-buy-your-game/)
- [Alissa McAloon, Game Developer: 5 questions you should be asking playtesters to get meaningful feedback (10 October 2016)](https://www.gamedeveloper.com/design/5-questions-you-should-be-asking-playtesters-to-get-meaningful-feedback)
- [Mike Ambinder, Valve: Valve's Approach to Playtesting, slides of a talk at the Game Developers Conference (GDC 2009)](https://cdn.fastly.steamstatic.com/apps/valve/2009/GDC2009_ValvesApproachToPlaytesting.pdf)
- [John P. Davis, Keith Steury and Randy Pagulayan, Game Studies 5(1): A survey method for assessing perceptions of a game: The consumer playtest in game design (October 2005)](https://gamestudies.org/0501/davis_steury_pagulayan/)
- [Jakob Nielsen, Nielsen Norman Group: Why You Only Need to Test with 5 Users (18 March 2000)](https://www.nngroup.com/articles/why-you-only-need-to-test-with-5-users/)
- [Jakob Nielsen, Nielsen Norman Group: How Many Test Users in a Usability Study? (3 June 2012)](https://www.nngroup.com/articles/how-many-test-users/)
- [Raluca Budiu and Kate Moran, Nielsen Norman Group: How Many Participants for Quantitative Usability Studies: A Summary of Sample-Size Recommendations (25 July 2021)](https://www.nngroup.com/articles/summary-quant-sample-sizes/)
- [Jakob Nielsen, Nielsen Norman Group: First Rule of Usability? Don't Listen to Users (4 August 2001)](https://www.nngroup.com/articles/first-rule-of-usability-dont-listen-to-users/)
- [Amy Schade, Nielsen Norman Group: Avoid Leading Questions to Get Better Insights from Participants (17 December 2017)](https://www.nngroup.com/articles/leading-questions/)
- [Maria Rosala, Nielsen Norman Group: Rating Scales in UX Research: Likert or Semantic Differential? (7 June 2020)](https://www.nngroup.com/articles/rating-scales/)
- [Therese Fessenden, Nielsen Norman Group: Net Promoter Score: What a Customer-Relations Metric Can Tell You About Your User Experience (21 June 2024)](https://www.nngroup.com/articles/nps-ux/)
- [Jeff Sauro, MeasuringU: Should You Use 5 Or 7 Point Scales? (25 August 2010)](https://measuringu.com/scale-points/)
- [W.A. IJsselsteijn, Y.A.W. de Kort and K. Poels, Technische Universiteit Eindhoven: The Game Experience Questionnaire (2013)](https://research.tue.nl/en/publications/the-game-experience-questionnaire/)
- [Effie L.-C. Law, Florian Brühlmann and Elisa D. Mekler, CHI PLAY 2018: Systematic Review and Validation of the Game Experience Questionnaire (GEQ) (23 October 2018)](https://dl.acm.org/doi/10.1145/3242671.3242683)
- [Vero Vanden Abeele, Katta Spiel, Lennart E. Nacke, Daniel Johnson and Kathrin Gerling, International Journal of Human-Computer Studies: Development and validation of the player experience inventory (online 19 October 2019)](https://doi.org/10.1016/j.ijhcs.2019.102370)
- [PXI Bench: User Guide of the Player Experience Inventory (read 19 November 2026)](https://playerexperienceinventory.org/docs)
- [Florian Brühlmann and Elisa D. Mekler: Surveys in Games User Research, chapter in Games User Research, Oxford University Press, pages 141 to 162 (2018)](https://irf.fhnw.ch/entities/publication/abb2533e-4bef-4216-b7cf-7b06a6a9afa1)
- [Steamworks documentation: Steam Playtest (Valve, read 19 November 2026)](https://partner.steamgames.com/doc/features/playtest)
- [Wikipedia: Van Westendorp's Price Sensitivity Meter (read 19 November 2026)](https://en.wikipedia.org/wiki/Van_Westendorp%27s_Price_Sensitivity_Meter)
- [Google Docs Editors Help: Publish and share your form with responders (read 19 November 2026)](https://support.google.com/docs/answer/2839588)
- [Tally Help: Hidden fields (read 19 November 2026)](https://tally.so/help/hidden-fields)

## Steam data

- [RPG games on Steam: benchmarks](https://manalens.dev/steam/genres/rpg)
- [Roguelike games on Steam: benchmarks](https://manalens.dev/steam/genres/roguelike)

## Related guides

- [Game marketing measurement guide](https://manalens.dev/blog/game-marketing-measurement-guide)
- [Playtest analytics: what to measure](https://manalens.dev/blog/playtest-analytics)
- [In-game feedback surveys: join survey answers to player data](https://manalens.dev/blog/in-game-feedback-survey-player-data)
- [Playtest to wishlist: track which playtesters wishlist and buy](https://manalens.dev/blog/steam-playtest-to-wishlist)
- [itch.io for playtesting: restricted pages and keys](https://manalens.dev/blog/itch-io-playtesting)
- [Playtest survey builder](https://manalens.dev/tools/playtest-survey-builder)
- [Steam Playtest or demo?](https://manalens.dev/tools/steam-playtest-or-demo)
- [Game tracking plan generator](https://manalens.dev/tools/game-tracking-plan-generator)
- [What is demo conversion rate?](https://manalens.dev/glossary/demo-conversion-rate)
- [Player analytics guides](https://manalens.dev/blog/topic/player-analytics)
- [How to run a Steam Playtest](https://manalens.dev/blog/how-to-run-a-steam-playtest)
- [Best playtest and feedback tools for indie games](https://manalens.dev/blog/best-playtest-tools-indie-games)
- [Build vs buy a game analytics pipeline](https://manalens.dev/blog/build-vs-buy-game-analytics-pipeline)
- [Best analytics tools for three.js games in 2026](https://manalens.dev/blog/best-threejs-analytics-tools)
