For a usability playtest on a mobile game, you need five participants per round, for each distinct type of player you are testing. If you need to compare two versions and trust which one players prefer, start with about 40.

Why five

The number five comes from research by Jakob Nielsen and Tom Landauer, published by Nielsen Norman Group in 2000. Across the projects they studied, a single participant found about 31% of the usability problems in an interface. Five participants together found about 85%. After that, each new participant mostly ran into problems the earlier ones had already found.

So Nielsen Norman Group recommends spending the budget for 15 participants on three rounds of five. The team fixes what the first round finds, and the second round checks those fixes and turns up problems that were harder to reach before. The third round repeats that.

For a first-session study of a game, six to eight participants per player group is a common choice, because it leaves room for a no-show or a broken recording and still lets you see whether a problem repeats. The first study we run with a studio at Emhance uses six. When four of the six get stuck on the same screen, that is worth fixing, while a problem only one person hits may be particular to them. See how Emhance playtesting works.

What changes the number

Usability versus preference

A usability study asks where players get stuck, confused, or lost. It is qualitative. You are looking for problems, and you do not need statistics to act on four of six participants failing to find the in-game shop.

A preference study asks which of two versions players like more, understand faster, or would choose. It is quantitative. With five participants, a three-to-two split means very little. For quantitative usability studies, Nielsen Norman Group recommends 40 participants in most cases, which measures a metric such as task success to within a 15% margin of error at 95% confidence. With around 30, you get the same margin at 90% confidence. NN/g says going to 20 or fewer is generally a lot riskier. Treat 40 as a starting point for a preference study: if the two versions turn out to be close, you will need more participants to tell them apart.

Moderated versus unmoderated

Moderation changes what you get from each participant more than how many you need. A moderated session, with a researcher asking questions as the participant plays, gives you more explanation per session. An unmoderated session, with the participant playing alone at home, gives you behavior closer to how they would really play.

For a qualitative study, five participants per round is the starting point in both cases. For an unmoderated study it is worth recruiting one or two extra people, since remote participants sometimes drop out, can't install the build, or have camera problems. Moderated or unmoderated: choosing for games compares the two in detail.

More than one type of player

Five participants covers one group of similar players. If your game serves groups that play very differently, such as new and experienced players, or casual and competitive players, each group needs its own participants. Nielsen's guidance for this case is three to four participants per group when testing two groups, and three per group when testing three or more.

Recruit each group against the criteria you would use for a paid UA campaign: genre familiarity, platform, device tier, and region.

A simple rule

• Finding problems in a flow, a tutorial, or a screen: five per round, per player group, and run more than one round.

• First-session study of a new game: six to eight per player group.

• Choosing between versions: about 40 to start, and more if the versions are close.

Our Beginner's Guide to FTUE Playtesting covers recruitment and how to tell a real pattern from noise in a first study.

For the wider context, see game user research: what it is and what teams use it for and playtesting methods: which one answers which question.

Book a demo to plan the sample for a study on your own game.