How Long Should a Steam Demo Be? Data and Benchmarks

Design your Steam demo to hold a player for 30 to 90 minutes, and expect the median player to quit after about 14 minutes anyway. Those two numbers sound contradictory, but they describe the same demo: a slice deep enough that your most engaged players can sink an hour into it, built so the hook lands inside the first 10 minutes for everyone else.

This post is about the length question specifically: what the playtime data shows, how played time differs from designed time, whether to end on the hook or hand over the full loop, and how genre changes the answer. Whether you should ship a demo at all, and what it does to wishlists and sales, is covered in our data companion on whether Steam demos increase sales - read that one first if you have not committed to a demo yet.

One provenance note up front. The brief version of this topic that circulates (“Valve says demos should be X minutes”) is wrong. Valve has never published aggregate demo playtime statistics, for Next Fest or otherwise. The official Steamworks demo documentation gives exactly one sentence on length: “You need to balance giving the customer enough content to get them excited, without giving away so much that they feel like they’ve experienced everything the game has to offer.” Every number below comes from community-collected data, mostly Chris Zukowski’s developer surveys, and I will flag which is which.

If you just want targets, here they are:

The rest of the post is the reasoning behind each row, because the targets only make sense once you see how differently the median player and your best player treat the same build.

The best public dataset on demo playtime is Zukowski’s median playtime benchmark, built from developer surveys around the 2022 Next Fests plus a general survey of released games. Self-reported data, so treat it as directional. Across roughly 130 games:

• •Median playtime across all demos: 14 minutes
• •Average playtime: 19.75 minutes
• •Bottom 30th percentile: 10 minutes
• •Top 70th percentile: 22 minutes

The interesting part is what happens when you split those games by how much money they eventually made. Zukowski groups games into revenue tiers, and median demo playtime climbs with every tier:

Games that went on to earn almost nothing had demos people abandoned in 7 minutes. Games that became hits had demos with median playtimes of 38 to 65 minutes. Correlation runs both directions here - good games hold players longer, and demos that hold players generate the wishlists that fund a good launch - but the practical reading is the same either way: a demo that loses the median player in under 10 minutes is a warning light, not a length choice.

One more split from the same dataset matters for the genre section later: linear, story-driven demos had a median of 14 minutes while replayable demos (roguelikes, sims, anything loop-based) held players for 18.5 minutes. Replayability buys you playtime without you building more content.

Here is the finding that should change how you build the demo. In Zukowski’s survey, most developers said they designed their demo as a roughly 30-minute experience. The measured medians came in at half that or less: 14 minutes overall, 22 minutes even at the 70th percentile.

The median player does not finish your demo. They were never going to. They play a demo the way you skim a sample chapter: long enough to decide, not long enough to complete. So “how long should my demo be” is really two questions with two different answers:

1. 01How long until the player knows if they want this game? That has to happen fast - inside 10 minutes, ideally inside 5. This is the part of the demo the median player actually experiences, and it is where your effort goes.
2. 02How much content is behind that moment? This is for the minority who fall in hard, the streamers who need an hour of material, and the Next Fest browser who comes back for a second session. This is where the 30-90 minutes of designed content lives.

You can watch this split in your own numbers. Steamworks gives every demo a “median time played” stat, and Mark Brown’s writeup of the Word Play demo is a good worked example: he recommends aiming for a median between 30 and 90 minutes, and Word Play landed at almost exactly 1 hour of median playtime. That is an exceptional number - his demo is a replayable roguelike-adjacent word game, the genre best suited to long medians - but the measurement habit generalizes: ship, check the stat, adjust.

Next Fest is where demo length gets stress-tested, because the fest algorithm feeds on engagement. The clearest public playtime-vs-wishlists numbers are in Zukowski’s October 2022 Next Fest survey. The median demo in that survey held players for 18 minutes. The fest’s biggest wishlist earners sat far above it: Soulstone Survivors posted a 3-hour median, Manor Lords 1 hour 20 minutes.

The pattern held with exceptions. Capital Command, a hardcore space sim in the same survey, earned 22,862 wishlists off a 10-minute median playtime - niche audiences sometimes decide fast and wishlist on concept. But as a default, the demos that won the fest were the ones players stayed inside.

Playtime is still the middle of the funnel, not the top. In the February 2026 fest survey covered in our demo sales data post, pre-fest wishlist count was the strongest predictor of fest performance (r = 0.825), and demos launched months early earned about 2.5x the median wishlists of fest-week shadow-drops. A long median does not replace showing up with an audience. If the fest is on your calendar, the Next Fest checklist sequences all of this. Note the demo-due date for the October 2026 fest was September 14, nine days before this post went up, so for most readers this is planning material for the February 2027 fest.

The classic advice is to cut the demo at a cliffhanger so players buy to see what happens next. The data above suggests most of that debate is academic: the median player leaves before your carefully placed cliffhanger. But for the engaged minority who do reach the end, the choice matters, and it splits by game structure.

End on the hook when your game is linear. A story game, a level-based platformer, a horror game with a scripted arc: give a complete emotional beat, then cut at the moment the next thing becomes visible. The demo should feel like the first act, not a random slice. What you are selling is “I need to know what happens,” and that only works if the cut lands on rising action. Cut just after a reveal, not just before one - players punish demos that feel like a tease with nothing delivered.

Give the full loop when your game is systems-driven. A roguelike, deckbuilder, factory game, or colony sim should hand over one honest, complete loop - a full run, a full first map - and gate depth instead: fewer characters, fewer perks, one biome, a progression cap. The Word Play demo shipped 25 of the full game’s 120 perks and one game mode. Players could finish a run and immediately see the shape of everything they were not getting. That is the model: complete loop, visibly incomplete breadth. Locked UI elements labeled “available in the full release” are not tacky here; they are the sales pitch.

The failure mode to avoid in both cases is the demo that satisfies. If a player can reach “I have experienced this game” - the exact outcome Valve’s documentation warns about - the demo consumed a sale. Short narrative games are the highest-risk category, which is why the demo sales post lists them as one of the few cases where skipping the demo is defensible.

Genre changes both the right designed length and the median playtime you should expect. A 20-minute roguelike demo can be complete; a 20-minute RPG demo barely clears character creation. These targets are my synthesis of the benchmarks above plus observed Next Fest demos - working estimates, not measured medians:

Two patterns run through that table. Replayable genres ship short loops and let repetition create length; linear genres ship one longer, curated slice and accept a lower median. If your genre sits on the linear side, do not chase the 60-minute medians that roguelikes post - Zukowski’s linear-game median was 14 minutes, and that included successful games. Judge your demo against its own genre’s ceiling. You can see how top pages in your genre handle this on our genre optimization pages.

Almost every demo that misses these targets misses long, not short. The instinct is to include everything you are proud of. Resist it, for three practical reasons.

First, the refund window. Steam’s refund policy covers anything under 2 hours of playtime, so a demo that delivers most of a short game’s experience stacks with that window to make buying optional.

Second, maintenance. Since Valve’s July 2024 demo update, demos can have their own store pages, and a demo with a store page collects its own review score. A big demo is a big surface for bugs, and demo reviews are permanent social proof attached to your launch. If the build is not review-ready, run a playtest instead - the playtest vs demo comparison is the decision guide, and the short version is: rough build, playtest; polished slice, demo.

Third, iteration speed. A 30-minute slice can be re-cut after you see the drop-off data. A 3-hour slice cannot.

Cutting honestly means the demo still represents the real game. Do not strip the systems that make your game distinctive to hit a time target; cut breadth (content count, progression length) rather than depth (mechanics). And say what you cut. A line on the demo page or an end-of-demo screen listing what the full game adds turns missing content from a disappointment into a pitch. Zukowski’s argument for keeping your demo up permanently rests on the same logic: a demo scoped this way keeps selling the full game indefinitely instead of substituting for it.

You will not get the length right on the first build, and you do not have to. The loop:

1. 01Ship the demo early - months before any fest, per the 2.5x timing data.
2. 02Check median time played in Steamworks weekly.
3. 03Add lightweight analytics for checkpoint reach rates (tutorial done, first boss, demo end). Steam will not give you drop-off points; instrument them yourself.
4. 04If the median is under 10 minutes, the problem is almost never total length. Look at the first 5 minutes: tutorial friction, slow opening, hook buried behind setup.
5. 05If players finish fast and playtime clusters at the end-screen, add depth for the engaged tail: an extra difficulty, a score chase, one more unlock.
6. 06Re-cut and update. Demo updates are low-stakes; nobody paid.

Run your demo’s store page through the Steam Page Analyzer while you are at it - the page converts the players your demo impresses, and a strong demo pointing at a weak page leaks wishlists. Our demo best practices guide covers the store-page half of the equation in detail.

Collapsing all of it into something you can act on this week:

• •Build 30-90 minutes of content, toward the low end for replayable genres, the high end for RPGs and strategy.
• •Put the hook inside the first 10 minutes. The median player gives you about 14; spend them on the thing only your game does.
• •Linear game: end on the hook, one complete beat, cut after a reveal. Systems game: give one full loop, gate breadth, label the locks.
• •Expect a median playtime of 14-20 minutes, per the survey benchmarks above, and treat under 10 as a fix-the-opening problem, not a make-it-longer problem.
• •Never let the demo satisfy. If a tester says “I feel like I’ve played it,” cut deeper.
• •Ship early, measure, re-cut. The Steamworks median stat plus your own checkpoint analytics are the whole feedback loop, and the Next Fest prep tool will keep the deadlines straight while you iterate.

The length question has a boring center: 30-90 minutes designed, hook up front, scope cut by genre. What separates demos is not the runtime on the tin. It is whether the first 10 minutes make a stranger add the wishlist.