Which games should I install in my internet cafe?

Three weeks out. The machines are on the desks, the disks are empty and there is a browser tab open on somebody's list of the fifty most played games with a growing suspicion that you are meant to install all of them.

You are not. That list is also the wrong list.

The mistake underneath the panic is treating the library as a taste question, as if you were programming a film festival. It is not a taste question. It is a budget allocation problem. You have four scarce things to spend and none of them is opinion: disk space, license and account money, the staff hours it takes to keep a title patched and launching, plus the attention a customer gives your menu before they give up and open the thing they already know. The last one runs out fastest and nobody counts it.

Two things I have to admit before you install anything

First, on day one you have no data. None. Your first library is a guess and anyone handing you a definitive list is guessing about your neighbourhood, your age mix and your prices while calling it research. The point of the first library is not to be right. It is to be wrong in a way you can measure and correct inside three months.

Second, installing a game is not the same as anyone playing it. A disk full of titles is just a backup you pay to patch. A good part of what you install on opening week will never be opened by a paying customer. You will keep patching all of it every month regardless. Nobody notices because installs are visible and play time is not.

Both of those are arguments for a smaller first library than instinct suggests.

The three buckets

Every title on your machines belongs to exactly one of these. If you cannot say which bucket a game is in, it should not be installed.

Anchors. The reason somebody walked through the door instead of playing at home. Anchors are never removed, not in a slow month, not to free disk. The test is simple: if this title vanished for a week, would somebody ask at the counter? If the honest answer is no, it is not an anchor.

Rotation slots. A fixed number of positions you re-cut every month. You keep the slot forever and swap the title out of it whenever the numbers say so.

Request slots. Deliberately empty positions your regulars fill. The bucket most owners never build.

How many of each, before you know anything

The ceiling on your library has nothing to do with disk. It is set by a customer scanning a menu at the start of a session. Past the third row of tiles almost nobody scrolls, which on a normal menu is somewhere around thirty titles. What sits below that line is decoration you still have to patch.

So work from slots, not from titles:

Notice the anchor column does not move at all. That is not laziness in the table. Anchors are chosen by genre coverage rather than by room size and there are only so many ways people play together: a team shooter, a MOBA, a battle royale, a football title, a racing title, two co-op games a group can start without practice, a fighting or party game for the seat nobody booked, plus whatever your local crowd treats as the default. A fifty-seat room does not need more anchors than a twelve-seat room. It needs more copies of the same experience running at once, which is a seat and hardware question rather than a library one. If you have not settled that yet, how many PCs to start with is the prior decision.

Rotation slots do scale, because they are a function of staff time. Every rotation slot is a title someone has to install, patch, test and eventually pull.

Filling the rotation slots without copying a chart

Anchors you can reason about from a chair. Rotation is where you need something outside your own head, because the whole point of a rotation slot is that it tracks what people are playing now rather than what you liked last year.

The obvious move is to open a global chart and take the top ten. Do not. A global chart averages a student in Istanbul with a software engineer in Oslo and reports the result as a fact about your room. It hands you titles that fill rooms in countries you do not operate in while missing the regional favourite that owns your Friday night.

The server has a Games catalog page, which comes with a Cloud subscription. It reads Steam charts and storefront data and pulls artwork from SteamGridDB, then sorts the result into rails: "Most played right now", "Trending games" and "New releases". The part that matters here is the Country selector sitting next to Global. Switch it and the top of the list reorders. That gap between your country and the world is the whole argument against copying a chart.

Adding a game from the catalog pulls in cover art, a poster, genres, rating, release year and a description in the client language, then pushes it to the client menu on every connected PC immediately. No restart, no walking the floor with a USB stick at nine on a Friday. A cut that does not involve walking the floor is a cut you will still be doing in month six.

One caveat worth knowing before it surprises you: delisted titles come back from the catalog missing genre, rating and description. The game still works. The tile just looks bare until you fill those in yourself.

Request slots and why you leave them empty

A regular asking you to install something is telling you they intend to come back. That is worth more than the fifteen minutes the install costs.

The rule I would use: two different named customers asking on separate days. One member who commits to a regular slot counts too. Somebody asking once at the counter is a conversation rather than a signal. Loud does not mean many.

Request titles are not permanent. They live in a request slot until the next quarterly review and then they either graduate into rotation on the strength of their play time or they come out. Tell the customer that up front. Nobody minds a game leaving if they knew the terms when it arrived.

What I would not do

Install everything because the disk is big. Storage is the cheapest thing in the room. Patch bandwidth, launcher updates and a menu nobody can read are not.

Buy a copy per machine. For paid titles, shared launcher accounts lent out for the length of a session cover most of what you need, as long as the same account can never be checked out twice at once. That is the difference between carrying four paid titles and forty. The mechanics are in how to share game accounts across cafe PCs. Software sold by concurrent seat is a separate system that counts seats and blocks overuse without handing out credentials. Mixing the two up is a common way to overspend.

Let staff install things at the counter mid-shift. It ends with twelve machines carrying slightly different libraries and a customer who cannot find the game they played last week because they sat at a different PC. Everything goes through the menu or it does not exist.

Delete a title the moment it looks quiet. Use the visibility toggle instead. Every app in the Client Menu carries one. It takes the title off the client desktop without deleting anything. Hide it for a month. If nobody asks, then remove it. If somebody does, you turn it back on and you have learned something for free.

While you are in there: the Hero carousel and the Quick Launch bar decide what gets seen at all. Your anchors belong in Quick Launch. The rotation slot you are testing this month belongs in the Hero carousel, because an untested title buried at position twenty-two is a title you have guaranteed will fail.

The ninety-day follow-up

Everything above is a structure with nothing proven in it yet.

Statistics has an Applications tab that ranks apps by cumulative usage time across every client PC and session, groups usage by category such as games, browsers, communication and productivity, then shows a time-series trend view over the date range you pick. This is your cafe, not an average of somebody else's.

Three months after opening, sit down with it and work through four questions.

1. Rank by usage time. Are your anchors genuinely at the top? Sometimes the one you were most certain about is sitting seventh.
2. Which rotation titles have near-zero hours? Cut them without sentiment. That is what the slot is for.
3. What does the category split say? A room where browsers and communication beat games is not a gaming cafe yet. No game list fixes that.
4. Is anything climbing quietly on the trend view? A title moving up across eight weeks is an anchor candidate before you notice it at the counter.

The other thing worth opening is Cloud Benchmark, which compares your cafe anonymously against other cafes in your country and includes a ranked list of the top items across them. It is the only view you get of rooms other than your own. It also carries the one condition in this article that may not clear for you. A cohort is only built once at least twenty cafes in your country are contributing. In a thinly covered market the tab tells you the cohort is still forming and you are back to your own numbers. I would rather say that here than have you go looking for a page that stays empty.

The short version

Do not pick games. Allocate slots, then run a calendar.

Week one: install the anchors and nothing else. Week four: fill the rotation slots from the catalog with the country filter on. Hand the request slots to whoever asked twice at the counter. Month three: open the Applications tab and cut everything sitting under an hour of recorded play. Then repeat that last cut monthly, on a weekday, before the room fills.

You will be making all of this inside whatever software you already chose to run the room. The practical tests for that choice are in how to choose internet cafe software.