A game shaped around where it lived
Facebook Instant Games got rid of the install. You could come across a game inside Facebook, open it straight away and go back to the conversation you came from. That removed a lot of friction, but it also changed what counted as a good idea.
Know Your Friends fitted that setting exactly. Nobody needed a tutorial. Your friends were both the subject of the game and the reason to pass it on, so people understood the premise before the first screen had finished loading.
That mattered more than using every feature the platform offered. If a game travels through conversations, it has to produce something worth taking back to the conversation. Otherwise the share button turns into a prompt people learn to ignore.
A share button is not a growth strategy
It is easy to describe growth as a share button doing its job. That skips the hard part. A social loop only survives if the person receiving it understands why they were included and gets something enjoyable before being asked to pass it on.
Know Your Friends produced a personal question with an obvious audience. The answer meant more because it came from someone you knew, and it gave you a reason to compare, react or play again. The thing being shared and the thing being enjoyed were the same thing.
So design the invitation and the landing screen together. The recipient should recognise the sender, understand the game and reach something fun quickly. Every generic screen in between spends goodwill that the sender just gave you.
“Instant” is a performance target, not a label
A game with no install still has a loading cost. It just gets paid after the click, at the exact moment the player is least invested. Big downloads, a progress bar that tells you nothing and a pile of startup requests all work against the one thing the platform promised.
For an HTML5 game that means watching asset size, the number of requests, the order things happen in and how much has to finish before the player can do anything at all. The number worth tracking is not page load. It is how long until something happens that the player recognises as playing.
At a few million players a day, a small startup improvement lands on an enormous number of sessions, and so does a small regression. Averages hide a lot here, so it is worth splitting the numbers by device, connection, version and language. A healthy average can easily contain an unusable experience for an entire country.
What four million players a day changes
At that volume, ordinary application behaviour becomes sustained load. A harmless extra request becomes millions of requests. A rare retry bug becomes a queue. A one-in-a-hundred edge case stops being an edge case.
The work becomes keeping the path the player waits on as short as possible, making repeated operations safe to repeat, pushing everything else into the background, and measuring the whole thing from the player’s side rather than the server’s. Capacity is only half of it. The game also has to behave sensibly when a platform API or one of your own services slows down.
Releases change too. A new version reaches a huge and varied audience within minutes, so you want to roll out gradually and see the effect quickly. The safest system is not the one that never changes. It is the one where you find out fast when a change was wrong.
The player and daily-user figures here come from the game’s own analytics. Third-party trackers put it at or near the top of the Instant Games charts at the time, but their audience numbers were estimates.
Twenty-seven languages, because that is where the players were
Social games do not grow one anonymous user at a time. They grow through groups of people who talk to each other in the same language. Translating the menus would not have been enough for a game whose whole subject was how well people know each other.
Twenty-seven languages is more than a string file. Text lengths change and break layouts, plural rules differ, names and alphabets break assumptions you did not know you had made, and a prompt that sounds playful in English can sound strange or rude translated literally. The review question is not “is this correct?” It is “would you actually send this to a friend?”
It is also worth measuring. If one language has a much worse completion rate than the others, that usually points at a bad translation, a missing font or a joke that did not travel. Treating every language as part of the product is more work, but it is what turns translation into distribution.
The part that stings: the audience was never ours
Know Your Friends is gone. That does not undo what it did, but it is a reminder that anything built inside someone else’s platform inherits their policy changes, their technical direction and their priorities.
A platform can hand you reach that would otherwise take years. It can also change the review rules, how games get discovered, what notifications you may send, how you make money, or the APIs your growth depends on. It is worth knowing which parts of the relationship with your players you actually own and which can disappear without your agreement.
The answer is not to avoid platforms. It is to go in with your eyes open: keep what you learn, know where your players came from, keep the parts you can take with you, and do not mistake a channel for a market. Know Your Friends shows both halves of that bargain better than any smaller project I have worked on.
What still applies, years later
The platform is gone and the game with it, but the product lessons have not aged.
- Take the friction out before you ask for commitment.
- Make sharing worth something to the person receiving it, not just the sender.
- Get the first good moment right before you add any depth.
- Translate the idea, not only the interface.
- Treat fast growth as an operational problem rather than a celebration.
Most of all, build around something people are already doing together. We never tried to manufacture a social network. We gave one that already existed a small, obvious thing to do, which is how a single question ended up travelling through 130 million people.
