- Project A launched with a step-by-step tutorial (steps 1, 2, 3 teaching the controls in order).
- For games with a novel core loop and no genre conventions, "what is this game?" must land before "how do you control it?" — otherwise the drop-off doesn't stop.
- After switching to a concept-first tutorial, 5-minute retention rose +28% and average session length rose +50%.
- There are five tutorial types — procedural, conceptual, contextual, environmental, and remedial. Choose by the game's nature, not by reflex.
Cast
Senior consultant at ZehnStudio26. Around since the early Roblox days. Reads the story behind the data well.
A marketer at a brand company. Has only recently started paying attention to Roblox. Asks "what does that mean?" every time jargon shows up — the reader's voice.
Why “changing the type” alone moved early drop-off
Adding a tutorial reflexively can backfire
“Adding a tutorial is the kind thing to do” — broadly correct in game development. But Project A surfaced a slightly different lesson. The wrong type of tutorial increases drop-off rather than reducing it. A procedural walk-through of controls is supposed to help. In a game with a novel core loop, that same form creates a state where instructions stream in without the player understanding what the game is — and accelerates drop-off.
What was happening in Project A
At launch, Project A used a textbook procedural tutorial. Step 1: move. Step 2: pick up the item. Step 3: head for the goal.
The data showed 5-minute retention at 24.5%. About 13% gone by 30 seconds, ~30% by 1 minute, ~65% by 3 minutes. Most of the drop-off was happening between minute 1 and minute 3 — a textbook “early drop-off” curve.
Why procedural didn’t work
For genres with established conventions — obbies, tycoons, shooters — the player already brings in the core goal: “dodge to the goal,” “build to earn,” “shoot the other player.” Procedural is the right format for “I know what I want to do, just tell me the controls.”
Project A was a novel-core-loop title. No genre convention. The player was being shown control steps before they had any framing for what the game was — back to the cookbook with no dish name.
Rebuilding it concept-first
So we rebuilt the tutorial’s structure itself. The first 30–60 seconds were dedicated to making the concept explicit: this is the game, this is the goal, this is what winning looks like. Control instructions were reordered into small, just-in-time hints surfaced after the player had the core loop in their head.
Outcome: 5-minute retention moved 24.5% → 31.3% (+28%), average session moved 5.0 → 7.5 minutes (+50%). The biggest gain came after the 1-minute mark — 3-minute retention rose +22% as well. The most plausible read: players now had a frame for “what am I doing here?” early, so the first few minutes became easier to tolerate.
Three lessons portable to other titles
1. First, judge whether the genre has conventions. If it does (obby, tycoon, shooter), procedural will work fine. If it’s a novel core loop, lead with conceptual.
2. Validate, on real hardware, that “what is this game?” lands within the first 30–60 seconds. Devs default to “obviously they know” and miss this. Hand the build to first-time players and check whether they can verbally describe the game after the first minute.
3. Read improvement on two axes — 5-minute retention and average session length. If both move in the same direction, it’s a content-side improvement. If only one moves, external factors (ad-traffic quality, for example) might be doing the work.
