Onboarding funnel for a Roblox game

Track the stepsCopy Anchor Link

Pick the handful of moments every new player should reach, and send one event per step from the server, where the game logic already knows they happened:

local RoChart = require(game.ServerScriptService.RoChart)

local function onStepReached(player, step, props)
  props = props or {}
  props.step = step
  RoChart.track(player, "tutorial_step", props)
end

local function onTutorialDone(player, startedAt)
  RoChart.track(player, "tutorial_completed", { seconds = os.clock() - startedAt })
end

One tutorial_step event with a step property keeps the schema small. Filter on step to use each one as a funnel step.

Build the funnelCopy Anchor Link

Create a funnel over the last 14 days, converting within 1 day:

1. $session_start
2. tutorial_step where step = 1
3. tutorial_step where step = 2
4. tutorial_step where step = 3
5. tutorial_completed
6. upgrade_bought

If the tutorial only runs for new players, the steps after the first are already about new players. To make the first step new players only too, filter the funnel by the New players cohort from the step below.

$session_start 18,240 → tutorial_step 1 15,902 (87.2%) → tutorial_step 2 13,110 (82.4%) → tutorial_step 3 7,288 (55.6%) → tutorial_completed 6,940 (95.2%) → upgrade_bought 3,105 (44.7%)
Overall conversion 17.0%. Biggest drop-off: tutorial_step 2 → tutorial_step 3 (44.4% lost).

Here, nearly half the players who equip their first tool never make a sale. That's the step to fix.

Break it downCopy Anchor Link

Break the funnel down by $device_type (platform). If phones drop much more at one step, the problem is often the UI on a small screen, not the game. Break it down by $join_source to see whether players who followed a friend get through better than players from discovery.

Segments with fewer than 30 players are stippled: treat them as anecdotes.

Make a cohort of new playersCopy Anchor Link

"First played in the last 14 days" is easiest in SQL. Save this in the SQL Editor and create a cohort New players from it:

SELECT person_id
FROM events
WHERE event = '$session_start'
GROUP BY person_id
HAVING min(timestamp) >= now() - INTERVAL 14 DAY

A rule-based cohort works for the rest, such as Completed tutorial: performed tutorial_completed at least once. Cohorts work as filters, breakdowns and experiment segments.

Does onboarding predict retention?Copy Anchor Link

Create a retention insight by day: start with tutorial_completed, return with $session_start. Compare it with the D1 on Game analytics → Retention, which counts every new player. If players who finish the tutorial come back much more often, getting more players through it is worth the work.

Fix it and checkCopy Anchor Link

Ship the fix, and leave an annotation on the day so everyone knows what changed. Next morning, the funnel and today so far on Game analytics already show the first day of new players. If you're not sure the fix is better, A/B test it instead.