Launch & promotion

Launching a Roblox Game in 2026: A Growth Plan to 100 CCU and Beyond

A recognizable game fantasy, a strong first session, a clear reason to return and measurable acquisition form the foundation of a controlled launch. Start with a small audience, fix the largest drop-off and expand channels based on results. Use the practical tactics, 14-day plan, advertising decisions and 100-concurrent-player calculation below.

MARPLA team · Published 2026-09-10 · Updated 2026-10-28 · 20 min read

Full guide

Start with the essentialsRoblox game launch checklist: from preparation to 100 CCU Knowledge sectionLaunch & promotion →

AI-generated thematic illustration: preparing a game world for launch. Not a MARPLA interface or a promise of results.

Practitioner tactics: a clear niche, strong first play and a reason to return

Start with a purple-ocean opportunity: recognizable demand served by an awkward experience. Play several comparable games on a phone and read recurring complaints. Look for one specific gap: waiting too long for the first action, unclear two-player play, or progression that ends after one round. Choose a difference players can notice immediately. A different visual theme alone rarely closes that gap.

Express the fantasy as a verb: steal and defend loot, raise a creature, escape with a friend. Align the image promise, the first player action and the moment shown in a short video. A strong hook reveals a real choice and result. Clarity matters more than visual complexity.

Keeping a familiar foundation while changing part of the format can reduce the risk of an incomprehensible product. Pick one noticeable mechanism: cooperation, unusual stakes or a clear progression path. Test it separately. A 20% novelty rule can help discuss scope; it does not measure originality or guarantee discovery.

Give players an honest reason for their next visit: a creature finishes growing, a new task becomes available, the world changes, or a shared goal progresses. Show what will be ready and when. A timer works when it represents a real game state. A pair-play reward can support social play; check that it does not make a solo start frustrating.

Schedule a shared event when your audience can actually attend. You can test 16:00–21:00 in its local time zone as an initial hypothesis, then compare it with your own history. Concentrating real players can help fill servers and support cooperative play. Evaluate the experience and subsequent return.

Give content creators a strong playable moment, a working build and an honest exclusive to demonstrate. Let them explain it in their own voice. For short videos, test a clear opening hook in the first two seconds and one complete gameplay moment. Choose a sustainable posting pace, compare outcomes and iterate on what works. Check music rights and the rules of the publishing platform.

Tactic	Action	Evidence it helps
Serve a clear niche	Fix one recurring complaint in a familiar format	An unfamiliar tester recognizes the difference and reaches the main action
Make the image a promise	Show a real task, stakes and result	Better acquisition is accompanied by a stronger first session
Create a reason to return	Show the next world change or available goal	Return improves in comparable cohorts, not just notification volume
Coordinate a shared start	Run a genuine event at a suitable time	Servers fill, the core loop works and participants return
Work with content creators	Start with relevant creators, separate links and a shared brief	New players fit the game and the cost of the result is visible

100 CCU: Translate the Target into Capacity and Session Volume

CCU is the number of players in a game at a given moment. Specify the target more precisely: an event peak, average concurrency during an evening, or a daily average. A 100-player peak for ten minutes and a steady 100 CCU all day require different products and acquisition volumes.

Mathematical guide: average CCU over a period = total player-minutes / period length in minutes. With a relatively steady flow, average concurrency can be estimated as sessions started per hour × average session length in minutes / 60. This estimate is unreliable during a sharp peak.

Illustrative calculation: 600 sessions per hour with a 10-minute average duration produce about 100 average CCU. Over a full day at the same rate, that is 14,400 sessions, not 14,400 unique people: one person may return several times. This is capacity arithmetic, not an ad-delivery forecast. Use CCU history to see the growth pattern and launch cohorts to assess its quality.

Choose an Access Mode: Testing and Organic Launch

Check the current publishing requirements and game access in Creator Dashboard. Use private access for internal checks and an appropriate limited mode for selected testers. A soft launch means a small, controlled release, not a separate button.

Experience Beta lets you collect feedback and analytics without normal Home distribution. Roblox staff clarify that you can invite your community and run Sponsored Ads in Beta, but the game is excluded from Recommendations. It is a testing mode, not free insurance for organic growth. Before public launch, also check Exclude from Recommendations and the game’s current availability.

Assign a release owner, a rollback lead, and a review time. The detailed launch operations guide helps document access, roles, and critical stop conditions.

One Week Before Launch: Product, Presentation, and Measurement

Express the game’s promise as one action: grow, rescue, build, escape. See how similar games deliver that action, where players get lost, and how your first loop differs. “Change 20%” is a creative heuristic, not a proven success formula. Use the growing games radar to choose examples based on observable momentum; public concurrency does not reveal their retention or profit.

The title should be recognizable, and the first lines of the description should explain the action and what makes the game different. An accurate genre helps people understand the game; lists of other games’ names and unrelated keywords get in the way. “Must be under 500 characters” is not a universal ranking rule. Check readability in the current interface.

The icon and thumbnail should show the actual game. A clear focal point is a useful design hypothesis; a close-up face and warm palette do not guarantee a CTR lift. Prepare distinct, accurate options and check them at small sizes. Home requires a custom thumbnail. Thumbnail personalization and adaptive delivery are covered in the thumbnail guide.

Check joining, saving, rejoining, death, purchase fulfillment, co-play, and mobile controls. The shop’s price, payment, and reward delivery must work correctly. Choose Game Pass or Developer Product based on the product; having both is not a recommendation requirement. Prepare a clear offer and economy tracking before advertising.

Minimum funnel: join → first useful action → first reward → next self-directed step. Record the version and release time. Roblox funnel events are sent from the published game server; a successful Studio test alone does not confirm that these events are delivered. Use separate funnel sessions for repeatable checkout analysis. See the first-session diagnostic for details.

A 14-Day Plan: Move Forward Based on Results, Not the Calendar

One controlled launch cycle: preparation, audience, measurement and decision — A working cycle: establish safety and measurement, then decide the next step.

Start with a small group that is sufficient to answer a specific question. Friends are useful for finding bugs, but an audience familiar with the game will overestimate how clear it is. Also invite new players from relevant communities, with moderator permission. A sample of “100–300 testers” does not guarantee statistical precision.

Do not launch ads, videos, and a major update at the same time just to create an impressive spike. If a joint event is essential to the game itself, document the combined launch in advance: that observation cannot isolate each channel’s contribution.

Window — example	Task	Condition for next step
T−7…T−3	Check the first loop, devices, saving, events, and images	Critical paths work; a stable rollback version is available
T−3…0	Run a soft launch and observe new players	The main drop-off cause is understood; mature D1 is available or uncertainty is explicitly documented
Days 0–3	Open the selected mode and check real joins and load	No critical harm; sources and version are recorded
Days 3–7	Test one channel and one primary hypothesis within a spending cap	There are enough observations to decide, and the cost of answering the question is within the cap
Days 7–10	Scale a validated channel or run a measurable event with creators	The game can handle the load; each integration has its own link
Days 10–14	Review mature cohorts and fix one weak point	A decision is made—continue, change the test, or stop—and the next review is scheduled

How to Judge Metrics

Review D1, D7, and D30 in the context of similar games and your own history. The 2026 GameAnalytics report found median values of 10.3% D1, 1.6% D7, and 0.5% D30. Its sample includes more than 500 integrated games with 1M+ MAU from August 2025 through July 2026; it does not represent all Roblox games and these are not recommendation minimums. Use the report as external context and compare its definitions with Creator Analytics.

The team’s threshold should answer a decision question. For example: “Do not scale the mobile campaign until we fix player loss at the first-reward button.” This is more useful than a blanket rule like “No marketing if D1 is below 15%.” The funnel, early exits, cost, and source quality help explain what to fix.

Illustrative sample-size example: 15 returning players out of 100 new players gives D1 of 15%; one more return changes the result by a full percentage point. Record the percentage, denominator, and first-session date. A blank D7 for a new cohort means the data is pending, not that retention is zero. Compare equally mature windows using consistent definitions.

Metric	What it helps you decide
CTR / clicks after impressions	Whether the creative gets noticed in the selected placement; do not treat it as a substitute for starts or retention
Home PTR / starts after an impression	Whether Home users proceed to the game; do not confuse it with ad CTR
First-play bounce: under 60 and 61–180 seconds	Where expectations, loading speed, and the first experience diverge; inspect source breakdowns where available
D1, D7, D30	Whether the cohort returns; do not compare a mature week with yesterday’s launch
Play days / playtime	Frequency and depth of play; not the same as the D7 retention rate
CPP, revenue, costs	Cost per acquired play and cohort economics; attribution alone does not prove incrementality

Ad Test: Configure It to Answer a Question

In the current Ads Manager, campaigns automatically appear on Home and Search. Do not rely without checking on older instructions to separately buy ten Search keywords for each account. If the interface differs, compare the available fields with current documentation before adding funds.

To test initial acquisition, consider Plays + New Players, if available. Earnings has limited availability and targets paying players; Engagement is a separate objective for a highly engaged audience, not simply “any retention.” Automatic bidding calculates the bid. Limit country and device when justified by the game’s language or functionality; targeting ages 18–24 does not by itself promise a return.

Record the total loss limit, test duration, and result source. An illustrative budget of $10–20 per day may be too small or too large; what matters more is the cost of obtaining an observation and the team’s spending cap. For example, at an illustrative $0.10 CPP, a $20 budget buys about 200 starts; at $0.50, about 40. These are arithmetic examples, not auction forecasts.

A technical defect or exceeding the agreed cap requires an immediate pause. A weak CTR without enough impressions or immature revenue data does not provide a complete answer. Reported plays and earnings may be delayed by up to 48 hours. See the advertising guide and budget and stop-rules analysis for specific rules on waiting, stopping, and raising the budget.

Roblox — Ads Manager: current objectives, settings, and reports

Why Ads Can Lower the Blended Rate Without a Home Penalty

Roblox’s current FAQ separates ad-acquired players from the organic Home cohort: activity from users first acquired through ads is excluded from RFY ranking signals, including on later returns. Therefore, the claim that “early advertising permanently damages the algorithm” does not follow from the official model. Ads can still spend money on an unsuitable audience and create load.

Illustrative blending example: 100 organic new players with D1 of 20% and 100 ad-acquired players with D1 of 8% produce a blended D1 of 14%. The organic group has not worsened. Analyze ad spend separately. If Home impressions fall after a campaign, also check the available audience, seasonality, competition, and version changes; timing alone does not establish causation.

In My Games, use the available metrics for your connected game. Verify the source in Roblox Acquisition; the traffic mix guide explains sources and limitations. A competitor’s public visits and CCU do not reveal its ad budget, retention, or profit. See the paid and organic attribution article for causal analysis.

When to Stop a Launch or Ads

Separate an emergency stop from a hypothesis review. Do not wait for D7 if data is being lost or purchases are broken. If there is no technical threat, wait for the selected window and check whether there are enough observations. A CCU decline after ads stop only means one traffic stream ended; it does not prove a retention problem.

In MARPLA, add a launch milestone: version, exact time, budget, channel, change, and expected effect. After review, add the observed result and decision. If the needed combined breakdown is unavailable, record that limitation and use the official export; do not infer it from overall averages.

Observation	Action
Widespread save loss, purchases without fulfillment, or inability to join	Stop acquisition, limit the rollout, restore a stable version, and verify the data
The total test cap or maximum agreed CPP has been reached	Pause spending and review the result before spending more
Players click but drop off before the first action	Check the creative’s promise, loading, and interface on the affected device
A new country lowered blended D1	Compare each country’s share and within-country metrics; check language and performance
A new cohort has no D7 yet, or the report is still processing	Record the expected maturity date; do not replace unknown with zero
Players remain online after ads pause	Check Home, Continue, friends, and external links; distinguish returning users from new organic players

Creators and Short Videos: Buy Measurable Learning

Choose a creator based on audience fit, not just follower count. Provide access to a working build, an accurate description of the gameplay, and creative freedom. Create a separate Roblox share link for each creator and platform. A unique Discord invite tracks community joins, but does not prove game starts; UTM tags do not replace Roblox attribution.

Do not treat “5–15 creators in 72 hours” or “2–5% of views convert to game visits” as industry standards. For an initial test, use a number of integrations the team can support and measure separately. Record the post, spend, new joins, returns, and available purchases. Video views, link clicks, and new players have different denominators.

Show an actual gameplay moment: a choice, a risk, a funny outcome, or a cooperative task. Test a distinct video hypothesis; posting frequency alone does not guarantee visits. A shared event launch may help fill a server and organize co-op play, but it is not a secret signal to “turn on Home.” The event should deliver real value to players.

Tie each update to an observed obstacle: “players cannot see the action on mobile” or “the next step is unclear after round one.” Validate a fix before release; daily releases for “freshness” do not replace improvement. Use Events & Updates to announce a player-accessible update, then compare arrivals by version and channel.

Example: A Small Launch for a Co-op Survival Game

Fictional scenario. A team invites new testers and sees that mobile players leave before role selection. It first checks the wording, button size, and loading speed, fixes role selection, and tags the new build. Friends progress faster, so their results are kept separate.

After D1 matures, the team exits Beta and runs one capped initial-acquisition test. The cap and stop rules are written down in advance. Two creators get different share links; their posts are staggered to make each source’s path visible. MARPLA milestones record the fix, the start of advertising, and both videos.

A week later, the team sees three different outcomes: the ad group completes the first loop, one video attracts people with mismatched expectations, and organic Home provides too little data for a confident conclusion. The decision: keep the effective spend within the cap, revise the second creator’s brief, and continue observing Home. There is not yet enough evidence to say “the algorithm picked it up.” Cohort measurement helps structure this review.

Payback: Measure Cohort Economics, Not the Price of Robux

As of the review date, the standard new DevEx rate is $0.0038 per eligible Earned Robux earned on or after September 5, 2025, at 10:00 a.m. PT; the $0.0035 rate applies to older balances. A separate rate of $0.0054 applies from June 8, 2026, to qualifying spend by adult players in the United States. This is not the rate for all Robux and is not a reason to automatically direct the entire budget to US players aged 18+.

For team decisions, compare spend with revenue from the corresponding cohort over a defined window, including fees and other costs. Do not confuse player purchase totals with developer earnings. Account separately for Creator Rewards and other applicable sources: “there can be no revenue without purchases” is too categorical. Ads Manager-attributed revenue is not the same as proven incremental revenue.

Illustrative scenario: if comparable expected cohort revenue is $40, acquisition costs are $50, and other costs are $5, the result before further taxes and adjustments is −$15. Hope for future Home traffic does not erase a known loss. See the DevEx guide for cash-out terms and net proceeds.

How to use numerical targets from practical checklists

Numbers help a team agree on a decision before a test. Apply them to a specific audience, placement, build and mature measurement window. For example, D1 of 12–15%, CTR of 1–3% and early exit of 10–15% can be initial working targets for your own test. They are not official admission thresholds for ads or Home; refine them against comparable history.

D7 of 15–20% is an ambitious product-specific goal, not a required threshold for sustaining organic traffic. A starting group of 100–300 testers, $10–20 daily spend, 5–15 content creators and a 48–72-hour event window can help plan resources. Define the answer you need and the cost you can accept. One successful case is not a promise of the same outcome for every game.

Checklist claim	A sound decision
“Wait for D7 above 15–20%, or ads will penalize Home”	There is no universal D7 threshold. Evaluate cohorts and the acceptable test cost; ad-acquired players are excluded from RFY ranking signals.
“A new game gets a 48–72-hour boost”	The volume and duration of initial impressions are not guaranteed. Plan to validate the product rather than capture a magic window.
“Turn on Beta and wait for organic Home traffic”	Experience Beta excludes the game from recommendations. Check this setting before a regular launch.
“A 3% CTR means you are ready to scale”	Impressions, clicks, and starts are different stages. Strong click performance alongside widespread early exits does not validate acquisition quality.
“CCU sustained after ads proves Home is working”	Ad-acquired players, friends, and video viewers may return. Verify the source in Acquisition.
“Every Robux is worth $0.0035 when cashed out”	The rate depends on the specific Earned Robux; check when it was earned and the DevEx terms.

Boundaries and Frequently Asked Questions

Do not: buy bots or fake popularity, simulate play with extra accounts to inflate metrics, promise content or rewards that do not exist, or disguise an ad as another project. An AFK mechanic in a game described honestly and artificial engagement inflation are different situations. No specific ban duration is promised here; enforcement depends on the violation and Roblox’s decision.

Can I bring friends together for one evening? Yes. A genuine group test or event is useful for a co-op game. Record where that audience came from and do not present it as an independent market sample. Should I urgently change every thumbnail? No. Change them to test a hypothesis and make sure the promise matches the actual game. Large unrelated changes undermine comparisons.

Do I need to wait 28 days for recommendations? No. Roblox uses early and later signals as data becomes available. Are two short sessions always better than one long one? There is no universal rule with a fixed number of minutes. Design a natural reason to return and a useful session, not a forced timer-based exit. When is a game “ready for the charts”? There is no such certification; there is readiness for the next limited step.

What to Do Today

Clarify the target: peak or average CCU, time period, budget, and maximum acceptable loss.
Complete the first loop on mobile and PC; check saving, purchases, and rollback.
Check Beta, access, custom thumbnail, and discovery restrictions in Creator Dashboard.
Set up the funnel, version tagging, and separate share links before acquiring players.
Compare mature cohorts with your own history and similar games; record the denominators.
Run one capped test, record it in MARPLA, and decide using criteria selected in advance.

Continue the Analysis in MARPLA

Open the step-by-step milestone guide to record the launch, then review traffic mix and Home test history to connect changes with the observed audience. Private analytics require signing in and access to your game; missing data does not mean a zero result.

Useful guides: how Home recommendations work, what to do when the audience expands to a new country, how to evaluate the first session, and when to stop ads. Return to the short checklist before the next stage.

Primary sources

Put this into practice in MARPLA

MARPLA tool diagram: Add a milestone and track its result

Mark a launch or update as a milestone, record your hypothesis and compare acquisition before and after. Check traffic sources so an advertising effect is not mistaken for an update effect.

Open the tool →Select your connected game. Sign-in and access to its data are required.

Discussion0

Loading comments…