Roblox Monetization and Live Ops

Target: Roblox Studio (October 2026 build) on macOS, driven through the scenario-roblox-expert toolkit.

Expert level here means every Robux path ends in one server receipt handler that grants only what the saved profile confirms, prices and odds come from one source the engine or a config owns, and every live change ships behind a config with a metric that decides it. The agent builds and tests the code in an unpublished place; product creation, dashboards and store art uploads stay with the owner.

If a sibling skill named here is missing from your available skills, ask the user to install it (npx skills add scenario-labs/skills --skill <name>); unattended, proceed from tool schemas and flag the gap.

REQUIRED BACKGROUND: scenario-roblox-expert (channels, review loop, toolkit, current-Roblox traps).

Stance (the expert delta)

1. Grant only in the receipt handler, and only after the save. PurchaseGranted means the saved profile holds the PurchaseId; anything else answers NotProcessedYet. Prompt-finished events are not proof of payment (Creator Docs, developer products; Stewiepfing). The docs sample answers granted when the handler returns nil; the receipts suite reproduces that bug.
2. One handler per server. New code uses BindReceiptHandler; legacy code assigns ProcessReceipt once, in one Script (MarketplaceService reference). Live, a second catch-all bind errors, while a second ProcessReceipt assignment is accepted without a warning.
3. Prices come from the engine, on the client. Typed prices block price optimization and show Plus subscribers a price they do not pay (Creator Docs, managed pricing, Roblox Plus). The server never needs a price to grant.
4. One odds table for the roll and the disclosure, gated per player. Show every outcome with numerical odds before spend, and check ArePaidRandomItemsRestricted per player (Creator Docs, paid random items). Prompt random items from the server after that check [added].
5. Monetize every intended mechanic at launch, cheaply. Starter bundles at 25 to 99 Robux; many small spenders beat one high price; never gate basics such as movement (Roblox monetization leads and studio heads, RDC 2025). Adding payment later reads as pay to win.
6. Every risky change ships behind a config, and data decides it. Read experiment values on the player snapshot where the player meets the feature, check the minimum detectable effect first, and let guardrails such as disconnects outrank a modest gain (Creator Docs, experiments; Roblox experiments team and studios, RDC 2025).
7. The click gates discovery. Play-through rate and first-play bounce replaced QPTR; ranking learns only from organic Home traffic, and thumbnails must show the real game (Roblox discovery leads, RDC 2026 and RFY Q&A; studios, RDC 2025).
8. Rewarded video pays a fixed developer product. It is user-initiated, disclosed, never Robux, never random, never a progress gate (Creator Docs). Reward obsolete paid items so ads do not cannibalize spenders (RDC 2025).

Establish first

• Published or not: analytics, configs, ads and product lookups need a published experience. Unpublished, test with synthetic receipts and record what Studio does.
• Product inventory: names, kinds (currency, bundle, pass, subscription), what each grants, and whether ids exist. Only the owner creates ids in Creator Hub (gui-paths.md).
• Data layer: which profile store, and its getProfile and isPresent functions for scenario-roblox-data Receipts.
• Audience: age range, regions with paid random restrictions, mobile share, DAU (experiments under about 1,000 DAU rarely reach significance).
• Ads: owner age-verified 13+ with 2FA and a public game, or no rewarded video.
• Defaults when unanswered: one catalog module, BindReceiptHandler through Purchases, prices via PriceText, a currency pack plus a starter bundle, no paid random items, telemetry and live config on from day one.

Procedure

Full code and test results are in procedures.md. Server modules (Catalog, Purchases, Entitlements, Telemetry) go in ServerScriptService.MonKit beside the data layer's Receipts; shared modules (Odds, PriceText, LiveConfig, ShopRank) in ReplicatedStorage.MonKit. Status: 16 of 16 live tests pass in Studio 0.743 on an unpublished place (2026-10-19); 61 suite tests pass in Lune and unchanged in Studio. Needs a published game, not yet run: a real RankProductsAsync ranking (unpublished: HTTP 400, catalog order kept) and no-description on your own items.

1. Audit. Goal: every purchase path, price literal and client-side analytics or config call found. How: rbx_monetization.audit_place(place) offline (Lune), then MonAudit.run(game) through rbx_monetization.run in Edit. GATE: zero errors, and the live tally equals the offline tally.
2. Catalog. Goal: products, passes and subscriptions in one server table, grants as data, no prices. How: Catalog.new(def), catalog:validate(); in a published test server, catalog:verify(fetch, game.CreatorId). GATE: no duplicate-id, bad-grant, unknown-id or foreign-id finding. A wrong pass id is not an error: live, pass id 1 resolved to another creator's pass.
3. Receipts. Goal: exactly one grant per PurchaseId. How: Receipts.new{getProfile, isPresent, handlers = catalog:handlers()} from scenario-roblox-data, then Purchases.new{receipts, catalog, telemetry} and purchases:bind(MarketplaceService). Test with synthetic receipt tables, never a real prompt. GATE: a duplicate grants once, an absent player and a failed save answer NotProcessedYet, and the retry grants once.
4. Perks. Goal: passes, subscriptions and Plus as server-checked perks. How: Entitlements.new{passes, subscriptions, owns, subStatus, onChanged}, ent:connect(MarketplaceService, Players); Entitlements.giveTool writes Backpack and StarterGear. GATE: a check that errors keeps the paid state; the tool survives respawn; the client shows "Owned" from the perk_ attribute.
5. Prices. Goal: every displayed price is the player's price. How: PriceText.fetch and PriceText.view in a LocalScript when the shop opens. GATE: no price-literal finding; rows show "..." and stay unbuyable until the price loads.
6. Paid random items. Goal: disclosed odds, a path for restricted players. How: one Odds.fromWeights module required by the server roll and the client panel; Odds.display behind a "Details" button with the disclaimer; restricted players get Odds.directPrice or no offer. GATE: displayed odds sum to 100% or carry the disclaimer; a restricted player cannot reach the prompt.
7. Rewarded video. Goal: a disclosed, optional ad that pays a fixed product. How: client GetAdAvailabilityNowAsync when the menu opens; server CreateAdRewardFromDevProductId then ShowRewardedVideoAdAsync in pcall; the reward arrives as a receipt with channel AdReward. GATE: the ad row hides when unavailable; the reward path is the receipt path.
8. Analytics. Goal: onboarding, shop funnel and economy events that survive exploiters and the limits. How: Telemetry on the server only; clients send step numbers, the server accepts forward steps only; log economy events after the save. GATE: invalid steps dropped and counted; event names, currencies and fields within the limits table.
9. Live config and experiments. Goal: values change without a release. How: LiveConfig.new{defaults} with UpdateAvailable and Refresh, mirrored to cfg_ player attributes plus a cfgReady marker; forPlayer for experiment keys. GATE: defaults apply when ConfigService fails (live: it fails after about 30 s unpublished); the client reads values only after cfgReady.
10. Store art. Goal: thumbnails, icon and item icons that pass the size rules and show the real game. How: capture a staged playtest view, thumb_from_capture, then Scenario art (see Handoffs); check_art(path, kind). GATE: zero errors from check_art; a contact sheet reviewed; the owner uploads.
11. Launch and live ops. Goal: a plan with metrics that decide each step. How: rbx_monetization.mde(enrolled, baseline) for each test, rollout at 10, 50 then 100%, ads campaigns run 1 to 2 weeks without targeting for a baseline. GATE: each experiment has a primary metric, guardrails, an MDE under the expected effect and a stop date. Launch checklist and content pacing: launch-liveops.md.

Numbers

Value	Relative to	Source
25 to 99 Robux	Starter bundle price	RDC 2025
100 PurchaseIds	Receipt log kept in the profile	scenario-roblox-data default [added]
60,000 transactions in 30 days	Price optimization eligibility; about 3 weeks at a 98/2 split	Docs
4 places below the first non-zero digit	Odds rounding, with the disclaimer	Docs
120 + 20 x CCU per minute	Analytics requests per server	Docs
3 fields, 10 currencies, 20 types, 100 SKUs	Analytics cardinality before "Other"	Docs
10 funnels, 100 steps, 100 event names	Analytics per experience	Docs
14 to 60 days, up to 2 variants	Experiment length and arms	Docs
-10% playtime, -20% ARPU or payer conversion	Early harm stop, first 24 h, 10,000+ enrolled	Docs
About 1,000 DAU	Below this, a useful MDE is hard to reach	Docs
10%, 50%, 100%	Rollout steps for risky features	RDC 2025 studios
Up to 4% of daily earnings	Ads reinvestment; ROAS read over 30 days	RDC 2026
1920x1080, 16:9, under 3 MB	Thumbnail; bottom strip covered by the player count	Docs
512x512, legible at 150x150	Experience icon	Docs
512x512 max, circle crop	Pass and product icons	Docs
5 Robux per Active Spender per day	Creator Rewards, first three games of the day, 10+ minutes	Docs

Quality gates

Measurable:

• Audit: zero errors offline and live; warnings fixed or justified.
• Suites: catalog, receipts, entitlements, pricetext, odds, telemetry, liveconfig and monaudit pass in Lune and, unchanged, in Studio with the same count.
• Receipts in a playtest: a duplicate PurchaseId grants once; a failed save answers NotProcessedYet and the retry grants once.
• Console: server analytics lines read "AnalyticsService: <Method> event fired."; no "can only be executed on the game server" line.
• Telemetry trace: funnel steps forward only; economy events only after a saved grant.

Visual:

• A client capture of the shop with prices, the odds panel and the restricted view, checked with rbx_review.image_checks.
• Store art on a contact sheet at 1,600 px or less, opened and read.

Score the result with critique.md before calling it done.

Common mistakes

Mistake	Symptom	Fix
Grant, then save, then PurchaseGranted	Paid but nothing after a crash	Save first; grant only on LastSavedData
Grant on PromptProductPurchaseFinished	Free items through a fired remote	Grants only in the receipt handler
pcall(handler) success taken as a grant	Granted while the handler returned nil	Return true only after the save; nil is not granted
Two scripts set ProcessReceipt	One product never grants	One Purchases:bind per server
Prices typed in UI code	Plus and regional prices wrong; optimization blocked	PriceText from GetProductInfoAsync
Odds in two tables	Disclosed odds differ from rolls	One odds module for both
Paid egg prompted from the client	Restricted players can still buy	Server prompt after the policy check
Analytics or ConfigService on the client	Silent no-ops; "only available in the game server"	Server calls; client reads attributes
Client-sent funnel steps trusted	Funnel inflated by exploiters	Forward-only, validated on the server
Experiment read at join for everyone	Enrollment diluted by players who never see it	forPlayer where the feature appears
Stopping a test on an early peek	Winner that does not hold	Fixed duration from the MDE
Thumbnail text or giveaway claims	Rejected video or discovery penalty	Real gameplay, plain titles

Handoffs

• With scenario-roblox-data: delivers product ids and grant amounts; receives the Receipts handler table, which owns persistence (PurchaseIds log, LastSavedData confirmation). This skill owns the receipt contract and the catalog.
• From scenario-roblox-networking: the purchase-signal finding and validated remotes for shop steps and egg requests.
• To scenario-roblox-ui: price and odds text, perk_ and cfg_ attributes, the "Details" panel contents; layout and scaling stay with UI.
• To scenario-roblox-performance and scenario-roblox-tooling: telemetry send rates; CI for the offline audit and suites.
• Scenario art: thumbnails start from a staged gameplay capture, reframed to 16:9 at 1920x1080 with scenario-formats or edited with scenario-image; experience icons (512x512 square) and pass or product icons (transparent PNG, subject inside the circle) come from scenario-game-assets; scenario-quality-gate scores them. Files land in the project, check_art runs, and only the owner uploads. Keep text, prices and "free" claims out of all art; moderation applies to every upload.

Current Roblox notes

• BindReceiptHandler (with Enum.ReceiptDecision and optional filters) is the current receipt API; ProcessReceipt still works.
• Cross-game product and pass sales ended on 2026-05-30; Robux transfers use their own receipt types.
• Roblox Plus replaced Premium for new subscribers: HasRobloxSubscription, PromptRobloxSubscriptionPurchase.
• Creator Rewards replaced Engagement-Based Payouts on 2025-07-24.
• Yielding calls carry Async: GetProductInfoAsync, UserOwnsGamePassAsync. Use Enum.InfoType.GamePass: the price-optimization page says GamePasses, which does not exist.
• Play-through rate and first-play bounce replaced QPTR; the thumbnails page still says "qualified play through rate".
• Open Cloud has Configs, Experimentation, Analytics Query and Ads APIs, all on owner-issued keys.
• Free trials for passes were announced at RDC 2026 for October 2026 and are not in the docs yet: verify before using.

References

• references/expert-notes.md: principles by source, with timestamps and live findings.
• references/procedures.md: full code for stages P1 to P11, each with its tests and live result.
• references/critique.md: blocking and quality rubric with how to measure each item.
• references/launch-liveops.md: item metadata, rewarded video eligibility, shop order, content pacing and power creep, paid-item trading.
• references/gui-paths.md: Creator Hub and Studio paths for what agents cannot do, plus Open Cloud routes.
• references/sources.md: every doc and video, with credentials and timestamps.
• scripts/rbx_monetization.py: audit, kit loader, Studio probe, art checks, thumbnail crop, MDE.
• scripts/agentkit/monetization/: Catalog, Purchases, Entitlements, PriceText, ShopRank, Odds, Telemetry, LiveConfig, MonAudit, lune_mon_audit.luau.