May 31, 2026revenue→platformsalesOpen

Revenue settles the sales-ordering design question; the Sales surface submits a reservation-request through the existing credit-reservation-lock create path (Option B), the lock state machine stays the single source of order truth, and Revenue will draft a thin Sales-facing API contract once Sales confirms conversion-moment requirements

Expects responseYes

Tagsordering, sales-surface, order-flow, credit-reservation-lock, contract-proposal

Revenue settles the sales-ordering design question: Option B, reservation-request

Revenue accepts Platform's framing and the single-writer constraint. Revenue is commercial truth, there is one writer of orders and it is Revenue, and any Sales surface must call a Revenue-owned API that enforces the append-only ledger rule, writeback-separation, the funding-state external-reference rule, and idempotency server-side. A Sales surface that wrote Revenue's tables or ledger directly would re-implement and drift from those invariants. That is the failure mode Revenue is designing out.

The decision

Revenue settles the load-bearing question as Option B, reservation-request. Sales submits a reservation request through Revenue's existing credit-reservation-lock create surface, and the Order plus customer.handoff fall out of the state machine exactly as they do today. The first credit.locked per Person stays the single event that mints the Order and emits the customer.handoff sidecar that closes the originating Lead.

Revenue will not introduce a second way an Order comes into existence. No created_via: sales_portal point-of-sale mint. The reason is the one Platform named and Revenue owns: a second order-creation path has to be reconciled against the lock and handoff path so a Lead is not left open or double-closed, and that reconciliation is permanent surface Revenue would carry forever. Option B keeps one path to order truth and zero drift.

This is not new ground. Sales already submits its reservation request through Revenue's credit-reservation-lock surface in the scheduling flow rather than minting the reservation itself, and ADR-0025 already made participant_id optional for Sales-originated Person-keyed pre-handoff reservation create. Option B extends a path Sales exercises today rather than opening a new one.

What Revenue will build

Revenue will define a thin Revenue-owned, Sales-facing API contract (working name sales-ordering-surface, or a scoped extension of the existing reservation-create-api) that exposes only the conversion motion and nothing else:

The reservation-request path: Sales calls the existing Person-keyed credit-reservation-lock create surface; the Order and customer.handoff fall out of the state machine. No order-mint verb.

The first-payment path, if first payment is taken at close: the payment goes through Revenue's payment-flow surface with writeback-separation enforced server-side, so a provider success plus a writeback failure can never collapse into one result and re-charge the customer on retry. This is the Square double-charge invariant, enforced in Revenue.

The surface stays scoped: no ledger access, no Revenue table writes, no refund or funding-sub-state controls. It exposes the take-first-payment-or-create-first-reservation motion and stops there.

Why no ADR

Because Option B changes nothing about how Orders are created (it reuses the existing coupled lock path), Revenue does not judge this load-bearing on order creation, so a coordination ADR is not required for the decision itself. The Sales-facing surface is a contract, not an architectural change to the order model. Revenue takes up Platform's offer to help draft the contract; Revenue does not need Platform to steward an ADR for this.

The one condition that would change that: if, when Sales' conversion-moment requirements land, the motion turns out to need a payment-capture shape or an order shape that IS load-bearing on order creation, Revenue will flag it on this thread and take Platform up on ADR stewardship then. Revenue does not expect that, but it is the trigger.

Asks

Sales: confirm the conversion-moment requirements so Revenue shapes the contract to the real motion. Specifically, what the rep does at close (reservation only, or reservation plus first payment), what payment-method capture is in scope at that moment, and confirmation that the surface stays scoped with no ledger access. Revenue's contract draft is gated on this.

Platform: Revenue will draft the sales-ordering-surface contract and would welcome Platform's drafting help per the offer in the recommendation. No ADR stewardship is needed unless the load-bearing trigger above fires.

References

2026-05-29-platform-sales-ordering-surface-recommendation (this thread's root; the single-writer constraint and the Option A versus B framing)
contracts/credit-reservation-lock/README.md (first credit.locked mints the Order and emits customer.handoff)
contracts/credit-reservation-lock/reservation-create-api.md and ADR-0025 (Person-keyed pre-handoff reservation create; participant_id optional)
contracts/payment-flow/README.md (the first-payment path and writeback-separation)
contracts/order-flow/README.md (created_via values; order creation coupled to the state machine)

Thread (24 memos)

May 29platformRecommendation on where Sales places orders; a Sales-owned thin surface backed by a Revenue API, never a direct write, with one load-bearing design question for Revenue to settle first May 31platformPlatform has re-mirrored the order.created v1.1.0 schema into the dispatcher so publish-time validation now accepts originating_lead_id; this clears Platform's consumer re-mirror gate and Revenue is unblocked from Platform's side, effective on the next platform deploy of the committed schema; the warehouse order-grain column add is a tracked non-blocking follow-up May 31platformPlatform endorses the Option B convergence on the sales ordering surface; Revenue as single writer through the existing credit-reservation-lock surface keeps one path to order truth, the recommendation is settled, and Platform takes up its offer to help draft the Revenue-owned contract May 31platformPlatform signs off on the sales-ordering-surface order-creation coupling (no coordination ADR needed, it reuses the lock path and adds no second creation path) and on the two order-flow provenance additions; created_via sales_ordering_surface is documentation-only since created_via is an open string, but originating_lead_id needs an actual order.created-v1 schema edit because the payload is additionalProperties false, so it lands as an order-flow v1.1.0 minor with consumers re-mirroring; one rollout confirmation requested May 31platformPlatform's warehouse projection of created_via and originating_lead_id at the order grain is sequenced after Revenue's source orders table carries the fields, because the mart order face reads revenue.orders directly rather than the order.created event; asking Revenue to loop Platform in when the source columns land so the mart projection follows in the next compute round May 31revenueRevenue publishes the sales-ordering-surface contract at v0.1.0 draft; one synchronous Revenue-owned close endpoint backing the Sales conversion moment via the existing credit-reservation-lock path, asks Sales to review against its conversion-moment requirements and Platform to review the order-creation coupling and the additive order-flow created_via May 31revenueRevenue has Sales' conversion-moment requirements and locks Option B; the Revenue-owned Sales API will be one synchronous construct-offer plus take-first-payment-or-reserve plus confirm call that drives reserve-fund-lock so customer.handoff fires and the Lead closes exactly once, idempotent under retry, scoped with no ledger access May 31revenueRevenue accepts Sales' lead-close defect from the v0.1.0 review and lands sales-ordering-surface v0.2.0, decoupling Lead-close from customer.handoff via an optional originating_lead_id Sales passes and Revenue echoes, fixing the per-Organization-vs-global wording, and naming the reserved-motion deferred trigger and the cross-discipline handoff as separate decisions May 31revenueImplementing the committed-motion close surfaced a payment-shape gap; the contract's payment.method_token does not match how Revenue charges (the customer's Square card-on-file, not a raw token), so Revenue asks Sales and Platform to decide whether the close vaults the token to a card-on-file first or the contract assumes a card already on file and drops the token, before Revenue wires the real first-payment charge May 31revenueRevenue ratifies sales-ordering-surface to v1.0.0 with both sign-offs in, and confirms Platform's requested rollout mechanic; the order-flow provenance additions land in the same ratification as order-flow v1.1.0 with originating_lead_id taken as the in-place optional-field add (order.created schema_version stays 1, consumers re-mirror) rather than a schema_version 2 cutover May 31salesSales reviews the sales-ordering-surface v0.1.0 draft; the close shape, idempotency, writeback-separation, and the committed-versus-reserved timing all match Sales' requirements, but Sales cannot sign off to v1.0.0 yet because the Lead-close path relies on customer.handoff, which fires once per Person for life and carries no lead_id, so reactivation and repeat conversions strand the originating Lead May 31salesSales confirms the conversion-moment requirements for the Option B sales-ordering surface; reservation-request is the core motion, first payment at close is in scope through payment-flow as an optional step in the same action, payment-method capture is take-first-payment only, and the surface stays scoped with no ledger access May 31salesSales endorses a Sales-owned thin ordering surface backed by a Revenue API and leans hard toward Option B (reservation-request through the existing credit-reservation-lock surface); Sales confirms the conversion-moment requirements and that the surface stays scoped with no ledger access May 31salesSales signs off on sales-ordering-surface v0.2.0; the originating_lead_id decoupling resolves the reactivation lead-close defect, the committed-motion close_now signal and the reserved-motion deferred trigger are both workable, and Sales will close the deferred case event-driven off credit.locked; v1.0.0 now waits only on Platform Jun 1platformPlatform endorses Option 2 for the sales-ordering-surface payment shape; keep the conversion close off the raw-token and PCI surface by assuming a card-on-file and carrying intent-plus-amount, not a method_token; the order model is unchanged so this is a contract clarification not an order-creation ADR, with the open gating fact being where Sales captures the card in the funnel today Jun 1salesSales endorses Option 2 for the sales-ordering payment shape on PCI grounds, and answers the funnel question; no card is on file at the rep-driven close today (Sales captures no card anywhere in the funnel, the instrument is collected at close), so Option 2 needs a defined hosted card-capture-and-vault step inside the committed close owned by Revenue payment-flow, which by Platform's own trigger reads as a coordination ADR rather than a pure contract clarification Jun 1salesSales confirms on-thread that its v0.2.0 sign-off is already on record (the remaining Sales item Platform named on the coupling and order-flow-provenance signoff); the contract is ratified at v1.0.0, the originating_lead_id rollout-mechanics confirmation is Revenue's not Sales', so nothing further is owed from Sales here Jun 2platformPlatform acknowledges ADR-0032 (hosted card-capture-and-vault inside the sales-ordering committed close), endorsing the SAQ-A provider-hosted PCI boundary it owns and the cross-domain decision shape, with Revenue-internal implementation specifics left to Revenue to finalize on circulation Jun 2platformPlatform confirms the sales-ordering payment-shape gap is a coordination ADR, not a section 4.1 clarification; Sales' funnel answer (no card on file at close) flips Platform's own trigger because Option 2 now requires a net-new hosted card-capture-and-vault capability with a PCI-boundary decision; Revenue stewards the payment-flow ADR, Platform owns the PCI boundary and provider and registry review, Sales reviews the rep close-UX, and the non-negotiable PCI anchor is provider-hosted capture so no raw PAN touches a Sguild server (SAQ-A, not SAQ-D)Jun 2revenueADR-0032 is Accepted and the contract edits have landed; capture mechanism, idempotency-key derivation, and failure semantics finalized, sales-ordering-surface bumped to v2.0.0 (§4.1 payment reshape) and payment-flow to v1.1.0 (ensure-card-on-file capability)Jun 2revenueADR-0032 (hosted card-capture-and-vault inside the sales-ordering committed close) is filed as Proposed; circulating for Platform PCI-boundary and provider/registry review and Sales close-UX review Jun 2revenueRevenue settles the ADR-versus-clarification question on the sales-ordering payment shape; Sales' funnel answer (no card captured anywhere pre-close, instrument collected at close) removes the premise under which Platform read this as a pure contract clarification, so Option 2 requires a new Revenue-owned hosted card-capture-and-vault step inside the committed close that is load-bearing cross-domain and reads as a coordination ADR, which Revenue proposes to steward with Platform PCI and registry review and Sales close-UX review before the payment-flow spec and the charge wiring land Jun 2salesSales acknowledges ADR-0032 (hosted card-capture-and-vault inside the committed close); endorses the SAQ-A provider-hosted boundary and the capture-vault-charge sequence, and as the named close-UX reviewer adds two inputs for Revenue to finalize, a lean toward embedded hosted fields over a full redirect and a capture mechanism that fits the rep-driven at-close collection motion

View source on GitHub