Six-month roadmap · Rocketizer Intelligence

PhaseS1S2S3S4S5S6S7S8S9S10S11S12

evidence and scope

discovery

data foundation

shadow product

safety controls

execution rail

alpha

outcome validation

workflow depth

commercial validation

capture validation

portfolio decision

Roadmap Gantt data table

Phase timing shown in the six-month roadmap
Sprint	Weeks	Phase	Sprint objective	Critical path
S1	1–2	evidence and scope	Truth audit, portfolio contract, and metric preregistration	Yes
S2	3–4	discovery	COD-heavy Power Merchant discovery and hero-wedge gate	Yes
S3	5–6	data foundation	Versioned data contract and retrospective baseline	Yes
S4	7–8	shadow product	Shadow-mode recovery decisioning and evaluation set	Yes
S5	9–10	safety controls	Policy, approval, trace, and handoff control plane	Yes
S6	11–12	execution rail	One idempotent recovery action with verification	Yes
S7	13–14	alpha	Three-merchant approval-gated alpha	Yes
S8	15–16	outcome validation	Controlled merchant-outcome pilot	Yes
S9	17–18	workflow depth	Exception depth, handoff, and optional WISMO bridge	Yes
S10	19–20	commercial validation	Packaging, willingness-to-pay, and unit economics	Yes
S11	21–22	capture validation	Shiprocket-assisted attach and GTM experiment	Yes
S12	23–24	portfolio decision	Scale-readiness review and invest/change/stop decision	Yes

Hero-wedge commitment boundary

COD-to-delivered-revenue control loop

The ranked opportunity portfolio places this first because it spans profitable conversion, delivery/RTO margin protection and COD cash realization. Potentially differentiated merchant-authorized Shiprocket order, risk, shipment and action context is conditional on verified Rocketizer access, permitted use, cross-merchant learning rights and measurable lift.

Commit risk queue, approval-gated confirmation/prepaid intent, delivered-order attribution and one verified action on the G0-supported storefront path. Add first-NDR writeback only after the first action, safety and merchant-outcome gates pass.

Sprint contract

S1weeks 1–2

Truth audit, portfolio contract, and metric preregistration

proposedevidence and scope · critical path

S2weeks 3–4

COD-heavy Power Merchant discovery and hero-wedge gate

proposeddiscovery · critical path

S3weeks 5–6

Versioned data contract and retrospective baseline

proposeddata foundation · critical path

S4weeks 7–8

Shadow-mode recovery decisioning and evaluation set

proposedshadow product · critical path

S5weeks 9–10

Policy, approval, trace, and handoff control plane

proposedsafety controls · critical path

S6weeks 11–12

One idempotent recovery action with verification

proposedexecution rail · critical path

S7weeks 13–14

Three-merchant approval-gated alpha

proposedalpha · critical path

S8weeks 15–16

Controlled merchant-outcome pilot

proposedoutcome validation · critical path

S9weeks 17–18

Exception depth, handoff, and optional WISMO bridge

proposedworkflow depth · critical path

S10weeks 19–20

Packaging, willingness-to-pay, and unit economics

proposedcommercial validation · critical path

S11weeks 21–22

Shiprocket-assisted attach and GTM experiment

proposedcapture validation · critical path

S12weeks 23–24

Scale-readiness review and invest/change/stop decision

proposedportfolio decision · critical path

Milestones and decision gates

Sprint	Milestone	Exit criterion	Gate
S1	Freeze roadmap evidence and portfolio boundary	Source register is versioned; Rocketizer/Copilot/MCP/RADAR/TwentyTwo/Fastrr ownership questions have named decision owners or remain explicit blockers.	—
S1	Select and verify a supported storefront execution path	Every marketed connector/action is classified live, sandbox, demo, announced or unknown; exactly one supported connected-storefront read path is selected and production-safe verified, and at least one candidate action rail on that path is verified before recruitment.	G0
S1	Preregister metric hierarchy and attribution plan	Primary, secondary and guardrail metrics have formulas, event owners, counterfactual design and no double counting.	—
S2	Run structured COD-heavy Power Merchant discovery	Target 12 interviews with connected-storefront Power Merchants on the G0-supported path across at least three operating shapes; current COD acceptance, confirmation, prepaid conversion, NDR/RTO leakage, buyer, approver, data access and willingness to pilot are recorded without assuming demand.	—
S2	Validate the ranked hero and top alternatives with merchant evidence	The ranked 93-point hero is updated with merchant observations and uncertainty; any replacement requires a documented scored decision.	—
S2	Pass hero-wedge gate	A single COD-heavy connected-storefront ICP, economic loss baseline, first confirmation/action rail and three recruitment targets on the supported path are specified.	G1
S3	Define order-risk-contact-payment-shipment-action data contract	Versioned schema covers eligibility, pre-dispatch risk, order/address, contact/consent, confirmation, prepaid conversion, hold/release, shipment/NDR, merchant policy, approval, external action response and terminal delivery/RTO verification.	—
S3	Build retrospective baseline and eligibility ledger	Eight-week baseline computes eligible exceptions, recovery/RTO, time, manual effort and merchant economic inputs with reproducible queries.	—
S3	Contract Cohort 1 operating terms	Consent, holdout, incident response, data deletion, action approval, success metrics and support burden are agreed before any live data/action.	—
S4	Ship read-only COD-to-delivered recommendation in shadow mode	For each eligible high-risk COD order or first NDR, the system proposes confirmation, prepaid conversion, hold/release, reattempt/cancel or handoff with reason and confidence but performs no external write.	—
S4	Freeze adjudicated evaluation set and failure taxonomy	At least 200 eligible historical high-risk COD/first-NDR examples are double-reviewed; disagreements, false holds, unsafe recommendations, missing context and unsupported actions have explicit labels.	—
S4	Pass shadow-quality gate	Quality report includes precision, recall, intent accuracy, unsafe recommendation rate, calibration and merchant overrides.	G2
S5	Encode merchant policy and permission matrix	Roles, allowed action, monetary/contact bounds, consent, rate limit, approval SLA, expiry, deny rule and escalation are machine-readable per merchant.	—
S5	Implement approval, immutable trace, safe stop, and contextual handoff	Every proposal captures input, decision, policy version, approver, external response and verification; denied/expired/ambiguous cases stop and hand off with context.	—
S5	Pass safety-control gate	All injected unsafe actions are denied or handed off; every attempted action is attributable and replayable.	G3
S6	Implement one lowest-risk COD/recovery action adapter	The selected verified action—such as approved confirmation/prepaid-link contact, hold/release, or first-NDR instruction—has authenticated request, idempotency key, deterministic input, explicit approval and terminal-state verification; selection comes from live rails, not marketing copy.	—
S6	Implement reconciliation, retries, rollback/compensation, and silent-failure detection	Reconciliation queue identifies unknown terminal state, duplicates and third-party failures; retry and compensation behavior is documented and tested.	—
S6	Pass execution-rail gate	Adapter is safe enough for an approval-gated merchant alpha.	G4
S7	Launch Cohort 1 with daily supervision	Each merchant runs shadow plus approval-gated action with holdout assignment and named incident contact.	—
S7	Run daily error review and weekly merchant outcome review	All overrides, handoffs, customer complaints, action failures and metric anomalies are reviewed; manual support minutes are logged.	—
S7	Pass alpha trust and burden gate	Merchant operators can explain decisions, manage approvals and verify outcomes without hidden intervention.	G4A
S8	Launch Cohort 2 with pre-registered treatment/holdout	Eligible exceptions are assigned to treatment and contemporaneous holdout or a justified matched design before the outcome is known.	—
S8	Measure merchant economics and uncertainty	MET-001 and eligible COD delivery lift are computed without double counting; prepaid conversion, first-NDR recovery, false holds, manual effort, complaints and safety guardrails include uncertainty by merchant.	—
S8	Pass outcome gate	A decision memo distinguishes signal, uncertainty, merchant heterogeneity and operational cost.	G5
S9	Close the first-NDR path and top economically weighted failure modes	The pre-dispatch COD path and first-NDR reattempt/cancel path have deterministic policy, verified action or contextual handoff; the long tail remains explicit.	—
S9	Add governed cross-channel state and WISMO handoff	Shipment state, buyer intent, prior contact and current recovery status are available to support/handoff; WISMO containment is measured but is not a second autonomous product.	—
S9	Pass repeatability and support-load gate	Reliability, merchant configuration time and support load are stable enough for packaging.	G5A
S10	Define one pilot package and one scale hypothesis	Package states eligible workflow, included actions, approval model, onboarding, support, metric contract and exclusions; no broad employee bundle.	—
S10	Test paid-pilot, subscription-plus-usage, and outcome-linked concepts	At least 8 qualified buyer conversations use explicit price/term tradeoffs; stated interest is separated from signed willingness.	—
S10	Pass commercial economics gate	A single commercial test enters Sprint 11.	G6
S11	Instrument an eligible Shiprocket-assisted referral funnel	Funnel events cover eligible merchant, contact, diagnosis, setup, first verified action, paid conversion and support cost.	—
S11	Run controlled assisted-onboarding test	A bounded qualified cohort receives the outcome-led pitch and assisted setup; non-eligible merchants are not counted as failures.	—
S11	Pass Shiprocket-capture gate	Product attach and activation are interpreted with cost and uncertainty; shipping share, retention and ARPU remain lagging hypotheses unless observed.	G7
S12	Audit scale readiness against product, data, safety, economics and capture	Readiness pack reports current truth, cohort outcomes, reliability, security/privacy gaps, support load, pricing evidence and causal limitations.	—
S12	Re-score committed wedge and options	RTO/NDR, WISMO, remittance and returns are scored with actual merchant and operating evidence; no option inherits hero status automatically.	—
S12	Make invest, change-wedge, extend-evidence, or stop decision	One decision, next six-month budget envelope and explicit de-scopes are documented.	G8

Launch cohorts are recruitment targets, not assumed customers

Entry timing and evidence gates for each proposed cohort
Cohort	Target	Entry	Exit gate
Offline and sandbox	Historical replay plus synthetic failure injection; no merchant-facing action.	S3	G2
Design-partner alpha	Recruitment target of 3 connected-storefront Power Merchants on the G0-verified supported path; none are assumed to exist today.	S7	G4A
Controlled pilot	Recruitment target of 6-10 merchants after alpha gate.	S8	G5
Limited availability candidate	Maximum 20-30 merchants only if outcome, safety, economics and operating gates pass; not a committed launch.	S12	G8