Decision brief · Oracle Fusion Incentive Compensation 26C · October 10, 2026

Where we stand against the current Oracle experience.

Our strongest opportunity is a reusable, customer-branded experience with traceable results and focused reviews. Oracle already supplies Redwood payee/team views and documents broader AI investigation and plan-building capabilities. A blanket claim that it is “much better than 26C” is not yet substantiated.

Evidence boundary. Oracle's current documentation establishes delivered capabilities and prerequisites. Local source review and browser checks establish our app's behavior on sample records. No native 26C customer session, matched real-data comparison, or human task-timing study has been observed. Native step counts, comparative task times and production capacity remain unmeasured.

Comparison scorecard

Area	Oracle baseline available by 26C	Our current experience	Defensible standing
Payee dashboard	Redwood dashboard centralizes compensation progress and previous earnings. Introduced 24B; requires exposure through Visual Builder Studio. Oracle source	Next payout, components, previous-period change, rate milestones and links to detail.	Documented capability overlap Information arrangement may help, but “modern dashboard” is not a unique benefit.
Team visibility and coaching	My Team displays period credit totals with team-performance drill-down and configurable content. Available by default for the participant manager role from 25A. Oracle source	Pace, a coaching-priority list, readable Team rows, approvals and plan acceptance.	Capability overlap / UX hypothesis Prioritization needs a same-data task test; team charts alone do not establish superiority.
Understanding compensation	26C Payee Performance Insights Agent supports earnings, credits, goals, attainment and trends with contextual responses and visual insights. Requires publishing the agent and exposing a smart action. Oracle source	Deterministic component comparisons and bounded reads of reported calculation inputs, outputs, raw rate-table values and record IDs. No conversational agent.	Oracle has broader capability Our opportunity is transparent evidence and predictable review, not a claim of richer AI explanation.
Credit investigation	Credits Investigation Advisor can explain missing credit by checking rules applicable on the transaction date against transaction attributes. Introduced 26A; setup required; available to Incentive Compensation Plan Administrator, Analyst and Manager roles. Oracle source	Dispute details distinguish the claim snapshot from explicitly linked credit or earning records; identity and transaction mismatches are rejected.	Oracle has broader investigation Our record comparison helps evidence review but does not diagnose rule eligibility or root cause.
Paysheet review	Fusion REST documents calculated earnings, payment totals, differences and nested payment transactions, including holds and manual adjustments. This establishes data availability, not the native screen layout or review effort. Summary · Payment records	Scoped payment records and summary in one drawer, held-line indicators, paging, exact selection confirmation and background approval contracts.	Candidate task advantage Need to measure fewer navigation changes, errors and missed exceptions against the actual native review.
Plan administration	26C Compensation Plan Assistant converts uploaded plan documents into proposed configuration across plans, components, performance measures and rate tables, with review, refinement and implementation. Oracle states no enablement is needed; access requires Compensation Plan Administrator. Oracle source	Basic draft-plan workflow; no document-to-plan builder.	Oracle ahead Keep complex plan design in Oracle. Do not position our draft screen as an equivalent replacement.
Partner delivery and branding	Oracle provides Visual Builder configuration and extension capabilities. Redwood adoption matrix	Packaged source, branding preview/export, role navigation, EN/ES bundles, contracts, readiness checks and handoff guidance.	Candidate packaging advantage Validate repeat-customer branding effort and actual delivery reuse. Do not claim native Oracle cannot be branded.
Mobile and scale	No matched mobile task or capacity test has been performed on a 26C pod.	Phone/tablet bottom tabs, desktop sidebar, labelled Team rows, bounded detail requests. Optional OIC profile is disabled until backend activation.	Unmeasured comparison Local layout checks show usability fundamentals; they do not prove superiority or 100,000-participant / 10M-per-hour throughput.

Verified Oracle source register

Primary Oracle pages checked on October 10, 2026. Release readiness pages establish documented availability and setup; they do not demonstrate achieved time savings, answer correctness or this customer's configuration. “Oracle ahead” above means a broader documented capability than the local implementation.

Source / release	Access and enablement	What this supports
Redwood payee dashboard · 24B	Participant, Participant Manager, Administrator; Visual Builder Studio exposure step.	Central compensation progress and previous earnings; no unique-dashboard claim.
My Team · 25A	Participant Manager; no enablement required; administrators can add custom content.	Team credit totals and performance drill-down; native team visibility is established.
Payee Performance Insights · 26C	Participant or Participant Manager; publish agent, expose smart action, configure security.	Conversational compensation exploration and visual insights; actual answer quality remains untested.
Credits Investigation Advisor · 26A	Incentive Compensation Plan Administrator, Analyst or Manager; publish template and expose smart action.	Rule/date/transaction investigation beyond our explicit record comparison.
Compensation Plan Assistant · 26C	Compensation Plan Administrator; no enablement required.	Document-to-plan configuration with human review and refinement beyond our basic draft flow.

The linked Sales/Fusion Service REST reference is a rolling contract reference, not an observed 26C pod response. Validate its supported fields, permissions and statuses on the target pod. This comparison concerns Fusion Incentive Compensation; it does not use E-Business Suite ICM or HCM Workforce Compensation as the baseline.

Concrete changes made from this review

Recorded earning explanation

Load one earning by earning and participant ID. Show the reported input, output, credit, rate-table value and tier count. Preserve its currency and posting status. Do not infer a percentage, formula, or payment confirmation.

webApps/icm/resources/js/icm/evidence.js · seller/partner statement drawers

Actual paysheet evidence

Refresh the selected native paysheet, then query its own nested payment records in pages of 50. The paysheet supplies the total; a displayed page is never used to reconstruct it.

manager reviewPaysheet action chain · native paymentTransactions child

Claim versus source

Keep the case's original amounts as a snapshot. Load only explicit credit and earning links, with participant and transaction checks. Missing or denied records remain visibly unverified. New-dispute prefill selects only a uniquely linked credit and does not infer an incorrect split or an expected amount.

manager dispute loadEvidence action chain

Priority improvements

P0 — validate Fusion mappings and actual outcomes. Reconcile native IDs, currencies, statuses, totals, periods and authorization on the target 26C pod. Test denied access and mismatched records under each real role. Missing values must remain unknown. Exit evidence: secured read results reconcile to the native record and wrong-scope cases are denied.
P1 — establish an explanation task advantage. Make it easy to move from a component change to its underlying earning. Add causal descriptions only when secured Fusion/report data explicitly establishes them.
P1 — make payment exceptions actionable. Review holds, manual-adjustment reasons and differences with exact native records and source links. Assess whether the native workflow already supplies the same context more efficiently. Exit evidence: reviewers identify all seeded exceptions without approving an unintended record.
P1 — strengthen dispute evidence. Clearly distinguish participant expectations, claim-time snapshots and verified current records. Do not equate a missing record with zero or a justified claim.
P2 — complement Oracle agents. Evaluate linking to Oracle's configured payee and credit-investigation assistants with role-appropriate access. Participant Manager and Incentive Compensation Manager are distinct documented roles; a team manager must not be assumed to have investigation access.
P2 — activate and load-test the optional Oracle backend. Bounded UI requests are implemented; secured report datasets, OIC admission budgets and durable jobs still require deployment and capacity tests.

The test that would substantiate “much better”

Use the same 26C business unit, participant scope, plan, period and source records in both interfaces. Test the native Redwood experience with the customer's intended agents and configuration enabled. Record actual security roles and agent availability: a missing optional setup is a deployment finding, not evidence of an absent Oracle capability. Recruit 6–8 representative people for each tested role. Counterbalance which interface they use first and use matched alternate tasks to reduce learning effects.

Task	Successful outcome	Measure in both interfaces
Seller: explain a payout change	Identify the changed component and inspect its source earning without treating a guessed rate as evidence or confusing posted with paid.	Unassisted success; active task time; incorrect explanations; help requests; confidence rating.
Manager: review paysheets	Review selected totals, held lines and adjustments; identify the records requiring investigation; confirm precisely the intended selection.	Time; selection mistakes; missed exceptions; incorrect totals/currency interpretation; navigation changes.
Manager: investigate a dispute	Separate the claim from verified source amounts, inspect the linked record and decide with the correct participant and transaction context.	Time; identity mistakes; unsupported conclusions; help requests; confidence rating.

Proposed decision gates — targets, not results: at least 90% unassisted completion; at least 30% lower median task time on two of the three primary tasks; no deterioration in correctness and no critical wrong-record, wrong-currency or wrong-selection errors. Report individual results and uncertainty; a small study identifies practical differences but is not a population-wide guarantee.

Human task comparison worksheet (unmeasured)

Automated task activations and request sizes are engineering checks. They are not human usability timings. Native benchmark cells remain empty until measured.

Local automated task evidence

Measured sample paths on desktop and phone, 2026-10-10T16:43:27.144Z. Scripted activations exclude sign-in, scrolling and typing. Record request counts cover the source-evidence step, not every request needed to enter the task.

Task	Viewport	Activations	Evidence reads	Largest requested page	Evidence displayed
Seller — earning explanation	phone	3	1	1	8 reported fields
Manager — selected paysheet	phone	3	2	50	5 payment records
Manager — linked dispute evidence	phone	4	1	1	2 reported fields
Seller — earning explanation	desktop	3	1	1	8 reported fields
Manager — selected paysheet	desktop	3	2	50	5 payment records
Manager — linked dispute evidence	desktop	4	1	1	2 reported fields

Earning and dispute counts above describe displayed facts, not fetched record counts. The sample paysheet contains five payment records; the request cap remains 50.

Validation: the full regression suite passed 253 tests; the added rate-unit formatting regression also passed. The integrated all-role screen walk passed 72 visits. Final EN desktop/phone experience checks passed 10 visits; Spanish phone experience checks passed five. Schema and JavaScript syntax checks report no errors. Component and statement totals now retain currency precision so displayed terms reconcile.

Machine-readable sample task results · Unmeasured native comparison worksheet

Command: node scripts/smoke.js BASE_URL --roles seller,manager --experience-only --evidence-report results.json. These checks are not human usability results, do not process production transactions and do not establish Fusion throughput.

Prepared from public Oracle documentation and this repository on October 10, 2026. The baseline is 26C availability, not a claim that every optional feature is enabled in a particular customer's environment.