Forward-cost analysis counts the work, delay, dependencies, operations, and uncertainty still ahead under each path while preserving the value of reusable assets.
Fix the present state and time boundary
Record the current revision, environments, supported workflows, incident load, active defects, team, contracts, operating cost, and target date. Prior development expense belongs in history. Include it only when it creates a current asset, obligation, dependency, or constraint. The decision concerns effort and consequences from the present state forward.
Use the same categories for every scenario: discovery, architecture, implementation, tests, data, integrations, migration, parallel operation, release, rollback, training, documentation, support, infrastructure, vendors, and contingency. State what each range includes. A rebuild estimate that omits migration and dual-running cost will look artificially comparable to an operational repair path.
Count failure-management work
Measure how much engineering time currently goes to incident response, flaky tests, manual recovery, dependency workarounds, repeated regressions, release repair, and explaining undocumented behavior. Separate one-time stabilization from burden likely to continue under the same structure. Use ranges and source periods instead of converting partial timesheets into false precision.
A high burden does not automatically require rebuilding. Test whether a bounded repair changes the ratio. Conversely, a low incident count can conceal stalled delivery if the team avoids touching the risky subsystem. Include deferred roadmap work and ownership concentration when there is evidence that the current structure constrains the target state.
Expose uncertainty and path dependence
For every range, list assumptions, evidence quality, dependencies, earliest reversible point, and decisions that close later options. A repair can become expensive if it must be undone before rebuilding. A rebuild can destroy useful behavior or require long parallel operation. A selective replacement may preserve the most assets while introducing temporary integration cost.
Reality Contact, LLC builds this comparison for Forward Cost Dossier but does not certify estimates or advise on accounting, procurement, or investment. The buyer validates labor rates, vendor quotes, staffing, opportunity cost, and risk tolerance. The output is a decision record with bounded ranges, not a fixed implementation bid or schedule promise.
Where the service stops
Reality Contact, LLC prepares technical decision evidence but does not certify security, compliance, reliability, valuation, delivery dates, or cost estimates; it does not give legal, investment, accounting, or procurement advice and does not guarantee that any selected path will succeed. The buyer defines the target state and constraints, supplies authorized evidence, challenges assumptions, approves the proof, chooses the path, assigns funding and owners, and commissions any repair, rebuild, selective replacement, or retirement under a separate scope. This technical decision service does not replace security, compliance, legal, accounting, investment, procurement, product, data, architecture, or production-operations review. The buyer owns the target state, estimates, constraints, proof authority, funding, staffing, implementation scope, and final repair, rebuild, selective replacement, retirement, or observation decision.
Sources: AWS guidance on the complexity and cost of refactoring; AWS phased approach to application modernization.