AWS Migration Assessment: Compare Scope Before Price

Compare a metadata-only AWS migration review with an isolated rehearsal using task-level evidence, effort ranges, customer inputs and acceptance limits.

Buy the smallest assessment that can change your next workload decision. Ask the delivery lead to connect each task to authorized evidence, an effort range and an acceptance artifact before comparing prices. A metadata review and a runnable rehearsal can both be useful, but they purchase different answers. Neither becomes a migration commitment because it appears in an AWS proposal.

A small hosting bill does not establish a simple migration. One required host executable, an unavailable application owner or an unobserved recovery dependency can determine the work. Equally, a broad transformation assessment can be unnecessary when a narrow evidence review would expose the decision-blocking issue.

This article provides a comparable assessment-offer record. It is for the buyer and delivery lead deciding what to commission next, not selecting a consulting company or calculating a migration business case. The worked effort ranges below are constructed teaching inputs. No customer assessment, quotation, account inspection or rehearsal was performed.

Name the question that the assessment must settle

Replace “assess our AWS migration” with a decision and a finite evidence population. For example: can the declared report-export workflow remain unchanged on the proposed database service, and which missing facts prevent a bounded cost comparison?

Name the supplied schema, scheduled jobs, application revision and required output. State who can confirm that the population is complete. An assessment of the supplied jobs cannot certify an undiscovered monthly export or an unavailable integration. Exclusions belong in the answer, not only the proposal's small print.

Use the database-led scope paper when the migration group itself is disputed. Use the migration strategy guide for later cutover, reconciliation and recovery design. Here the purchase is the next evidence task, not another placement matrix or migration runbook.

Separate a metadata review from an isolated rehearsal

A metadata-only offer can reconcile supplied inventories, review exact vendor support statements, identify conflicts and expose missing inputs in a cost comparison. Its authority can be limited to receiving approved exports. Do not silently turn that into permission to connect a discovery tool, inspect a live account, run a job or upload source material to an external service.

A rehearsal adds execution against a declared fixture and environment. It needs an approved candidate artifact, permitted data and identities, safe destinations, observation and cleanup. It can provide evidence about that tested behavior. It does not automatically establish production compatibility, whole-workload performance or recovery.

There is a concrete distinction for SQL Server. AWS documents that RDS SQL Server Agent does not support command-line scripts through ActiveX, Windows command shell or PowerShell. Reviewing a supplied job that depends on that behavior can identify a conflict with leaving it unchanged on that service; running a larger test does not remove the documented restriction. RDS SQL Server Agent limitations.

If a replacement is needed, design or implementation of that replacement is another task. A rehearsal offer must not hide that work inside “validate compatibility.” It can instead require the customer to supply a separately reviewed candidate, or include a separately scoped engineering item with its own estimate and acceptance.

Compare offers against the same purchased answer

Request a short record for every task: question, evidence population and version, authorized method, customer input, provider effort range and basis, output, acceptance owner, exclusion and stop trigger. A sentence saying “assessment included” cannot answer those questions.

Keep four amounts distinct. Provider active effort describes the people doing the work. Customer active effort covers preparing exports, explaining behavior and reviewing findings. Elapsed waiting depends on access and decisions. Third-party consumption covers resources or licenses actually permitted for the exercise. A person-hour is one person's hour of work, not a calendar promise or a price.

Have the actual estimator date the range and explain its endpoints. The lower bound might assume the agreed exports are complete and readable. The upper bound might include one specified clarification cycle, not unlimited investigation. Name the additional condition that breaks the upper bound. An absent range remains UNKNOWN, not zero effort or a free service. Commercial price, taxes and payment treatment require their own offer terms; this record supplies no hourly rate.

Worked offer M: review the supplied evidence first

The following is a constructed purchasing example, not Ampity pricing, a real estimate or an industry benchmark. Stipulate one report-export workflow with two supplied job definitions, an application-owner description and a dated cost export. The proposed service is standard RDS for SQL Server, not RDS Custom or engine conversion; exact source and target profiles still belong in the actual offer. A fictional review lead assigns the allowances below solely to demonstrate a traceable estimate. No credentials, runnable data or cloud invocation are included.

Offer M asks whether the supplied workflow has an unchanged-target support conflict, which facts remain unknown and what evidence is needed before comparing stay, improve or migrate. It excludes executing jobs, designing replacements and producing a fixed implementation price.

Reconcile one approved export bundle and its declared population

Stipulated provider effort: 1–2 person-hours, including one clarification cycle

Acceptance artifact and boundary: Versioned manifest, missing items and owner-confirmed coverage or explicit gaps; no live collection

Review two job behaviors against the named target's support boundary

Stipulated provider effort: 1–2 person-hours per job, 2–4 total

Acceptance artifact and boundary: One disposition per job with source/version and reason; no execution or replacement implementation

Bound one current-versus-proposed cost question

Stipulated provider effort: 1–2 person-hours

Acceptance artifact and boundary: Quantity/rate-source ledger, retained obligations and unavailable estimates; no invented savings total

Review one combined finding with the workload owner

Stipulated provider effort: 1–2 person-hours

Acceptance artifact and boundary: Accepted evidence record, specific correction request or unresolved decision with next owner; no migration approval

The constructed provider range is 5–10 person-hours: one intake allowance, two job-review allowances, one cost-boundary allowance and one finding review. It is an endpoint sum under the stipulated scope, not a confidence interval or evidence that a real team can complete it in that time. If the two job definitions call an undisclosed shared component, the estimate must be revisited rather than pretending it covers that component.

Separately stipulate 2–3 customer person-hours to prepare the already authorized exports, explain the required output and review findings. Do not add these to the provider invoice or erase them from the buyer's planning. Calendar waiting, professional fee and third-party charges remain UNKNOWN in this example because none was stipulated. A real lead must replace all teaching allowances with a dated estimate against the actual packet.

Worked offer R: add a separately authorized rehearsal

Suppose offer M identifies a required command-shell job. Its accepted answer may be that the job cannot remain unchanged on the proposed RDS execution surface. That is useful decision evidence, not a failed assessment. It does not prove that a replacement will work.

For a second constructed offer, assume a separately reviewed candidate replacement is already available. Offer R would examine one declared output and one agreed interruption case in a disposable, isolated fixture. The candidate implementation is an input, not work silently included in the rehearsal estimate. If it does not exist, hold R and scope its creation separately.

Prepare the agreed disposable environment and isolation checks

Stipulated provider effort: 3–5 person-hours

Acceptance artifact and boundary: Permitted identities/destinations, fixture population and stop owner recorded; readiness must pass before execution

Prepare synthetic inputs and checks for the supplied candidate

Stipulated provider effort: 2–4 person-hours

Acceptance artifact and boundary: Version-bound inputs and expected output/interruption rules; no replacement development or production data

Observe the one approved normal case and one interruption case

Stipulated provider effort: 1–2 person-hours

Acceptance artifact and boundary: Actual observations separate from expectations; failed or unobserved behavior stays visible

Reconcile fixture state, cleanup and the decision record

Stipulated provider effort: 1–2 person-hours

Acceptance artifact and boundary: Retained evidence, disposition of test resources and unresolved effects with owners; no assumed complete recovery

Those invented incremental allowances total 7–13 provider person-hours. Only if M is required, reusable and not already purchased would the combined constructed allowance be 12–23. Do not charge the intake twice or assume an older packet is reusable after the candidate or target changes. These sums exclude customer effort, waiting and resource consumption, and say nothing about a commercial price.

R's customer inputs are the candidate artifact, the accepted output rule, permission for the isolated environment, approved synthetic data and an available acceptance owner. Its estimator must explicitly address environment readiness, specialist availability and cleanup work. No teaching range authorizes resource creation, a paid service call or a fault against a live workload.

The buyer can choose M now and defer R. Buying R is sensible only if runtime evidence can change the next decision and its inputs and isolation are ready. It should not repeat a known unsupported operation merely to produce a red test. Nor is a passing normal fixture enough when the purchased interruption question remains unanswered.

Accept the evidence even when it blocks migration

Agree what constitutes a conforming assessment before work starts. For M, that can mean every admitted job has a supported, conflicting or unresolved disposition, a source reference, a reason and a next owner. An unresolved result is acceptable only where the offer permits it and identifies the missing evidence. Omitting a promised check is an incomplete deliverable, not a successful UNKNOWN.

For R, accept the traceable observations and their limitations, not a promise that every test will pass. A reproducible incompatibility may satisfy the purchased diagnostic task while preventing the proposed implementation from advancing. Acceptance of the assessment, acceptance of a candidate design and authorization to migrate are separate decisions.

Define rework, reviewer response time and dispute/escalation treatment in the actual agreement. The broader outcome-linked delivery paper owns those commercial principles. This record should supply the task-specific evidence those terms refer to, without making a universal payment or warranty promise.

Stop or revise the offer when its premise changes

If the approved export omits a required executable, preserve that finding and stop the affected conclusion. The team can finish unrelated admitted checks if the contract permits, return a limited record and ask the sponsor whether to buy a separate implementation investigation. Do not run the executable to fill the gap, omit it to preserve the fee or silently enlarge the assessment.

If customer inputs arrive late, separate active effort already spent from calendar waiting and any agreed rescheduling cost. Name the next review point and stop condition; do not promise that an indefinite wait is included. The commercial consequence must come from the agreement, not from this example's effort arithmetic.

During a later authorized rehearsal, an unexpected live destination, sensitive input, unbounded consumption or missing stop/observation path ends further test actions. Preserve the approved evidence and assign state reconciliation and cleanup. Removing a test artifact does not establish that every external effect or charge was reversed. No production rollback, deletion or retry instruction is supplied here.

The same modest hosting bill can therefore produce a different assessment offer after one required behavior is uncovered. That change comes from evidence and scope. It is not a reason to enlarge the project to fit an incentive narrative.

Keep funding separate from the offer you can accept

AWS's partner funding page makes eligibility dependent on partner progression and program participation. Its MAP overview distinguishes assessment from later mobilization and migration phases. Neither public page is this buyer's award or this delivery team's quoted fee. This article asserts no minimum spend, funding entitlement or participation status.

AWS promotional credit terms restrict credits to designated eligible charges and state that they have no cash value. Do not assume a prospective credit pays a consultancy invoice. Any different customer/partner commercial arrangement needs its own evidence and applicable terms. The incentive-independent business-case paper owns benefit scenarios and the unfunded investment comparison; do not recreate those spreadsheets to buy this assessment.

Next action: send one comparable assessment record to the delivery lead

Use a short offer record rather than a large questionnaire. Fill the actual values with the estimator and customer owners; the examples above are not defaults.

Decision and population

Name the next decision, workload/output, exact evidence population and revisions, and the owner who confirms coverage.

Task and authorized method

For each question, specify supplied-export review versus separately authorized collection or execution. Identify permitted handling and excluded systems.

Estimator basis and effort range

Record estimator role/date, task quantities, lower and upper provider person-hours, included review/rework cycles and the condition that invalidates the range. Missing estimates remain UNKNOWN.

Customer input and timing

Name the input owner, active preparation/review effort, availability date, permissible waiting and the action if the dependency fails.

Consumption and commercial treatment

Separate resource/license exposure, fee and payment terms from effort. State what is included, capped, excluded or still unestimated; assume no funding offset.

Acceptance and stopping

Name the artifact, reviewer, allowed unresolved findings, incomplete-deliverable rule, stop owner, retained evidence and separate cleanup/recovery obligations.

Scope-change decision

Record the new fact, affected task, work that can still close, proposed additional evidence, revised estimate and sponsor decision. No implementation is authorized by an assessment finding.

Ask the lead to walk one task from question to evidence, effort and acceptance. If two offers answer different questions, request a common metadata-review scope and a separately stated rehearsal option before comparing their prices. You can stop after a useful compatibility finding without committing to migration. If this is an active decision, share the workload and the evidence already available to scope the missing assessment task; contact is optional, and the record above is usable as it stands.