cloud migration consulting services cloud migration consultant cloud procurement request for proposal statement of work

Cloud Migration Consulting: An RFP and SOW Guide

By Peter Korpak, Founder · Last updated

Cloud migration consulting is external support for defining, governing, and delivering a cloud move under an auditable scope. Buyers should compare proposals against the same workload inventory, acceptance gates, named delivery team, risk allocation, and post-cutover handoff—not against headline rates or partner badges alone.

The cloud migration strategy program guide covers workload paths, readiness gates, and wave logic. Once those decisions are agreed, use this guide to turn the program baseline into an RFP, compare responses, select a team, and write a statement of work with acceptance terms.

Use the service purchasing map to define the engagement, compare two fictional bids, or go directly to the SOW acceptance example. The editable comparison CSV contains the worked rows and blank rows for your own bids.

Choose the services you are buying

Decide whether you are buying assessment, migration delivery, or delivery with an operations handoff before requesting a price. An assessment proposal will leave production build and cutover to a later engagement unless they are expressly included.

PurchaseOutputs to requireBuyer prerequisitesAcceptance ownerExclusions to resolve
Assessment and advisoryValidated inventory, dependency gaps, workload recommendations, target-state assumptions, and a delivery scope packAccess to inventory and system owners; business constraints; authority to approve recommendationsBuyer program sponsor with architecture and application ownersProduction build and cutover, remediation, and ongoing support unless separately included
Migration deliveryDefined platform changes, workload moves, test evidence, cutover and rollback artifacts, and as-built recordsApproved workload paths; access and license decisions; funded buyer-owned work; test reviewersApplication and data owners for workload acceptance; platform/security owners for their allocated controlsRefactoring, unidentified dependencies, legacy shutdown, and post-hypercare operations unless included
Delivery plus operations transitionDelivery outputs plus coverage schedule, runbooks, access transfer, operational exercises, and signed handoffNamed receiving team, on-call model, operating procedures, and time to participateReceiving operations owner, with application owner approval for unresolved defectsIndefinite managed operations, later optimization, and support outside the agreed coverage

A buyer commissioning assessment may not have a complete inventory yet. In that case, buy a bounded discovery output and define how it will be accepted before requesting a fixed delivery price. For the technical work behind those outputs, use the cloud migration execution practices.

Start the RFP with a scope baseline

Give every bidder the same dated scope pack and require an exceptions schedule against it. Include the inputs below so bidders can price the work, assign staff, and identify what remains unknown.

Scope inputWhat to provideWhat the bidder must return
Workload inventoryOwners, business criticality, runtime, data stores, lifecycle, target disposition, and known technical debtIncluded workload IDs, excluded IDs, assumptions, and discovery gaps
Dependency evidenceApplication, data, identity, network, certificate, DNS, batch, and third-party connectionsValidation method, unresolved dependencies, and remediation ownership
Target environmentPlatform decision, landing-zone status, regions, identity boundary, network pattern, and policy constraintsTarget architecture, required platform changes, and responsibility split
Migration wavesProposed dependency groups, business calendar, blackout windows, and cutover orderWave plan, entry criteria, exit criteria, and resource conflicts
AcceptanceRequired test evidence, data reconciliation, performance baseline, security evidence, and business sign-offTest method, evidence format, acceptance owner, and defect treatment
OperationsMonitoring, incident response, backup, recovery, patching, cost ownership, and change processHypercare model, knowledge-transfer plan, runbooks, and support handoff

Keep three lists separate: in scope, buyer-owned prerequisites, and excluded. A proposal that silently relies on the buyer to complete discovery, remediate applications, or build the landing zone is not comparable with one that includes those tasks.

Define the unit of scope

Choose stable identifiers that survive proposal and contract revisions: workload ID, application ID, server or cluster group, database, interface, and migration wave. Avoid scope phrases such as “the finance environment” unless an attached inventory resolves exactly what that means.

For each unit, state the intended migration path, current evidence quality, acceptance owner, and treatment of newly discovered dependencies. Use these identifiers in change requests to show which work and assumptions have changed.

Allocate engagement risk deliberately

Choose the contract structure according to what is known about the work. The table shows which risks each party carries and the controls to agree before signature.

Engagement structureBuyer carriesSupplier carriesBest fitRequired guardrail
Time and materialsVolume, productivity, and most discovery riskStaff availability and professional performanceAssessment or genuinely uncertain remediationRole-level rate card, approval limits, burn reporting, and backlog ownership
Fixed deliverable or milestoneErrors in buyer-provided assumptions and approved changesDefined delivery effort and acceptance reworkBounded work with stable inputs and measurable outputsAssumption schedule, acceptance test, cure period, and change formula
Managed capacityBacklog priority and value realized from the teamProviding the agreed capacity and skill mixRepeatable waves where scope order changesNamed roles, allocation, throughput evidence, substitution controls, and exit rights
Outcome-linked componentBaseline quality and dependencies outside supplier controlThe specifically defined outcome measureA narrow result both parties can measure and influenceAuditable baseline, attribution boundary, cap, exclusions, and dispute process

A fixed fee still needs acceptance criteria. Outcome-based fees need a metric the supplier can influence; future cloud use, internal adoption, and wider business conditions may sit outside its control. You can combine structures: bounded assessment, milestone-based foundation, then capacity-based migration waves.

Verify hyperscaler programs without assuming an incentive

Provider programs, names, eligibility rules, and available investments change. Treat any proposed credit or co-funding as a conditional commercial input until it is verified in writing through the provider or an authorized channel.

Use current official starting points:

Ask each bidder to complete the same verification schedule:

  1. Program and route: current program name, provider contact, partner route, and official source.
  2. Eligibility: customer, workload, geography, partner, consumption, and timing conditions that apply.
  3. Commercial effect: form of support, eligible costs or services, approval state, claim milestones, and expected timing.
  4. Ownership: who submits, tags, documents, monitors, and claims each benefit.
  5. Failure case: what the buyer owes and what scope changes if the provider declines, delays, or changes the support.

Keep the written confirmation in the contract file. Show the underlying services price without the incentive and state what the buyer owes if funding is declined, delayed, or changed.

Normalize proposals before you score them

Create a single comparison workbook and map every response into it. Do not compare vendor-native phase names or bundled totals directly.

Normalize these fields:

  • workload IDs, disposition, and migration-wave assignment;
  • discovery, remediation, landing-zone, migration, testing, and decommission scope;
  • buyer, supplier, hyperscaler, software vendor, and subcontractor responsibilities;
  • named roles, allocation, delivery location, availability, and replacement terms;
  • included tools, license terms, data-transfer charges, and other pass-through costs;
  • deliverables, evidence format, acceptance owner, cure period, and payment trigger;
  • schedule assumptions, buyer prerequisites, blackout dates, and critical dependencies;
  • hypercare coverage, incident path, knowledge transfer, and operational exit criteria;
  • exclusions, options, change-control mechanics, and termination assistance.

Mark every cell as included, excluded, buyer responsibility, optional, or not stated. Leave omissions as not stated until the bidder clarifies them. Return the matrix to bidders for correction before final scoring.

Worked comparison: two bids for the same workloads

Every bid, person, allocation, and coverage term below is fictional. This is an editorial example, not a supplier ranking, a real quote, or evidence of completed work. The references A-1 through A-4 and B-1 through B-4 identify the fictional response excerpts reproduced in the table and CSV.

The buyer’s common baseline has two workloads: APP01, an order-entry application and its database, and APP02, a reporting application and its database. Both require application and data acceptance, rehearsed recovery, and operations handoff. The buyer supplies an approved target environment, access, application/data reviewers, and a receiving operations team. Application refactoring, ongoing managed operations, and legacy shutdown are excluded from this example; they need separate decisions if required.

Keep each original bid status visible. The final column is the buyer’s requested common boundary, which still needs written bidder agreement and any revised price or schedule.

Workload / itemFictional Bid A responseFictional Bid B responseCommon boundary and clarification
APP01 application and database cutoverIncluded. A-1: “Application and database migration, testing, and cutover included.”Buyer responsibility for database. B-1: “Application cutover included; buyer migrates database.”Require supplier delivery of both components, with buyer data approval. Ask B to include and price the database work, or revise both bids to the same buyer-owned boundary.
APP02 application and database moveIncluded. A-1 includes both components.Database optional; application included. B-1 separates the database option.Obtain B’s database option and incorporate it into the base comparison. Do not compare A’s combined scope with B’s application-only total.
APP01 / APP02 discovery gapsBuyer responsibility. A-1 assumes the supplied dependency map is complete.Included. B-1 includes dependency validation; remediation excluded.Require a validation report and exception register from both; buyer approves remediation before a change proceeds. Ask A for the effort and schedule effect.
APP01 / APP02 rollbackNot stated. A-2 mentions a cutover runbook but gives no rollback owner or test.Included restore test. B-2: “Restore test and results included”; rollback decision authority not stated.Require supplier rehearsal evidence, data reconciliation, and latest safe decision point. Name the buyer go/no-go owner. A restore test alone does not settle cutover rollback.
APP01 / APP02 migration leadNot stated for named staffing. A-3: “Migration lead assigned”; person and allocation absent.Named staffing included. B-3: “Morgan Lee, migration lead, 60% allocation through workload acceptance”; CV and interview evidence not stated.Require name, allocation by phase, availability, relevant-work evidence, interview, and replacement controls. B supplies more staffing detail; neither response yet proves delivery capability.
APP01 / APP02 acceptance and paymentAcceptance evidence not stated. A-2: “Invoice at migration completion.”Included review step. B-2: “Invoice after buyer accepts the test pack”; review window and cure terms not stated.Define the tests, reviewer, review window, rejection/cure process, and milestone before signature. Ask A to tie the delivery invoice to accepted evidence.
APP01 / APP02 hypercareIncluded. A-4: “Five business days, business-hours coverage”; start and exit conditions not stated.Included. B-4: “Ten calendar days, round-the-clock incident coverage”; exit conditions not stated.Ask both to quote ten calendar days of incident coverage from accepted cutover with the same agreed coverage hours and timezone, response duties, extension rule, and operations acceptance. Separate routine operations from incident support.
APP01 / APP02 handoff and exceptionsIncluded document handoff. A-4: “Runbooks delivered”; receiving-team exercise not stated.Included walkthrough. B-4: “Runbooks and operator walkthrough”; demonstrated recovery and open-defect treatment not stated.Require the receiving team to complete agreed operating exercises and accept access, artifacts, and the exception register. Define who owns unresolved defects and any extension charges.

Before an award decision, Bid A needs staffing, rollback, acceptance, and coverage detail. Bid B needs a consistent database boundary and stronger rollback and handoff terms. Both need the common discovery and acceptance requirements confirmed. Keep the original responses alongside revisions so changes in scope remain visible when prices change.

Download the worked comparison and blank bid rows as CSV. Each row keeps workload IDs, inclusion status, assumptions, delivery roles, named-team evidence, acceptance owner and evidence, payment milestone, hypercare, exceptions, and clarification status. Duplicate the blank rows for each work item and bidder; replace the fictional values. Leave missing responses as not stated. Add dated proposal, attachment, or interview references as evidence arrives.

Use a scorecard tied to procurement risk

The scorecard below is an editorial starting point, not a market benchmark. Change the weights before issuing the RFP, keep their total at 100, and publish them to bidders so the decision rule does not move after proposals arrive.

DimensionStarting weightEvidence to score
Scope comprehension20Exceptions schedule, dependency treatment, assumptions, and completeness of the normalized response
Named delivery team20Relevant work by assigned people, allocation, interviews, subcontractor disclosure, and replacement control
Delivery method and quality gates20Wave method, automation, test evidence, rollback, defect process, and acceptance design
Commercial and contract fit15Risk allocation, payment triggers, pass-through transparency, change control, and termination support
Hypercare and knowledge transfer15Coverage, operational artifacts, shadowing, access transfer, and exit criteria
Funding verification10Official source, written status, eligibility conditions, claim owner, and failure-case treatment

Use scoring only after bidders confirm the same scope and resolve mandatory conditions. You can select through an acceptance and exceptions review without a numerical score.

For these weights, the suggested scale is 0 = not stated or excluded against a requirement; 1 = assertion without supporting detail; 2 = relevant detail or artifact with a material verification gap; 3 = evidence meets the published requirement and the buyer has verified it. Keep the supporting reference and reason beside each score. You can edit this scale before issuing the RFP; it is a decision aid, not measured market performance.

Calculate each weighted contribution as weight × score ÷ 3 and add the contributions for a total out of 100. To illustrate the arithmetic only (these bids have not cleared the mandatory checks), A-3 lacks a named person (score 0); B-3 states a name and allocation but lacks capability verification (score 2). At weight 20, their contributions would be 0 and 13.33. Record any unresolved mandatory condition separately; a high total cannot clear it.

Supporting references can be proposal pages, attachments, interview notes, reference responses, or contract clauses. A certification or partner tier may qualify a firm for a program, but it does not prove that the named team has delivered a comparable workload.

Run technical interviews with the people assigned

Require the proposed architect, migration lead, security lead, and operations handoff owner to attend the working session. Give them a dependency scenario, a failed cutover, and a late-discovered data constraint. Ask who decides, what evidence they need, and which artifact changes.

Capture for every named role:

  • employer or subcontractor status and delivery location;
  • allocation by phase and conflicting commitments;
  • role in comparable work, with a reference where permitted;
  • credentials only where they are relevant to assigned responsibility;
  • start availability, replacement standard, and buyer interview rights;
  • on-call and time-zone coverage during cutover and hypercare.

“Equivalent replacement” should mean equivalent against stated criteria, not a supplier’s unilateral judgment.

Write acceptance and risk allocation into the SOW

The SOW should turn the normalized proposal into delivery obligations the buyer can review and accept. Attach the inventory, assumptions schedule, responsibility matrix, architecture baseline, milestone plan, acceptance matrix, and fee schedule under version control.

SOW areaMinimum contract contentRed flag
ScopeStable workload IDs, included work, exclusions, prerequisites, and treatment of discoveriesBroad outcome language without an attached inventory
DeliverablesFormat, required contents, owner, due point, and reuse rights“Documentation” or “migration completed” without artifact definitions
AcceptanceTest method, evidence, reviewer, review window, rejection grounds, cure, and deemed-acceptance rulePayment triggered by delivery rather than accepted evidence
StaffingNamed roles, allocations, subcontractors, replacement process, and key-person treatmentSupplier can swap the core team without notice or buyer review
Security and dataAccess path, environments, data handling, incident notice, deletion or return, and evidence dutiesGeneric compliance promise with no project control mapping
CommercialsFees, invoicing unit, pass-throughs, caps or approvals, taxes, and funding treatmentBundled total with undefined tools, travel, transfer, or license charges
Change controlTrigger, impact evidence, approvers, rate or pricing rule, and emergency pathWork can proceed before written scope and commercial approval
Cutover and rollbackGo/no-go owner, triggers, latest safe decision point, communications, and state recoveryRollback is mentioned without decision authority or data reconciliation
Intellectual propertyOwnership and licenses for code, infrastructure templates, scripts, diagrams, and generated artifactsBuyer receives access only through a proprietary supplier portal
Exit and terminationTransition assistance, artifact delivery, access removal, data return, and open-work treatmentOperational dependency survives contract termination

Avoid security absolutes. The supplier should implement and evidence the controls allocated to it; the buyer still owns business risk acceptance and the controls assigned under the cloud shared-responsibility model.

Turn the comparison into SOW acceptance terms

The following is an editable starting point for the fictional APP01/APP02 comparison. It proposes terms for contracting review; it does not imply either bidder has accepted them. Replace the bracketed fields and attach the test specifications before signature.

Milestone / workload IDsSupplier output and acceptance evidenceBuyer acceptance ownerPayment and exception treatment
M1: APP01 / APP02 validationInventory and dependency validation report; each gap has an owner and proposed resolution[Program owner], with application and data ownersInvoice the agreed M1 fee after written acceptance. Newly discovered remediation needs an approved change; reporting a gap does not authorize the extra work.
M2: APP01 cutoverApplication/database test results against [test specification version]; data reconciliation with approved exceptions; cutover and rollback rehearsal record[Application owner] and [data owner]; [go/no-go owner] authorizes production cutoverInvoice the agreed APP01 delivery milestone after workload acceptance. Supplier cures failures against agreed scope; buyer accepts residual risks explicitly.
M3: APP02 cutoverThe same evidence categories, using APP02’s own test specification and reconciliation rules[Application owner] and [data owner]; [go/no-go owner] authorizes production cutoverInvoice the agreed APP02 milestone after workload acceptance. APP01 approval does not accept APP02.
M4: APP01 / APP02 operations handoffCoverage record; transferred access, repositories, runbooks, and alerts; receiving-team operating exercises; assigned open-defect and exception register[Receiving operations owner], with application owners for unresolved defectsInvoice the agreed handoff milestone after operations acceptance. State extension duties, pricing, and approvals if exit conditions remain unmet.

Copy these terms into the acceptance schedule and complete the blanks:

For each milestone, the supplier submits the evidence listed in the attached acceptance matrix to the named buyer reviewer. The buyer has [review period] to accept in writing or identify the unmet criterion and supporting evidence. The supplier corrects in-scope failures within [cure period] and resubmits. Deemed acceptance: [state the rule, notice period and treatment of unresolved defects, or “None”]. The invoice trigger is the acceptance event named in the fee schedule.

The buyer’s [go/no-go owner] retains cutover and rollback decision authority. The supplier provides the agreed rehearsal results, rollback triggers, latest safe decision point, and data-state recovery evidence before that decision. A successful restore test does not replace the workload’s rollback acceptance criteria.

Hypercare starts at [defined accepted-cutover event] for each workload and provides [coverage timezone and hours], [incident duties and response targets], and [escalation route]. The receiving operations owner accepts exit evidence. If exit is delayed, the exception register identifies the cause, owner, continuing coverage, charging rule, and approval path; elapsed days alone do not accept the handoff.

Attach a responsibility schedule naming the supplier, buyer application/data owners, platform/security owners, and receiving operations owner. Keep the fee schedule, workload inventory, and acceptance matrix on the same version so each payment trigger refers to the agreed work and evidence.

Define hypercare and knowledge-transfer exit criteria

Define hypercare coverage by workload, with a start event, support hours, incident duties, and exit criteria. Close it when the operating team accepts the agreed evidence.

The handoff pack should include:

  • target and as-built architecture, configuration decisions, and exception register;
  • source repositories, infrastructure code, pipelines, secrets-transfer procedure, and administrative access;
  • dashboards, alerts, log locations, incident routes, support contacts, and escalation paths;
  • backup, restore, failover, rollback, patching, and routine maintenance runbooks;
  • cost-allocation tags, budgets or alerts defined by the buyer, and ownership of optimization reviews;
  • known defects, accepted risks, technical debt, licenses, renewals, and vendor dependencies;
  • recorded walkthroughs, paired operations, restore or failover exercises, and named internal owners;
  • criteria and approvals for legacy shutdown, archive, data retention, and access removal.

Have the receiving team use the runbooks in agreed operating exercises before accepting the handoff. Document delivery alone does not demonstrate that the team can run the workload.

Run the selection in this order

  1. Lock the dated scope pack and bidder instructions.
  2. Find candidates in the Cloud Consulting Intel directory, then filter by platform and workload requirements.
  3. Issue the same RFP and response template to every bidder.
  4. Normalize responses and return gaps for written clarification.
  5. Verify provider programs and named-team evidence.
  6. Hold scenario interviews with the people assigned to delivery.
  7. Review acceptance and exceptions against the published criteria; if scoring, retain the evidence behind each score.
  8. Negotiate the SOW from the normalized matrix, not from the supplier’s sales summary.
  9. Reconfirm staffing, assumptions, funding status, and attachments immediately before signature.

Frequently Asked Questions

What should a cloud migration consulting RFP include?

Include the workload and dependency baseline, in-scope and excluded work, target-state assumptions, required deliverables, acceptance evidence, security and data constraints, staffing requirements, commercial response template, change-control rules, hypercare, knowledge transfer, and legacy decommission responsibilities.

How should buyers compare cloud migration proposals?

Use the same workload IDs and mark each bid item included, excluded, buyer responsibility, optional, or not stated. The worked two-bid comparison and downloadable CSV cover assumptions, roles, team evidence, acceptance, payment milestones, hypercare, and exceptions. Resolve differences in writing before comparing totals or scoring.

How should hyperscaler funding appear in a migration SOW?

Treat funding as conditional until the provider or authorized channel confirms eligibility, amount, timing, eligible spend, claim owner, and what happens if approval changes. Keep the consulting scope commercially viable without assuming an incentive will arrive.

What team evidence matters when hiring a cloud migration consultant?

Evaluate the named people assigned to the work: role, employment or subcontractor status, relevant delivery examples, current credentials where material, allocation, location and time-zone coverage, replacement controls, and participation in technical interviews.

P

Peter Korpak

Founder

Data-driven market researcher with 15+ years helping software agencies and IT organizations make evidence-based decisions. Former market research analyst at Aviva Investors and Credit Suisse. Built the 50 cloud consulting firm profiles published on cloudconsultingfirms.com from publicly available evidence.

Connect on LinkedIn

Stay ahead of cloud consulting

Quarterly directory updates, pricing benchmarks, and new research — delivered to your inbox.

No spam. Unsubscribe anytime.