The fixed-price Microsoft 365 estimate arrived before anyone inspected the system

The proposal has one price, one preferred platform and a delivery date. The estimate pack lacks a verified inventory, user journeys, dependency map, target architecture and acceptance model. A precise number without scope evidence relocates uncertainty into contingency, exclusions, variations and untested outcomes.

The situation: a precise price for an uninspected estate

COMPOSITE SCENARIO: documented product behaviour combined with recurring operational failure patterns. No client-specific delivery claim is attached. Consider a Perth member-services organisation replacing a legacy SharePoint estate used by forty staff, external assessors and several thousand members. A short request describes document migration, online applications, approvals, reporting and a public portal. After a ninety-minute demonstration, a supplier proposes a fixed price to move the service to SharePoint Online, Power Apps, Power Automate and Power Pages within six months.

The price assumes a content volume inferred from storage, ten workflows reconstructed from screenshots, standard external identity, clean member data and no material integration beyond email. Nobody has run a migration assessment, enumerated InfoPath forms, traced SOAP or BCS dependencies, sampled unique permissions, observed exception handling, reconciled records obligations, tested external-user journeys or confirmed licensing. “Like for like” appears in the scope even though the proposed platform behaves differently.

The buyer reads the fixed number as transferred risk. The supplier reads the exclusions as retained uncertainty. Once delivery begins, one form contains managed code, a payment reconciliation depends on a nightly database export, assessors share accounts, historical decisions must remain discoverable and a manual spreadsheet is the real source of status. The commercial disagreement began before the contract: both parties priced different systems under the same nouns.

Freeze commitment and recover the estimate basis

Before accepting, rejecting or negotiating the amount, request the estimate basis: scope decomposition, quantities, source evidence, assumptions, exclusions, dependencies, resource model, rates, contingency, confidence, schedule logic, licensing basis, acceptance conditions and change mechanism. Preserve the request, clarification answers, workshop notes and proposal versions. A total without this basis cannot be independently reviewed or safely updated.

Classify every statement as observed fact, measured quantity, stakeholder requirement, design decision, assumption, constraint, dependency, risk or exclusion. A navigation screenshot supplies a provisional count of ten labels. Measure triggers, actions, connections, owners, run volume and exception paths before using the count for effort. “Power Pages authentication” names a candidate implementation. The requirement must identify the users, assurance level, session rules and recovery path. Turn “migrate all records” into a source population, disposition rule, metadata map, version rule, permission rule and reconciliation tolerance before pricing acceptance.

Do not demand a free, exhaustive solution design from bidders. Define the smallest paid evidence cycle needed to decide whether to proceed and how to procure the next stage. Discovery can itself have a fixed time, price, team, inputs and deliverables. Its job is to reduce decision-critical uncertainty, expose irreducible uncertainty and produce comparable evidence. Remaining unknowns stay in the assumption register.

Decompose uncertainty into priceable planes

Separate at least eight planes. Outcome covers the policy, service and measurable value sought. Users covers roles, access needs, volumes, accessibility and high-friction journeys. Information covers sources, quality, ownership, sensitivity, records and disposition. Behaviour covers rules, exceptions, approvals and manual work. Technology covers customisations, integrations, identities, environments and non-functional constraints.

Migration covers quantities, transformations, reconciliation, downtime and coexistence. Operations covers support, monitoring, recovery, release, licensing and capability. Commercial covers responsibilities, client inputs, third parties, acceptance, warranties, risk ownership and the price mechanism. An unknown needs an owner, evidence method, decision date and treatment: resolve, prototype, bound with a range, allocate contractually or exclude explicitly.

Classify each unknown by likelihood, impact, owner and retirement test. File count can be measured by a scan; workflow complexity needs behavioural observation and dependency tracing; identity feasibility needs a controlled spike; adoption effort depends on role change and support; platform licensing requires verified usage assumptions. A single percentage contingency applied to all of them creates a precise number without a causal model.

Run discovery against the riskiest assumptions

Microsoft's migration guidance recommends assessing and remediating content before migration. SharePoint Migration Assessment Tool reports identify classes such as apps, InfoPath, BCS, large sites and other conditions that require interpretation. Microsoft specifically notes that some InfoPath forms migrate as files but still fail without remediation because of unsupported SOAP calls, managed code or identity differences. A storage total alone cannot price that work.

Join tool evidence with service evidence. Inventory sites, libraries, lists, forms, flows, custom code, databases, endpoints, identities, secrets, reports, devices, owners and obligations. Build the sample from the highest-impact and least-understood paths. Observe users completing ordinary and exception journeys. Trace the authoritative record, hand-offs and corrections. Quantify volumes and seasonal peaks. Confirm what can be retired, archived, migrated, rebuilt or temporarily rehosted.

Use proof-of-capability spikes only where a decision depends on them: the hardest external identity journey, cross-record isolation, a representative legacy integration, accessibility with real assistive technology, migration reconciliation, search retrieval and recovery. A successful prototype establishes the recorded result for its tested slice. Untested journeys remain in the risk register. A failed spike is valuable when it prevents a larger mispriced commitment.

Minimum estate capture

Run read-only inventory under the approved tenant administrator. Save module versions, operator object ID and capture UTC with the exports.

Connect-SPOService -Url https://contoso-admin.sharepoint.com
Get-SPOSite -Limit All |
  Select-Object Url,Template,StorageUsageCurrent,Owner,SharingCapability

Get-AdminPowerAppEnvironment |
  Select-Object EnvironmentName,DisplayName,EnvironmentType,Location

Join those rows to flow/app ownership, connection references, migration populations and the candidate acceptance journeys. An estimate remains gated when a production environment has no owner, a site population has no disposition, or the hardest integration cannot be observed.

Price outcomes after comparing architecture options

Reframe the request from products to outcomes and constraints. Compare viable options such as Microsoft 365 configuration, Power Platform, custom Azure services, specialist SaaS, retained legacy components or a staged hybrid. For each, record functional coverage, data and identity boundaries, accessibility, security, reliability, performance, integration, migration, licensing, support, exit and whole-of-life implications.

Microsoft Power Platform Well-Architected organises workload review across reliability, security, operational excellence, performance efficiency and experience optimisation. ISO/IEC 25010 supplies a technology-neutral product-quality model for specifying and evaluating ICT quality. These are useful completeness lenses; neither selects Power Platform or any other implementation automatically.

Carry at least one credible non-platform alternative into the decision record. If the evidence supports a simpler document-and-workflow uplift, do not price a new transactional portal. If the hardest requirements exceed low-code boundaries, do not hide custom engineering inside configuration rates. If every option retains a material unknown, price another bounded proof and withhold the full-build estimate.

Build an estimate that exposes its confidence

Create a work-breakdown structure tied to controlled deliverables and acceptance evidence. Estimate effort using quantities, reference classes and team composition; include analysis, design, engineering, migration, testing, accessibility, security, environments, release, training, operations transition and management. Model client and third-party work separately. Show licensing and consumption as verified assumptions or ranges in their own cost rows.

For each work package record basis, low/most-likely/high effort, dependency, confidence and contingency driver. Aggregate ranges with documented correlation assumptions. A sum of every worst case overstates independent risks, while the most-likely sum cannot be presented as a guarantee. State the estimate's effective date, currency, GST treatment, rate basis, schedule assumptions and validity window. Version it when evidence or scope changes.

Calibrate with completed comparable work where the organisation has trustworthy data. Track forecast versus actual at the same decomposition level, including rework and client delay, then update the reference class. A range with an explicit basis is an honest priced result. False precision is weaker: it prevents the buyer from seeing which decisions would most improve price confidence.

Match the price mechanism to residual uncertainty

Use fixed price for a genuinely bounded output where scope, quality, dependencies, acceptance, responsibilities and change rules are controlled. Current UK government sourcing guidance states the core relationship directly: fixed price requires fixed scope and is unsuitable when the output specification is unknown. Fixed-price sprints, capped time-and-materials, target cost, milestone funding and staged option contracts allocate different risks; choose the model against the documented uncertainty and risk owner.

A practical sequence is a fixed-price discovery, an optional fixed-price alpha or proof, then fixed-price delivery tranches for repeatable packages. Keep high-uncertainty integration or remediation under capped effort with explicit stop/replan points until evidence supports conversion. Price risk with the party best able to control it. The buyer owns timely decisions, access and authoritative data; the supplier owns professional method, transparent evidence and delivery within accepted boundaries.

Write variation triggers before the first dispute: quantity outside tolerance, undisclosed dependency, changed requirement, delayed client input, regulatory change, failed third-party assumption or acceptance change. A variation must cite the affected baseline, causal evidence, cost and schedule effect, alternatives and approval. It must not become the routine mechanism for completing work that the original controlled scope already required.

Release the price only when the evidence is decision-ready

The pricing-readiness gate requires named outcomes and owners; verified inventory and quantities; prioritised functional and quality requirements; architecture options and decisions; dependency and responsibility maps; migration and coexistence strategy; acceptance measures; operating model; risk and assumption registers; estimate basis; and contract-model rationale. Every unresolved item needs an explicit treatment and commercial owner.

Western Australia's procurement guidance warns that poorly defined specifications and unenforceable performance requirements undermine value for money. That principle is useful beyond government procurement, but WA rules apply only where legally in scope. The buyer should score whole-of-life value, non-cost factors and the headline implementation amount separately.

Passing the gate authorises fixed-price discussion. Price, schedule and platform remain governed by the recorded assumptions and acceptance model. It demonstrates that the parties are pricing the same controlled outcome and can explain residual uncertainty. Failing the gate has legitimate outcomes: commission discovery, narrow the scope, split the procurement, change the price mechanism, defer the decision or stop. Price certainty follows a controlled scope, output specification, acceptance set and risk allocation. Platform choice remains open until the proof resolves it.