Geode
Unlabelled cost distribution curves and bars beside a notebook on a desk
Back to Geode Perspectives
Delivery & Commercial Excellence
2026-11-04 11 min read

Building a Defensible Cost Estimate for a Digital Business Case

From initial planning ranges to risk-adjusted P90 confidence: how structured cost architecture, evidence triangulation and quantitative risk analysis establish an auditable investment baseline.

Geode logo
Geode

Advisory & Product Consulting

Share this perspective
Building a Defensible Cost Estimate for a Digital Business Case

Developing a cost model for a major digital initiative is fundamentally an exercise in supporting an informed investment decision. When public-sector agencies and enterprise organisations embark on a Detailed Business Case (DBC), the central objective is not to justify a predetermined platform or procurement preference. The business case must define the operational capability the organisation requires, evaluate viable delivery pathways, establish a defensible baseline of costs and assumptions, and recommend an accountable investment envelope.

At its core, a statewide or enterprise-level operational platform aims to help an organisation see the system as a whole, coordinate workflows across divisional boundaries, act on operational insight, and improve outcomes. Achieving that ambition requires an estimating discipline that looks well beyond the price of software. As explored in The Rise of the Digital Cost Architect, establishing a credible business-case estimate demands that leadership deconstruct early planning envelopes, test commercial evidence against delivery realities, and model uncertainty with statistical rigour.

The initial planning estimate: why vendor pricing is not the project cost

Early in an initiative's lifecycle, sponsors often rely on an indicative planning estimate to understand the order of magnitude of the required investment. These preliminary estimates frequently use commercial proposals from software suppliers as market reference points. While commercial proposals provide valuable data on licensing models, cloud hosting, and core vendor services, they represent only a fraction of the total investment needed to establish an enduring operational capability.

A commercial software quote does not account for the extensive customer-side effort required to prepare the enterprise environment, establish governance, build complex data integrations, execute multi-site rollouts across regional operating divisions, and embed new tools into frontline practice.

The dashboard below illustrates an indicative planning envelope for a multi-year enterprise platform initiative across 8 regional operating divisions, separating one-off implementation investment from recurring annual operating funding.

Initial Planning Baseline · Illustrative Example

Indicative Implementation and Annual Operating Costs

AUD, excluding GST · Not approved budget

Indicative Enterprise Platform Implementation

$18.5M–$36.5M

Indicative whole-of-enterprise project investment envelope across multi-year delivery.

Indicative Annual Operating Funding

$3.5M–$6.9Mper annum

Recurring operational funding commencing from initial production cutover.

Implementation Cost Components ($M)

ComponentRangeWhat it covers
Platform establishment & commercial mobilisation$0.7M–$2.1MInitial licensing, commercial onboarding, environments and enterprise onboarding
Vendor implementation services$4.7M–$7.5MSolution design, baseline configuration, implementation, testing and deployment support
Customer enterprise delivery team$4.1M–$8.3MProgramme delivery, governance, architecture, business analysis, procurement and assurance
Enterprise integration and data services$1.8M–$3.2MEnterprise integration, interoperability, data services and source-system connectivity
Enterprise rollout (8 operating divisions)$2.5M–$6.1MDivisional implementation coordination, local readiness, testing, cutover and go-live support
Operational change, training and adoption$1.5M–$2.7MChange management, training, communications, operational engagement and adoption
Security and assurance$0.7M–$2.1MSecurity assessment, independent testing and assurance
Implementation risk contingency$2.5M–$4.5MImplementation uncertainty, additional requirements and currently unforecast costs
Total Estimated Implementation Cost$18.5M–$36.5MIndicative overall project investment

Annual Operating Funding ($M p.a.)

ComponentAnnualWhat it covers
Hosting and SaaS subscription$1.8M–$3.2MEnterprise software subscription, cloud infrastructure and hosting
Platform operations & service management$0.5M–$1.1MService management, environment operations, monitoring and administration
Vendor support and managed services$0.5M–$1.1MVendor support, specialist operational services and service-level arrangements
Continuous enhancements & optimisation$0.7M–$1.5MOngoing enhancements, workflow optimisation and minor releases
Total Annual Operating Funding$3.5M–$6.9MRecurring each operational year

Key Estimating Assumptions

  • Estimates based on market references and comparable enterprise implementations.
  • Covers rollout across all 8 regional operational divisions.
  • Includes commercial software, vendor services, and customer delivery effort.
  • Excludes ongoing business-as-usual local divisional operations.

Executive Takeaway:

The enterprise platform requires an estimated $18.5M–$36.5M in implementation investment, with recurring operational funding of $3.5M–$6.9M per annum. A supplier quote captures only commercial software and vendor services; customer delivery management, enterprise integration, multi-site rollout, change, and risk represent the majority of customer effort.

Deconstructing the implementation cost build-up

To establish a defensible planning envelope, the total implementation effort must be broken down into discrete functional components, defining clear boundaries so that responsibilities are neither omitted nor double counted:

  • Platform establishment and commercial mobilisation ($0.7M–$2.1M): Captures one-off commercial onboarding, initial environment provisioning, enterprise tenant establishment, configuration baselines, and legal transition activities not covered by recurring subscription charges.
  • Vendor implementation services ($4.7M–$7.5M): Covers solution design, baseline configuration, supplier testing, and cutover support. Because multi-entity deployments introduce substantial local workflow variations, this allowance typically exceeds a supplier's standardised product quote.
  • Customer enterprise delivery team ($4.1M–$8.3M): A dedicated internal delivery organisation is essential to govern and manage a multi-year transformation. This team encompasses programme leadership, enterprise and solution architecture, business analysis, procurement, change oversight, data governance, and independent technical assurance.
  • Enterprise integration and data services ($1.8M–$3.2M): Connecting a new operational platform to legacy administrative and transactional systems requires substantial effort. This allowance provides for API development, event processing, master data mapping, data quality remediation, interface testing, and integration middleware infrastructure.
  • Enterprise rollout across regional divisions ($2.5M–$6.1M): Reflects the incremental local coordination required across 8 operational entities, including site readiness audits, workflow discovery, cutover coordination, user acceptance testing, and temporary operational backfill.
  • Operational change, training and adoption ($1.5M–$2.7M): An operational coordination platform changes daily frontline routines. Benefits realisation depends on comprehensive change management strategies, dedicated operational champions, role-based training programmes, and post-go-live adoption monitoring.
  • Security and assurance ($0.7M–$2.1M): Covers security assessment, independent testing and assurance activities.
  • Implementation risk contingency ($2.5M–$4.5M): An initial planning provision for emergent scope, integration complexity, technical dependencies, and dual-running requirements. At the planning stage, this is an early allowance rather than a statistically validated risk percentile.

Phasing implementation capital and recurring operational funding

A common pitfall in digital business cases is treating operational expenditure as an afterthought that begins only when implementation concludes. In a phased, multi-wave rollout across 8 operating divisions, operational funding must commence from the initial production cutover—meaning the organisation will fund both ongoing deployment waves and live platform operations concurrently.

Recurring operational costs—comprising cloud hosting, software subscriptions, enterprise service management, vendor managed services, and continuous enhancement allowances—must be modelled on their own multi-year timeline. Distinguishing early capital expenditure from enduring operational baselines ensures Treasury and enterprise finance committees understand both the cashflow profile and the long-term cost of ownership.

Developing defensible cost estimates and P90 confidence

An indicative planning range provides an initial boundary, but it cannot withstand deep commercial scrutiny or Treasury evaluation on its own. Converting a planning envelope into a defensible funding request requires a structured, multi-stage methodology aligned with recognised public-sector standards, including the Queensland Government Business Case Development Framework (BCDF) Cost Benefit Analysis Guide and the Project Cost Estimating Manual (PCEM).

The five-stage framework below outlines how component-level base estimates, triangulated evidence, quantified risk distributions, and independent validation combine to produce an auditable P50 and P90 investment position.

Governance & Methodological Framework

Developing Defensible Cost Estimates and P90 Confidence

Aligned to Government Business Case Standards

Component-level base estimates, triangulated evidence, quantified risk and independent validation produce a traceable P50/P90.

Stage 1

Establish Component-Level Base Estimates

Hybrid estimating across commercial pricing, first-principles resource models, unit rates and comparables.

Platform establishment:$1.4M
Vendor services:$6.1M
Customer delivery:$6.2M
Integration & data:$2.5M
Divisional rollout:$4.3M
Change and training:$2.1M
Security and assurance:$1.4M
Base Total:$24.0M

Base only: not P50 or risk-adjusted.

Stage 2

Triangulate and Validate Evidence

Reconcile independent comparisons against normalised equivalent scope.

Vendor A:$5.4M
Vendor B:$7.4M
Comparable C:$6.8M
Selected Reference:$6.1M

Reconcile, don't average. Justify selected figure.

Stage 3

Quantify Component Uncertainty

Define probability distributions for inherent estimating uncertainty and discrete risks.

Low $4.7MMode $6.1MHigh $7.5M

Triangular(Low, Likely, High)

Risks already in ranges are not counted again.

Stage 4

Run Quantitative Risk Analysis

Monte Carlo simulation samples component distributions and discrete risks.

Total Cost =

Σ Sampled + Σ Risks

Base:$24.0M
P50:$24.0M
P90:$25.8M

Allowance is modelled, not a flat %.

Stage 5

Validate, Assure and Report

Transparent reporting of P50, P90, confidence intervals and governance.

Base & methodology
P50 & P90 percentiles
Sensitivity drivers
Independent assurance

No generic contingency on top of P90.

Where does an RFI contribute to establishing P90 confidence?

An RFI can strengthen commercial assumptions and uncertainty ranges where it provides new, relevant and comparable supplier evidence. It does not independently establish P90 confidence. Its value should be assessed against the material cost uncertainties it can resolve.

Executive Takeaway:

A defensible P90 is developed through structured cost estimating, evidence triangulation, quantified risk and independent validation, not simply the number or range of vendor responses.

Stage 1: Establish component-level base estimates

The foundation of the cost model is a hybrid estimating approach that matches each component to its most reliable estimating method:

  • Commercial pricing and market benchmarks for software subscriptions and vendor deployment services.
  • First-principles resource models (planned roles × duration × fully burdened labour rates) for internal delivery teams and governance functions.
  • Unit-rate and complexity-based models (interface inventory × complexity tier + shared middleware services) for integration and data engineering.
  • Comparable historical project metrics for divisional rollout and change management effort.

Base estimates must reflect target scope only; they are neither risk-adjusted nor padded with informal buffers. Every component must explicitly record its scope boundaries, labour rates, physical quantities, exclusions, and evidence maturity.

Stage 2: Triangulate and validate evidence

Estimators must cross-examine commercial quotes, market benchmarks, and historical project costs against one another. The cardinal principle of defensible estimating is reconcile, don't average.

Averaging differing quotes simply conceals underlying scope variances. If Vendor A quotes $5.4M, Vendor B quotes $7.4M, and Comparable Project C cost $6.8M, the analyst must normalise all three to an equivalent scope baseline. Does Vendor A exclude customer interface testing? Does Vendor B include extended warranty support? Once scope discrepancies and evidence maturity are accounted for, the selected reference (e.g. $6.1M) represents a justified, auditable position rather than a mathematical compromise.

Stage 3: Quantify component uncertainty

Uncertainty is inherent in enterprise digital delivery, but it must be modelled systematically. For each material cost component, analysts define an inherent probability distribution—typically a triangular distribution defined by a realistic Low, Most Likely (mode), and High value.

Critically, estimators must separate inherent estimating uncertainty (the plausible variance in labour productivity or configuration effort) from discrete risk events (such as an interface redesign or regulatory delay, modelled by probability of occurrence multiplied by financial impact). Risks already accounted for within component ranges must not be duplicated in the discrete risk register.

Stage 4: Run quantitative cost risk analysis

Once component distributions and discrete risks are established, Monte Carlo simulation repeatedly samples the entire model across 10,000 or more iterations. The simulation samples correlated cost drivers simultaneously—recognising, for example, that an extended deployment timeline expands both vendor support costs and internal delivery team burn rates.

The simulation output generates a cumulative total-cost distribution. The P50 estimate represents the median (the cost at which 50% of simulated outcomes fall at or below). The P90 estimate represents the 90th percentile (90% of simulated outcomes fall at or below). The difference between the base estimate and the modelled P90 is the modelled risk allowance (P90 risk allowance = P90 − base estimate). It is a calculated, statistically derived requirement, not an arbitrary flat percentage.

Stage 5: Validate, assure and report

The final estimate must undergo independent peer review, commercial validation, and sensitivity testing before presentation to decision-makers. The business case report must transparently document:

  • The base estimate and its underlying evidence sources.
  • Modelled P50 and P90 confidence thresholds.
  • Key sensitivity and variance drivers.
  • Outstanding commercial or technical evidence gaps.

Most importantly, no generic contingency should be added on top of a modelled P90. Because P90 already accounts for inherent variance and quantified risks at a 90% confidence level, stacking arbitrary contingency percentages on top inflates funding requests and destroys executive credibility.

Applying the methodology: from planning assumptions to cost drivers

Moving from high-level planning assumptions to an evidence-based model requires deconstructing the initial planning numbers and rebuilding them from traceable cost drivers.

The pipeline below demonstrates how the planning envelope is deconstructed, validated through first-principles mathematics, and reconstituted into a risk-adjusted distribution.

Translating the Model

Applying the Methodology: From Planning Assumptions to P50/P90

Worked Estimating Mechanics
Stage 1 · Planning Estimate

Deconstruct the Planning Range

The published low and high values are not statistically derived probability percentiles, and the range midpoint does not establish P50.

ComponentLowHigh
Platform establishment$0.7M$2.1M
Vendor implementation$4.7M$7.5M
Customer delivery team$4.1M$8.3M
Integration and data$1.8M$3.2M
Divisional rollout (8 sites)$2.5M$6.1M
Change, training & adoption$1.5M$2.7M
Security and assurance$0.7M$2.1M
Base Range Sum$16.0M$32.0M
Planning risk contingency$2.5M$4.5M
Indicative Published Range$18.5M$36.5M

Arithmetic sum of component bounds. Implementation only; operating costs estimated separately.

Stages 2–3 · Evidence-Based

Rebuild from Cost Drivers

Validate each component using first-principles calculations and complexity metrics:

A. Customer Delivery Team
C_delivery = Σ (FTE_i × Months_i × MonthlyRate_i)

Rebuilt from planned role profile, duration and fully burdened labour rates.

B. Integration and Data Services
C_integration = Σ (Quantity_i × UnitCost_i) + SharedServices

Interface inventory, complexity tiers, data transformation and migration.

C. Multi-Site Enterprise Rollout
C_rollout = Σ (Effort_h × Rate_h) + SharedRolloutCosts

Reflects 8 divisional site sizes, workflow maturity and wave sequencing.

Stages 4–5 · Risk-Adjusted

Reconcile Contingency & Derive P90

Monte Carlo simulation combines component distributions and discrete risks:

C_total = Σ C_components + Σ R_incremental
  • Components sampled from justified distributions.
  • Incremental risks modelled separately where not already included in ranges.
  • Correlations between related cost drivers applied.
  • Existing planning contingency is not automatically added atop modelled P90.
P50P90

P50 = Q0.50(C_total) · P90 = Q0.90(C_total)

Executive Takeaway:

The planning estimate provides an established starting point. The business case improves its reliability by validating individual cost components, triangulating available evidence, and quantifying uncertainty to derive defensible P50 and P90 estimates.

The fallacy of sum-of-extremes

A critical insight highlighted by the methodology is that planning ranges are not percentiles. Adding the lows across the seven base components ($16.0M) and adding all the highs ($32.0M) produces arithmetic extremes. These numbers do not represent P10 and P90, and their midpoint ($24.0M) is not a P50 calculation. It happens to match the rounded P50 in this symmetric illustration; that will not generally be true.

In the real world, the probability that every single independent delivery component simultaneously encounters its absolute worst-case outcome is statistically negligible. Conversely, the probability that every component achieves its absolute lowest cost is equally remote. Evaluating investment decisions on arithmetic sums misrepresents the true probability distribution of total delivery cost.

Rebuilding components through first-principles mathematics

Rather than debating top-down range numbers, the business case rebuilds each major cost category from physical cost drivers:

  1. Customer delivery team: Rebuilt by establishing an explicit resource schedule across procurement, architecture, analysis, and assurance: Delivery cost = sum of (FTE × months × monthly rate) across all roles and phases. This exposes exactly how many full-time equivalents are required in each delivery phase, ensuring labour rates reflect total cost including contractor margins and payroll on-costs.

  2. Enterprise integration and data services: Rebuilt by categorising every external interface and data source by technical complexity: Integration cost = sum of (quantity × unit cost) across interface types + shared services. Simple REST APIs, complex bi-directional event streams, and legacy batch interfaces are priced by unit rates informed by past enterprise integrations.

  3. Multi-site regional rollout: Rebuilt by evaluating each operating division's size, operational complexity, and technical readiness: Rollout cost = sum of (effort × rate) across the eight divisions + shared rollout costs. This allows deployment sequencing to be modelled across successive deployment waves, reflecting local backfill requirements and regional cutover support.

Quantifying investment uncertainty: Monte Carlo simulation results

To evaluate how component uncertainties interact across an enterprise delivery model, we executed a 50,000-iteration Monte Carlo simulation sampling the seven base cost components ($16.0M–$32.0M base range) using independent symmetric triangular distributions, with each midpoint as its most likely value and no separate discrete risks in this illustration. A fixed random seed (522) makes the example reproducible.

The dashboard below displays the resulting total-cost probability distribution, cumulative S-curve, and variance contribution breakdown.

Quantitative Cost Risk Analysis

Quantifying Investment Uncertainty: Monte Carlo Simulation

50,000 Iterations · Illustrative Model

50,000 iterations sample the seven base cost components ($16.0M–$32.0M) from symmetric triangular distributions with midpoint modes and no separate discrete risks (seed 522). Independent components partly offset, so the modelled total-cost range is substantially narrower than the arithmetic sum of component lows and highs.

Expected Cost (Mean)

$24.0M

Mean of simulated totals · SD $1.4M

P50 Estimate (Median)

$24.0M

50% of simulated outcomes at or below

P80 Estimate

$25.2M

80% of simulated outcomes at or below

P90 Estimate

$25.8M

90% of simulated outcomes at or below

Planning Range Ref

$18.5M–$36.5M

Planning envelope (not a confidence interval)

B · Probability Distribution (Simulated Totals)

Each bin spans $1M; P90 falls within the $25M–$26M bin. Rounded frequencies exclude outcomes below $20M or at/above $28M.

Cost in $M · dashed line: P90 $25.8M · vertical grid lines mark bin boundaries.

View histogram data
Simulated total-cost histogram, AUD millions. Lower boundaries included; upper boundaries excluded.
Cost bin ($M)Frequency
20 ≤ cost < 211.3%
21 ≤ cost < 225.9%
22 ≤ cost < 2316.4%
23 ≤ cost < 2426.3%
24 ≤ cost < 2526.4%
25 ≤ cost < 2616.2%
26 ≤ cost < 276.0%
27 ≤ cost < 281.3%

C · Cumulative Probability S-Curve

Probability of not exceeding cost
  • P50: $24.0M (50%)
  • P80: $25.2M (80%)
  • P90: $25.8M (90%)
View cumulative probability data
Cumulative probability of total cost at or below each amount, AUD millions.
Cost ($M)Probability at or below
$19.9M0.1%
$20.8M1%
$21.7M5%
$22.8M20%
$24.0M50%
$25.2M80%
$25.8M90%
$26.3M95%
$27.2M99%
$29.2M100%

D · Uncertainty Drivers: Share of Total-Cost Variance

Range width, rather than component size, drives uncertainty:

Customer enterprise delivery team38.8%
Enterprise rollout (8 divisions)28.4%
Vendor implementation services17.3%
Platform establishment & mobilisation4.3%
Security and assurance4.3%
Enterprise integration and data4.3%
Change, training and adoption3.2%

Key Simulation Insights & Contingency Reconciliation

1. Portfolio cancellation effect

90% of simulated outcomes fall between $21.7M and $26.3M (P5–P95). Because independent components partly offset, the modelled range is much narrower than the $16.0M–$32.0M arithmetic sum of lows and highs.

2. Range width drives variance

Customer delivery team contributes 38.8% of variance. Regional rollout contributes 28.4% of variance—more than vendor implementation (17.3%) because their uncertainty envelopes are substantially wider.

3. Contingency comparison

Modelled P90 allowance above expected cost is $1.8M ($25.8M − $24.0M). Planning contingency ($2.5M–$4.5M) and modelled allowance are not equivalent and must not be added together without reconciling scope.

Executive Takeaway:

Monte Carlo analysis converts documented cost assumptions into a probability distribution, identifying potential investment requirements, key uncertainty drivers, and where additional evidence can most improve confidence.

The portfolio cancellation effect

The simulation demonstrates the powerful portfolio effect in capital project estimating:

  • The arithmetic sum of component lows and highs spans $16.0M to $32.0M—an uncertainty spread of $16.0M.
  • However, when simulated across 50,000 iterations, 90% of all total-cost outcomes (P5 to P95) fall between $21.7M and $26.3M—an uncertainty spread of just $4.6M.

Because independent cost components partly offset one another—cost pressures in integration being balanced by efficiencies in rollout—the modelled spread of total cost is substantially narrower than the arithmetic extremes imply. Decision-makers who rely on summed extremes overestimate volatility, while those who rely on single-point estimates ignore it entirely.

Range width, not component size, drives uncertainty

Sensitivity analysis reveals a vital principle for project sponsors: range width drives variance more than component size.

In our simulation, the customer delivery team contributed 38.8% of total project variance. Furthermore, enterprise rollout ($2.5M–$6.1M range) contributed 28.4% of variance—more than commercial vendor implementation ($4.7M–$7.5M range, contributing 17.3%). Even though vendor implementation has a comparable base cost, the wide uncertainty envelopes of the internal delivery team and multi-site rollout make them the dominant risk drivers.

This finding dictates where leadership should focus its evidence-gathering efforts. Sponsoring agencies often spend months negotiating minor licensing variations with software vendors while leaving internal delivery staffing and site rollout plans unquantified. Defensible estimating directs management attention to where uncertainty is widest.

Reconciling planning contingency against modelled P90

The simulation yields an expected mean cost of $24.0M, a median P50 of $24.0M, and a risk-adjusted P90 of $25.8M. The required risk allowance above the mean is $1.8M ($25.8M − $24.0M).

When transitioning from the initial planning estimate to the formal business case, the preliminary $2.5M–$4.5M planning contingency must be reconciled against this modelled allowance. The planning contingency was an early placeholder; the modelled P90 allowance is a statistically grounded risk requirement. Adding the planning contingency on top of the modelled P90 would double-count risk and artificially inflate the funding request by several million dollars.

Detailed Business Case maturation and the investment decision

As a Detailed Business Case nears completion, the cost model transitions from an exploratory analytical tool into an auditable investment baseline. Achieving that transition requires four concluding governance disciplines:

  1. A versioned basis of estimate: Every iteration of the model must be supported by an explicit basis of estimate document. This ledger records option boundaries, price base dates, labour rates, physical interface counts, distribution parameters, and documented rationale for all exclusions.
  2. Expenditure cashflow profiling: The total implementation envelope must be mapped onto an annual financial profile matching the procurement, design, build, and rollout schedule. Capital expenditure and recurring operational funding must be clearly delineated across financial years.
  3. Targeted market validation: A Request for Information (RFI) or early supplier engagement should be deployed selectively to resolve specific commercial evidence gaps, such as subscription pricing tiers, SaaS hosting inclusions, and implementation responsibility splits.
  4. Independent assurance and sensitivity testing: The model must be subjected to external review, benchmarked against comparable public-sector initiatives, and stress-tested against structural delivery delays.

A digital business case is ultimately an investment recommendation. By replacing top-down speculation with structured component estimating, evidence triangulation, and quantitative risk simulation, organisations provide Treasury, executive boards, and delivery teams with a cost baseline they can trust, defend, and deliver.

Topics

Investment Planning
Cost Estimation
Business Case
Risk Analysis
Digital Transformation
Geode logo
Geode

Advisory & Product Consulting

Geode is an Australian management and technology consultancy providing strategic advisory, complex program delivery, and digital product consulting across government, healthcare, and infrastructure.