Purpose
Translate a proposed analytics or AI initiative into a finance-grade business case — converting efficiency gains and revenue uplift assumptions into NPV, IRR, payback, and benefit-cost ratio, with explicit sensitivity ranges so decision-makers can see how robust the return is to optimistic assumptions. Solves the recurring problem of analytics proposals that promise value but can't survive a CFO's scrutiny.
When to Use
Invoke this skill when the user asks to:
- Estimate the ROI or business impact of a proposed analytics, data, or AI project
- Build a business case or investment justification for a steering committee or finance review
- Compare two or more competing analytics initiatives for funding prioritization
- Pressure-test the assumptions behind a vendor's or team's claimed value
- Calculate payback period, NPV, or IRR for a data/AI investment
- Decide whether an initiative clears the company's hurdle rate
Inputs Required
Before starting, collect:
- Initiative description: What the investment is (e.g., demand forecasting model, customer churn engine, BI platform migration)
- Cost inputs: One-time build/implementation cost, plus annual run/maintenance cost (licenses, cloud, FTE, vendor fees)
- Time horizon: Analysis period in years (default: 3 years)
- Discount rate / hurdle rate: Company WACC or required rate of return (default: 10% if unspecified)
- Efficiency gain estimate: Hours saved, error reduction, or cost avoided per period — with the unit cost to monetize it
- Revenue uplift assumption: Expected incremental revenue or margin gain, with the mechanism (conversion lift, churn reduction, pricing optimization)
- Ramp/adoption profile: How quickly benefits are realized (e.g., 30% Year 1, 70% Year 2, 100% Year 3)
If the discount rate, time horizon, or ramp profile are missing, apply the defaults above and flag each assumption explicitly in the output.
Instructions
Step 1: Structure the Cost Model
Separate costs into two buckets and lay them out across the time horizon:
- One-time costs (Year 0): Build, implementation, data engineering, integration, change management, initial licensing
- Recurring annual costs (Years 1–N): Run-rate licenses, cloud/compute, ongoing FTE allocation, vendor support, model retraining/maintenance
- Hold recurring costs flat unless the user provides an escalation rate; if cloud/compute scales with usage, note that as a risk to the cost line.
Step 2: Monetize the Benefits
Convert each benefit into an annual dollar figure, then apply the ramp profile:
- Efficiency gains: Hours saved × fully-loaded hourly cost, OR errors avoided × cost per error, OR process cost reduction. Always state the conversion math.
- Revenue/margin uplift: Incremental units or revenue × contribution margin — never count top-line revenue as benefit; use margin. State the mechanism (e.g., "2pt churn reduction on a $4M ARR base = $80K retained revenue × 65% margin").
- Apply adoption ramp: Multiply each benefit by the realization % for that year. Full-value-from-day-one is a credibility killer — force a ramp unless the user justifies otherwise.
Step 3: Build the Cash Flow and Run the Financials
Construct a year-by-year net cash flow (benefits minus costs) and compute:
- NPV: Sum of discounted net cash flows at the hurdle rate. Show the formula and the discounted value per year.
- IRR: The discount rate at which NPV = 0. If cash flows don't produce a clean IRR (e.g., all positive after Year 0), state that and rely on NPV/payback instead.
- Payback period: The point at which cumulative net cash flow turns positive — express in years and months.
- Benefit-Cost Ratio (BCR): PV of benefits ÷ PV of costs.
- Verdict against hurdle rate: Does IRR clear the hurdle? Is NPV positive? State plainly.
Step 4: Sensitivity Analysis (Three Scenarios)
Re-run NPV and payback under three assumption sets so the reader sees the range, not a single fragile number:
- Conservative: Benefits at ~60% of base estimate, costs at ~120%, slower ramp
- Base: The user's stated assumptions as entered
- Aggressive: Benefits at ~120%, costs at ~90%, faster ramp
Present these side by side. Then run a one-way sensitivity on the two most uncertain drivers (typically revenue uplift % and adoption ramp): show how much each can degrade before NPV goes negative — the breakeven threshold.
Step 5: Risk Flags and Recommendation
- List the 3–5 assumptions the entire case hinges on, ranked by how much they swing the result.
- Note any benefits that are soft/unverifiable and should be tracked as KPIs rather than booked as savings.
- Give a clear, opinionated funding recommendation: Fund / Fund with conditions / Defer / Decline, with one sentence of reasoning tied to the numbers.
Output Format
- Structure: Cost model table → benefit model table → cash flow table with financial metrics → sensitivity scenario table → risk flags → recommendation
- Length: 700–1,000 words plus tables
- Format: Structured document with three core tables (costs, benefits/cash flow, sensitivity) and bulleted risk/recommendation sections
- Sections: Assumptions & Defaults → Cost Model → Benefit Model → Financial Summary (NPV, IRR, Payback, BCR) → Sensitivity Analysis → Key Risk Assumptions → Funding Recommendation
- Tone: Finance business partner — numerate, skeptical of optimistic claims, decision-oriented
Quality Checklist
Before delivering output, verify:
- [ ] Every benefit dollar figure shows its conversion math (units × unit value), not just a stated total
- [ ] Revenue benefits use contribution margin, not gross top-line revenue
- [ ] An adoption ramp is applied — benefits do not hit 100% in Year 1 unless explicitly justified
- [ ] NPV, IRR, payback, and BCR are all reported with the discount rate and time horizon stated
- [ ] Sensitivity table shows all three scenarios (conservative/base/aggressive) with NPV and payback for each
- [ ] At least one breakeven threshold is identified (the point where a key driver turns NPV negative)
- [ ] A single clear funding verdict is given (no hedging across multiple options)
Notes
- The most common failure in analytics business cases is double-counting: the same FTE hours saved are claimed as both cost reduction and capacity for revenue work. Pick one and note the other as upside.
- Soft benefits (better decisions, faster insight) are real but should be listed as strategic rationale, not monetized — monetizing them invites a CFO to discount the whole case.
- When comparing competing initiatives, rank by NPV for absolute value and by BCR for capital efficiency; they often disagree, and that disagreement is the decision.
- If the user has no revenue uplift mechanism — only cost savings — say so and build a cost-avoidance-only case; padding it with speculative revenue weakens credibility.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills