“NPV vs IRR vs Payback Period: Which Should You Use for Project Selection?”

What is the difference between NPV, IRR, and payback period?

npv vs irr vs payback period is a key topic in modern project management and teamwork. NPV measures value created in currency, IRR measures the annualised return as a percentage, and payback period measures time to recover the investment. They use the same cash flow forecast but answer different questions: how much, at what rate, and how fast.

That distinction matters because a project can be excellent on one measure and mediocre on another without any of the numbers being wrong.

NPV: the value test

NPV discounts every future cash flow back to today at your required rate of return, then subtracts the initial investment.

NPV = Σ [CFₜ ÷ (1 + r)ᵗ] − Initial investment

If NPV is positive, the project earns more than the required return and creates value. If it is negative, the project destroys value even when it looks profitable on a cash basis, because the capital could have earned the hurdle rate elsewhere.

The discount rate is the part teams argue about most. It is usually the weighted average cost of capital plus a premium for project-specific risk. A safe efficiency project might be discounted at 8%; a first-of-its-kind market launch in a new country might need 18%. Changing that input changes the answer, which is why the rate should be agreed before proposals are scored, not after.

IRR: the return test

IRR is the discount rate that makes NPV exactly zero. You do not solve it algebraically for most cash flows; spreadsheets iterate to it.

Decision rule: accept if IRR exceeds the hurdle rate. It is popular because “this project returns 18% against a 12% hurdle” is a sentence executives understand immediately, while “NPV of $168,000” requires a follow-up question about scale.

The trouble is that IRR is a ratio-like measure. It says nothing about how much capital is deployed, and it assumes interim cash flows are reinvested at the IRR itself — an assumption that becomes absurd when the IRR is 40%.

Payback: the recovery test

Payback is the time until cumulative cash inflows equal the initial outlay. For an even cash flow stream, divide the investment by the annual inflow.

It ignores everything after the payback date and ignores the time value of money. Those are serious flaws for a decision metric — and completely acceptable for a risk screen. If a project’s payback is 18 months in a market that may shift in 24, that is useful information no discount rate will give you.

A worked example: one project, three answers

Take a $250,000 automation project with expected savings of $70,000 a year for five years, discounted at 10%.

Read the four numbers together and the project is understandable in a way no single metric allows. It creates about $15,000 of value, earns roughly 2.4 points above the 10% hurdle, takes three and a half years to return the cash, and almost five years to return the value. Margin of safety: thin. A 10% shortfall in annual savings drops NPV to roughly zero.

That last point is the most useful sentence in the analysis, and none of the four metrics states it directly. Sensitivity testing does.

Why do NPV and IRR disagree?

They disagree for exactly two reasons: projects of different size, and cash flows with different timing. When either is present, the two measures can rank the same pair of projects in opposite orders — and the conflict is not a calculation error.

Consider two mutually exclusive options at a 10% discount rate.

Project X wins on IRR and payback. Project Y wins on NPV by a wide margin — $119,407 more value. If you can only fund one and your objective is to maximise value, Y is correct despite the lower percentage return.

The crossover rate here is about 23%. Above a 23% hurdle, X’s NPV exceeds Y’s; below it, Y wins. Since the hurdle is 10%, Y is the answer. That crossover rate is the single most useful number in the comparison, because it tells you how wrong the discount rate would have to be before the recommendation flips.

Timing produces the same effect. A project that returns cash early has a higher IRR for the same total cash, because IRR compounds the early inflows more aggressively. NPV at a realistic rate does not reward that front-loading as heavily.

> Rule: for mutually exclusive projects, rank by NPV. For independent projects under a capital constraint, rank by profitability index or NPV per unit of scarce resource.

When does IRR fail completely?

IRR has three failure modes, and two of them are common enough to matter in ordinary project portfolios.

Non-conventional cash flows. If a project’s cash flow changes sign more than once — a build followed by a decommissioning cost, or a loan with a balloon repayment — the mathematics can produce more than one IRR. A stream of −$200,000, +$1,100,000, −$1,000,000 has two roots, roughly 15% and 335%. Neither is wrong; the metric has simply stopped identifying a unique answer. The standard fix is to use NPV at the actual cost of capital, or MIRR.

The reinvestment assumption. IRR implies interim cash flows are reinvested at the IRR. For a project showing 36.9%, that means assuming you can redeploy every returned dollar at 36.9% for the remaining life. Most organisations cannot.

Scale blindness. A 300% IRR on a $20,000 pilot is a rounding error in a $50 million capital plan. Percentage returns do not aggregate.

MIRR addresses the first and second problems. It compounds positive cash flows forward at a stated reinvestment rate, discounts negative flows at the finance rate, and solves for a single rate.

For the automation example at a 10% reinvestment rate: the five $70,000 inflows grow to $427,357, and MIRR = ($427,357 ÷ $250,000)^(1/5) − 1 = 11.3%. That is a more defensible number than 12.4%, because it does not assume the company can reinvest at 12.4%.

What about payback and discounted payback?

Payback’s weakness is structural: it stops measuring at the recovery date. Two projects with identical 2.5-year paybacks can differ by millions in total value, because one keeps generating cash for a decade while the other stops.

Discounted payback fixes the time-value problem but keeps the cutoff. In the automation example, simple payback is 3.57 years and discounted payback is 4.65 years — a full year of difference that matters if the equipment has a five-year life.

Where payback genuinely earns its place:

• Cash-constrained periods. When the business needs capital back quickly, recovery speed is a real constraint, not a proxy for quality.
• High-uncertainty markets. The further out the forecast, the less it deserves weight. Payback concentrates attention on the near term.
• Regulated or policy-driven investments with mandated maximum payback thresholds.
• As a tie-breaker between projects whose NPVs are within measurement error of each other.

Set the threshold policy in advance. A common pattern is a maximum payback of three years for technology investments and five for infrastructure, with an explicit exception route for strategic projects. Without a stated threshold, payback becomes an argument about whose optimism is more persuasive.

Which measure should decide the project?

Use them in a fixed sequence, and write the sequence into your selection policy so that every proposal is judged the same way.

The order matters more than the metrics. Teams that start with IRR end up defending a percentage; teams that start with NPV end up defending a value figure, which is a much better argument to have with a CFO.

One more rule: never let a metric override a strategic requirement. A compliance project with negative NPV may still be mandatory. Say so explicitly in the business case rather than manipulating the inputs to force a positive number.

How does this fit a stage-gate process?

Financial selection is not a one-time event. In a stage-gate model, each gate should re-run the analysis with updated cash flows and a re-validated discount rate.

Practical gate discipline looks like this:

• Gate 1 (concept): rough order-of-magnitude NPV and payback, clearly labelled as estimates with a stated accuracy range.
• Gate 2 (business case): firm NPV at the approved hurdle, IRR and MIRR for transparency, payback against policy.
• Gate 3 (execution approval): confirm the cash flow forecast still holds; re-run sensitivity on the two assumptions that drive most of the value.
• Post-implementation: compare realised cash flows with the business case. This is the step almost everyone skips, and it is the only one that improves future estimates.

Sunk costs must be excluded at every gate. Money already spent is irrelevant to whether the remaining investment is worth continuing — but emotionally it is the hardest rule to follow, which is why failing projects rarely get stopped.

What mistakes do teams make with these metrics?

Comparing IRR across project sizes. A 45% IRR on a $50,000 project and a 22% IRR on $5 million are not comparable. Rank by value.

Using one discount rate for every project. A regulated upgrade and a market expansion do not carry the same risk. Using the corporate WACC for both systematically over-funds risky projects.

Treating the forecast as data. Every metric inherits the error in the cash flow estimate. If the revenue forecast is a straight line with no ramp-up, the NPV is fiction with three decimal places.

Ignoring working capital and tax timing. Cash flows are not accounting profit. A project that bills quarterly, pays suppliers in 30 days, and books revenue on delivery has a materially different cash profile from the one in the spreadsheet.

Quoting payback as if it were profitability. “It pays back in 14 months” says nothing about whether it creates value after month 15.

Forgetting the option value of staging. A pilot that costs $80,000 and buys the right to invest $2 million later is often worth more than its standalone NPV suggests. Real options analysis is the formal version, but the practical version is simply asking what decision the pilot preserves.

> Common mistake: letting the finance model become the whole argument. The best project selection meetings spend half their time on the numbers and half on execution risk, capability, and strategic fit — the things a discount rate cannot capture.

Do you need a finance team to run this analysis?

No, but you do need three inputs that are frequently missing in project teams: a cash flow forecast by period, an agreed hurdle rate, and a record of what actually happened on comparable projects.

A spreadsheet handles the arithmetic in minutes. The hard part is keeping the forecast and the delivery plan in one place, so that when the schedule slips six weeks the cash flow model reflects it. When the business case lives in a finance folder and the delivery plan lives in a task board, the analysis ages badly and gate reviews become archaeology.

That is the gap Doitify is designed to close: milestones, budgets, dependencies, and progress reporting on the same data, so the cost and benefit assumptions behind a business case can be revisited with current information. We build the product, so weigh that accordingly — for a single small investment, a spreadsheet and a calendar reminder are entirely sufficient.

Conclusion

NPV, IRR, and payback period are not competing answers to the same question. NPV asks how much value the project creates, IRR asks what rate it earns, and payback asks how quickly the capital comes back. Only the first is a complete decision rule.

Use them in that order. Screen for strategic fit, accept or reject on NPV at a documented hurdle rate, test risk with payback, and communicate with IRR and MIRR while being honest that IRR assumes reinvestment at its own rate. Then do the thing most organisations skip: compare realised cash flows with the business case after delivery. That single habit is what turns a financial model from a funding ritual into a forecasting capability.

If this post on npv vs irr vs payback period was helpful, you might also enjoy Free Resource Plan Template and How to Successfully Switch Project Management Tools.