ROI is the most quoted number in business and one of the most casually abused. The formula is a single division, yet two people can compute “the ROI” of the same project and get different answers, both technically correct, because they disagreed silently about what counts as gain, what counts as cost, and over how long. This guide pins all three down, with worked examples you can re-run in our free ROI calculator.
In this guide
The formula
ROI = (gain − cost) / cost, usually shown as a percentage.
Spend $5,000 on a project that brings back $8,000: ROI = (8,000 − 5,000) / 5,000 = 60%. Positive ROI means the project returned more than it consumed, negative means it destroyed money, and 0% means it exactly paid for itself. So far, one division. Every complication that follows comes from deciding what the two inputs really are.
What counts as gain, what counts as cost
The most common inflation of ROI is using revenue as the gain. Revenue is not gain; what the project actually returned is profit after the costs the revenue dragged along with it (goods, fulfillment, fees). On the cost side, the equivalent mistake is counting only the obvious invoice and skipping labor, software, or your own time. The honest version of both inputs is:
- Gain = incremental profit attributable to the project, not gross revenue.
- Cost = everything the project consumed, including the costs that do not arrive as invoices.
Margin is the bridge between revenue and gain, and if margin vs markup is fuzzy, fix that first with our margin guide, because an ROI built on the wrong margin is wrong by the same amount.
The missing ingredient: time
Plain ROI ignores duration, and duration changes everything. A 100% ROI sounds like it beats a 30% ROI, but if the 100% took ten years and the 30% took two:
- 100% over 10 years = (1 + 1.00)1/10 − 1 = 7.18% per year
- 30% over 2 years = (1 + 0.30)1/2 − 1 = 14.02% per year
The “smaller” return compounds nearly twice as fast. The general conversion is annualized ROI = (1 + ROI)1/years − 1, and it should be applied any time two opportunities run on different clocks. A 50% ROI over three years, for instance, is 14.47% a year. Annualized figures also plug directly into compounding projections, which is where the investment calculator takes over.
ROI vs ROAS in marketing
Ad platforms report ROAS, return on ad spend, which is revenue divided by spend. A campaign that spends $2,000 and generates $8,000 of revenue has a ROAS of 4. That sounds spectacular until margin enters: at a 40% gross margin, that revenue carries $3,200 of gross profit, so the ROI is (3,200 − 2,000) / 2,000 = 60%. Excellent, but a very different number than “4×”. The practical rule: a campaign is profitable only when ROAS exceeds 1 / margin. At a 40% margin the break-even ROAS is 2.5, and any agency quoting ROAS without asking about your margin is reporting someone else’s success metric.
Three worked examples
| Scenario | Inputs | ROI |
|---|---|---|
| Marketing campaign | $5,000 spend, $8,000 incremental gross profit | 60% |
| Stock position | Bought $10,000, sold $13,100, $400 dividends, $60 fees | (13,100 + 400 − 60 − 10,000) / 10,000 = 34.4% |
| Equipment purchase | $12,000 machine saving $400/month for 3 years = $14,400 | 20% total, 6.3% annualized |
The stock row shows the full-income version: gains include everything received (sale proceeds plus dividends) minus everything paid (purchase plus fees). The equipment row shows why annualizing matters: 20% looks healthy until it is spread across three years.
Where ROI misleads
- It ignores risk. A 60% expected ROI with a real chance of total loss is not better than a certain 20%. ROI ranks outcomes, not probabilities.
- It ignores scale. A 300% ROI on a $100 experiment earns $300; a 15% ROI on $100,000 earns $15,000. Percentages do not pay bills, profits do.
- It invites cherry-picked windows. Measured from the bottom of a dip, almost anything has heroic ROI. Fix the measurement window before computing, not after.
- It hides ongoing costs. Projects with maintenance, renewals, or churn need their ROI recomputed over the full life, not the launch quarter.
None of this makes ROI useless; it makes ROI a starting question. For decisions with long horizons and compounding, follow it up with the time-aware math in our compound interest guide, and for debt-vs-invest decisions, remember that avoiding interest is a guaranteed return, as the numbers in our mortgage guide show.
Frequently asked questions
Is a good ROI 10%, 50%, 100%?
There is no universal threshold; it depends on the alternatives and the risk. The honest comparison is always against your next best option of similar risk, annualized over the same period. A 9% annualized ROI is poor for a risky venture and excellent for a near-certain process improvement.
What is the difference between ROI and IRR?
IRR handles projects with many cash flows at different times and finds the single rate that makes them balance. ROI is the simple end-to-end version. For one outlay and one payoff they tell the same story; for multi-year cash flow streams IRR is the more truthful tool.
Should I subtract taxes when computing ROI?
If you are comparing options that are taxed differently, yes, compare after-tax to after-tax. If the options share the same tax treatment, pre-tax comparisons rank them identically and are simpler to build.
Can ROI be below −100%?
Only when a project consumes more than its original cost, for example an investment with follow-up liabilities. For a simple purchase the floor is −100%, total loss of what you put in.