Return on Investment (ROI): The Four Decisions Behind the Number

Return on Investment (ROI) compares what an activity returned with what it cost to produce that return. The arithmetic is not the hard part. Divide return by cost, multiply by one hundred, and a percentage appears. What it never shows is the four decisions behind it: what counted as return, which costs went into the denominator, what period was measured, and whether the result was credited to the activity or shown to have been caused by it. Change one and the same campaign can report thousands of percent, or a loss.

What Return on Investment Actually Measures

The Marketing Accountability Standards Board, in its Universal Marketing Dictionary entry for ROI, describes the metric as one way of relating profit to capital invested, measuring per period rates of return on money invested in a business. Two details carry the weight. The numerator is profit, not revenue. And ROI is read against an expected or required rate of return, so it delivers no verdict by itself.

ROI is a ratio, not a grade. The same 40 percent reads well against a low required rate and badly beside a better alternative.

The same entry notes that ROA, RONA, ROC and ROIC differ mainly in how investment is defined. A standards body is saying, in its own reference work, that the denominator is a choice.

The ROI Formula and Its Accepted Variants

The core construction

Return on investment (%) = Net profit / Investment x 100

Net profit on top. Capital invested underneath. A positive result means measured return exceeded measured cost over the period reviewed. The same structure sits behind Arfadia's SEO ROI calculator.

Gain, ending value, and net income

Other constructions are in legitimate use. The Corporate Finance Institute presents ROI as investment gain over investment base, as net income over original capital cost, and in the value form where initial cost is subtracted from final value.

They are not interchangeable. A gain can be realized through a sale or unrealized on an asset still held, so two analysts can compute different figures for one position. Net income sits after interest and tax, importing financing and tax policy into an operating ratio. No variant is wrong. The label alone is insufficient.

Annualized return, and where that rule belongs

Simple ROI carries no time term, so one percentage can describe three months or ten years. Annualized ROI takes ending value over beginning value, raises it to the power of one over the years held, then subtracts one.

A stricter rule exists, with a boundary worth respecting. Under the CFA Institute's Global Investment Performance Standards, returns for periods shorter than one year must not be annualized. That governs investment performance reporting by GIPS compliant firms and does not bind marketers. The reasoning still travels: annualizing four weeks extrapolates a noisy window into a claim about a year.

Defining Return, Cost, and Period Before You Calculate

Most of the damage happens before anyone touches a calculator. Three definitions come first, and each is a choice.

What counts as return. Revenue is not profit. Sales, gross profit, operating profit, net profit and contribution each produce a different ROI from identical trading. A revenue numerator describes sales efficiency, not economic return.

What counts as cost. The MASB entry on Return on Ad Spend shows how wide this boundary legitimately runs, from ad creation and media placement only, out to salaries of in-house and contracted staff, vendor and affiliate costs. Two honest analysts on one campaign can publish very different figures without either doing anything improper. Agency retainers sit in the same denominator, so check published service pricing first.

What period is measured. The IAB and MRC Retail Media Measurement Guidelines of January 2024 require attribution windows to be empirically supported, logical for the purchase cycle, disclosed before the campaign runs, then applied consistently. They record 3, 7, 14, 28 or 30 days as typical digital practice and decline to prescribe any. That document is scoped to retail media, so treat it as the clearest statement of the principle rather than a universal mandate. Long cycles in B2B marketing programs are the usual reason a default window misleads.

The window is a choice. A number without one is not checkable.

How Marketing ROI Is Constructed Differently

Marketing ROI is not general ROI with marketing vocabulary attached. The MASB entry for Marketing Return on Investment, also called ROMI, defines it as contribution to profit attributable to marketing, net of marketing spending, over the marketing invested:

MROI (%) = 100 x [ (Incremental revenue attributable to marketing x Contribution margin %) - Marketing spending ] / Marketing spending

Notice what the standards body built in. The revenue term is incremental, not total and not attributed. It is multiplied by contribution margin, revenue less variable costs, so the numerator is profit contribution rather than sales. Only then is spending subtracted. Anything weaker is a simplification.

One arithmetic trap deserves a warning. Spending is subtracted in the numerator because the numerator to that point is gross of marketing spend. Swap in a figure already net of that expense, such as net profit, subtract again, and the same cost has been removed twice.

Marketing funds are risked in the current period rather than tied up in plant and inventory, so a marketing ROI and a factory ROI are not two readings on one scale.

ROI and ROAS Are Not the Same Ratio

ROAS divides sales attributable to ads by the cost of those ads. MASB adds that the sales term may be incremental sales or exposed sales depending on the need, which is a lot of flexibility for a metric quoted to two decimals.

Google Ads describes target ROAS as the average conversion value wanted per dollar of spend, with a worked example of five dollars in sales over one dollar of spend giving a 500 percent target. Conversion values are advertiser supplied and follow the account's attribution settings, so platform ROAS is an optimization signal first, an accounting figure second. Read any search engine marketing services report the same way.

ROAS is not a profitability measure. Converting it into ROI needs contribution margin and the full cost base, and both are usually missing from the report in front of you. A ROAS of three can sit on either side of break even on margin alone.

Incremental ROAS exists because sales not attributable to the advertising were being counted anyway, overstating the contribution. Calculated correctly, MASB says, ROAS and incremental ROAS are the same metric. Platforms report them as two numbers, and Google's Conversion Lift documentation defines incremental return on ad spend as incremental conversion value over total spend, measured by experiment rather than attribution rules.

ROI, MROI, ROAS, and Incremental ROAS Compared

Attributed Return Versus Incremental Return

This is where marketing ROI usually goes wrong, and the strongest evidence comes from inside the attribution field. The IAB and MRC guidelines define attribution as assigning credit for consumer actions, such as sales or visits, to specific marketing efforts. Google Analytics documentation describes an attribution model as a rule, a set of rules, or an algorithm that distributes credit across touchpoints. Both describe credit assignment. Neither describes causation.

The same guidelines then concede the central problem. The goal of attribution is the difference between how an exposed group behaved and how it would have behaved unexposed, which cannot be observed, because one person cannot be both exposed and unexposed. Every approach approximates a counterfactual it can never see. The document adds that attribution can be biased by data availability, so a digital only method may credit digital for outcomes driven by television.

Attribution is bookkeeping, not proof. It is still useful within one fixed model and window.

Incrementality asks the harder question: what happened because of the activity. The IAB and MMA guidebook states that simple weight based allocations such as first click, last click or time dependent weights do not get to true incrementality. The methods named for closing that gap are randomized controlled trials, preferred where feasible, plus synthetic controls, matched market tests and uplift models. A holdout group is the operational counterfactual, and a geo experiment applies that logic at market level.

What happened when one company measured both ways

The clearest demonstration comes from a field experiment at eBay, reported by Blake, Nosko and Tadelis in Econometrica in 2015. They switched paid search off instead of modeling it. With brand search suspended, roughly 99.5 percent of the click traffic was retained through natural search. In a non brand test across 68 treated and 65 matched control US media markets over 60 days, attributed sales in the test markets fell by more than 72 percent, while the estimated effect on actual sales was 0.66 percent, with a 95 percent confidence interval from -0.42 percent to 1.74 percent. For one channel, one company and one period, ROI came out at 4,173 percent by ordinary least squares, 1,632 percent with fixed effects, and -63 percent experimentally, with an interval of -124 percent to -3 percent.

That gap is not a rounding error.

The caveats are not optional. The experiments ran in 2012 and paid search has changed since. The result is short run, and the authors do not rule out longer term effects. eBay is a large established brand with high organic visibility, which is exactly why suspended brand ads were absorbed by organic results. Spend and revenue were estimated from public figures because actuals were proprietary. Effects varied sharply across customer groups. It does not generalize to every brand or channel.

Is There a Good ROI?

No universal benchmark for a good ROI exists in any defensible source, and the structure of the metric explains why. This is not a gap in the research. It is a property of the ratio.

Margin sets the break even point, since a revenue based ratio needs a different level to clear costs in a high margin category than in a low one. Cost boundary sets the denominator. Time horizon decides what got counted. Method decides the answer, as the eBay figures make clear. Risk decides whether a point estimate means anything. And the required rate of return sets the threshold, which is firm specific because it depends on cost of capital and alternatives.

So the honest comparators are internal: your own required rate of return, and your own earlier result on an unchanged method. A rule of thumb that skips margin, method, window and cost boundary is a number without units.

Where ROI Falls Short

Time. The formula has no time term and no discounting, so a return in year five counts the same as one today.

Risk and precision. ROI is a point estimate with no dispersion attached. Across 25 large field experiments with major US retailers and brokerages, covering about 2.8 million dollars of ad spend, Lewis and Rao found in The Quarterly Journal of Economics that the median confidence interval on ROI ran over 100 percentage points wide. That holds both a strong profit and a serious loss.

Scale. A percentage discards magnitude. A very high return on a tiny budget produces less cash than a modest return on a large one.

Incomplete costs and opportunity cost. Anything left out of the denominator inflates the ratio invisibly, and the ratio says nothing about the alternative you passed up.

Forecast uncertainty. The retail media guidelines record that when a shorter sales period is measured, providers commonly forecast the full twelve month effect, and require the basis of that forecast to be disclosed. Some Conversion Lift studies include modeled delayed conversions by default. Label the forecast as a forecast.

Long term effects and incompatible methods. A window truncates carryover, so a short one understates a slow burning effect while a long one absorbs outcomes the campaign had nothing to do with. Slow compounding work such as search engine optimization services suffers most here. Attributed ROAS, measured incremental ROAS and modeled marketing ROI are different estimands, so setting them side by side implies a shared scale that does not exist.

Payback period, customer acquisition cost, lifetime value, net present value and internal rate of return each answer something ROI cannot.

Reporting an ROI Figure That Can Be Checked

A percentage alone is not reviewable. Five labels fix that.

1. Return definition. Profit, contribution, gain, net income or revenue.
2. Cost boundary. What sits inside the denominator, and what does not.
3. Measurement period. The window, fitted to the purchase cycle and disclosed in advance.
4. Attribution basis. Credited by a model, or estimated by an experiment.
5. Measured or modeled. Whether any part is projected beyond the observed window.

State those five and the number becomes checkable. They also work when you evaluate a GEO agency.

Frequently Asked Questions

What is return on investment?

Return on investment is a ratio relating the return produced by an activity to the cost of producing it, usually shown as a percentage. MASB describes it as relating profit to capital invested, read against a required rate of return.

How do you calculate ROI?

The standards body construction divides net profit by the investment and multiplies by 100. Finance references also use gain over investment base, and the value form where initial cost is subtracted from final value. These do not produce the same number, so state which you used.

What is the difference between ROI and ROAS?

ROI compares profit with the investment that produced it. ROAS compares sales attributable to ads with the cost of those ads. ROAS has a revenue numerator and a cost term that may cover media only or extend to staff and vendor costs, so it measures efficiency, not profitability.

What is a good ROI?

There is no defensible universal answer. A threshold would have to hold across different margins, cost boundaries, windows, methods and required rates of return, none of which are constant between firms.

Does attributed revenue prove a campaign caused the sale?

No. Attribution assigns credit for an observed outcome according to a model, and the IAB and MRC guidelines state plainly that the counterfactual cannot be observed. Estimating causation needs a control group, through an experiment, a holdout or a geo test.

Related Terms

Marketing ROI (MROI), Return on Ad Spend (ROAS), Attribution, Incrementality, Contribution Margin.

References and Sources

1. MASB, Return on Investment
2. MASB, Marketing Return on Investment (MROI)
3. MASB, Return on Ad Spend (ROAS)
4. Corporate Finance Institute, ROI Formula
5. CFA Institute, GIPS Standards Handbook for Firms, 2020
6. IAB and MRC, Retail Media Measurement Guidelines, January 2024
7. IAB and MMA, MMM and MTA Guidebook, November 2019
8. Google Ads Help, About target ROAS bidding
9. Google Ads Help, About Conversion Lift
10. Google Analytics Help, Get started with attribution
11. Blake, Nosko and Tadelis, Econometrica 83(1), 2015
12. Lewis and Rao, Quarterly Journal of Economics 130(4), 2015