ROI Measurement
A practical framework for measuring ecommerce ROI: which profit number goes on top, which cost base goes on the bottom, and how to avoid the common mistakes that make the same campaign look like a 4× win or a 30% loss.
ROI Measurement
ROI measurement is the methodology for computing return on investment — choosing the right profit numerator, cost denominator, and time window for a given decision.
Return on investment looks like a single ratio, but in practice operators use the word to mean a dozen different things. Some divide gross profit by ad spend. Others divide contribution margin by fully-loaded marketing cost. A finance lead might mean net profit over total operating expenses. The arithmetic is trivial; the methodology is where mistakes compound.
ROI measurement is the discipline of making those choices explicitly: what costs you include, what profit layer you measure against, what window you measure over, and how you handle returns, discounts, and shared overhead. Done well, it gives you one number that actually drives decisions. Done sloppily, it hides losing campaigns behind flattering ratios.
Most stores have an ROI number on a dashboard somewhere. Very few can tell you, in one sentence, what's in the numerator and what's in the denominator. That gap is where bad budget decisions get made — campaigns get scaled because the ratio looks good, when really the methodology was hiding the cost of returns, the platform fees, or the agency retainer.
The framework below splits the problem into four choices: which profit layer goes on top, which cost base goes on the bottom, what time window you measure, and how you attribute across channels. Get those four right and your ROI numbers stop arguing with your P&L.
1. Choosing the numerator: which profit layer?
Revenue is the wrong numerator. Revenue divided by ad spend is ROAS, not ROI — see ROI vs ROAS for the full contrast. ROI is a profit ratio, so the question is which profit: gross profit, contribution margin, or net profit. Each tells a different story and answers a different question.
Contribution margin is usually the right default for an ecommerce store evaluating a marketing investment. It's revenue minus COGS, payment processing, fulfilment, and returns — every cost that scales with the order. It's the cleanest measure of whether the next euro of ad spend actually generates incremental profit, before you pile on fixed overhead the campaign didn't cause.
2. Choosing the denominator: which cost base?
The denominator question is even more contested. The narrowest version is platform spend only — what you paid Meta, Google, or TikTok. The broadest is fully-loaded marketing cost: platform spend plus agency fees, creative production, affiliate commissions, influencer payments, and the salary of whoever runs the channel.
If you're benchmarking a single campaign, platform spend is fine. If you're deciding whether your marketing function as a whole earns its keep, you need the loaded number. The same Meta campaign can show a 3.2× ROI on platform spend and 1.4× once you include the €4k/month retainer and the freelance editor — and the second number is the one that matches your bank account.
The most common ROI mistake
Mixing layers between numerator and denominator. If you put contribution margin on top, you must put marketing cost (not COGS, fulfilment, or fixed overhead) on the bottom — those costs are already subtracted out of the numerator. Double-counting them collapses the ratio and makes profitable campaigns look like losers. Pick one accounting layer and stay inside it.
3. Time horizon and customer lifetime
ROI measured on the first order will systematically understate the value of acquisition campaigns for any store with repeat purchase. An apparel brand with a 35% 90-day repeat rate looks unprofitable on first-order ROI and healthy on 12-month ROI. Pick the window that matches the decision: short windows for performance optimisation, longer windows for budget-level decisions.
Two practical horizons cover most cases. A 30-day window gives you fast feedback for paid-channel tuning and matches Meta's default attribution. A 12-month window gives you the truth for annual planning and channel-mix decisions — it folds in second orders, subscriptions, and the return rate that only shows up after 60 days. ROI Benchmarks by Industry is built on the 12-month view for exactly this reason.
Same campaign, five different ROI numbers
4. Channel attribution and blended ROI
Single-channel ROI overstates winners and understates losers because platforms claim credit for orders they only assisted. Meta's in-platform ROAS routinely shows 1.4–2.0× higher than the same campaign measured against your blended numbers. The fix is to compute blended ROI — total contribution margin divided by total marketing spend across all channels — and use it as the truth-teller above per-channel reports.
Use the Marketing ROI Calculator to run both views side by side. The per-channel numbers help you reallocate budget; the blended number tells you whether the whole marketing engine is profitable. When the two diverge sharply — channel ROI up, blended ROI flat — you're seeing attribution inflation, not real growth.
ROI measurement: frequently asked questions
ROAS is revenue divided by ad spend — a top-line efficiency metric. ROI is profit divided by cost, so it accounts for margin, returns, and fees. A 4× ROAS on a 25% margin product is a 1× ROI before any other costs. See ROI vs ROAS for the full breakdown.
Contribution margin is the better default for ecommerce because it strips out fulfilment, payment fees, and returns — all of which scale with order volume. Gross profit (revenue minus COGS only) overstates profitability for stores with high fulfilment cost or return rates.
For campaign-level optimisation, no — use platform spend so you can compare channels apples-to-apples. For channel-mix and budget decisions, yes — include the loaded cost so you're measuring what the marketing function actually consumes.
Subtract returned orders from revenue and refund the COGS — both belong in the contribution margin calculation. For categories with high return rates (apparel often runs 20–30%), measuring ROI before returns settle will systematically overstate results. Wait 60 days or apply an expected return rate.
On contribution margin over loaded marketing cost, healthy stores run 1.3–2.0× for blended marketing ROI. Below 1.0× means marketing is consuming more contribution margin than it produces. See ROI Benchmarks by Industry for figures by vertical and order-value tier.
In blended ROI, yes — organic and email revenue contribute to total contribution margin, and the salaries/tools that produce them contribute to loaded marketing cost. Excluding them inflates the ratio because you keep the revenue but drop the cost.
Weekly at the channel level for budget pacing, monthly at the blended level for trend, and quarterly with a full reconciliation against your P&L. The reconciliation matters: dashboards drift from accounting reality, and a quarterly check keeps both honest.
Meta reports revenue with its own attribution window and claims credit for orders it influenced rather than caused. Compute blended ROI — total contribution margin divided by total marketing spend — and the gap between in-platform and blended numbers tells you how much attribution inflation you're absorbing.
Both, for different decisions. First-order ROI tells you whether each acquired customer is immediately profitable — important for cash flow and scaling caution. LTV-based ROI (typically over 12 months) tells you the true value of acquisition and is the right view for annual budget allocation.
Numerator is the incremental contribution margin from the lift (treatment minus control, multiplied by post-test traffic and AOV and margin). Denominator is the cost of designing, building, and running the test — including tool cost. For most tests on a €1M+ store, even a 2% lift on checkout pays back within weeks.
Track CAC, channels, and funnel conversion in one place
Metricuno connects ad spend, funnel events, and revenue so you can see CAC by channel, cohort, and campaign — without stitching together five tools.