Build it in two halves: PPC ROI is incremental profit divided by ad spend, listing ROI is the conversion lift multiplied by existing sessions and unit margin. Model both on profit, never revenue, and hold the calculation to a fixed window. Most calculators fail because they credit ads with sales that would have happened anyway.
The short version
- Two calculations, not one. Ads and listing work use different inputs and settle over different windows.
- Profit is the numerator. A return measured in revenue flatters every campaign and decides nothing.
- Subtract a baseline. Units you would have sold regardless are not returns on the campaign.
- Listing gains compound, ad gains stop. One keeps paying after the invoice ends, the other does not.
- Put the management fee inside the model. If it does not fit in the profit line, the answer is no.
Start from the error that breaks most calculators
Almost every ROI model sent to me makes the same mistake. It compares sales in the managed period against sales before it, then credits the whole difference to the campaign. That difference contains price changes, seasonality, a competitor stockout, review velocity, and whatever listing work happened in the same weeks.
The cost is real. A seller I spoke with had kept a campaign structure running for eleven months because a spreadsheet said it returned four to one. The same units had been selling organically before the campaign existed. The ads were harvesting demand, not creating it, and the spend was pure margin loss dressed up as growth.
So the first rule of a usable calculator is that it measures incremental profit, and the second is that it states the window before you look at the result.
The PPC half
| Input | Where it comes from | The usual error |
|---|---|---|
| Attributed units | Campaign report, fixed window | Taking the 14-day attribution figure and calling it causal |
| Baseline units | The same ASIN before spend started, or a paused period | Skipping it entirely |
| Contribution margin per unit | Sell price less Amazon fees, landed cost, and returns | Using gross margin, which ignores fees |
| Ad spend | The same window, same campaigns | Mixing brand defense spend with acquisition spend |
| Return rate | 60 to 90 days after the window | Measuring profit before returns land |
The arithmetic: incremental units equal attributed units minus baseline units. Multiply by contribution margin per unit to get incremental profit. Divide by ad spend. A result of 1.0 means the advertising paid for itself and nothing more.
The listing half
Listing work does not have its own spend line, so the calculation runs the other way. Take the conversion rate before the change and after it, over comparable traffic. Multiply the difference by sessions in the period, then by contribution margin per unit. That is the monthly profit the work created. Divide the one-time cost of the work by that figure and you have payback in months.
A listing improvement that moves conversion by one point on a page doing 8,000 sessions a month at $9 of contribution is about $720 a month in new profit, and it keeps producing after you stop paying for it. That is the asymmetry buyers miss when they compare an ad management quote against a creative quote.
Failure modes, ranked by what they cost you
- Revenue in the numerator. The single most expensive error. A 3x return on ad spend at a 20 percent contribution margin is a loss, and the calculator will still show green.
- No baseline. Costs almost as much, because it makes bad campaigns immortal. If you cannot pause to measure, use the pre-launch period or a matched ASIN.
- Mismatched windows. Ad spend lands today, returns land in 60 days, and listing gains build over a quarter. Comparing a 30-day cost against a 30-day gain understates listing work and overstates ads.
- Return rate left out. A product with a 12 percent return rate loses more than the refund. It loses the fees, the inbound shipping, and the rating trend that drives future conversion.
- The fee outside the model. The cheapest error to fix and the one people skip. Management cost is part of cost of service. Put it in.
Where the management fee belongs in the model
Add the monthly fee to the denominator of the whole account, not to any single campaign. If the fee is $800 for one product, the model has to produce $800 of incremental profit before anything else counts. At a 25 percent contribution margin that is $3,200 of incremental sales a month, which is a target you can check against in 30 days rather than argue about.
That threshold is also how you should read staffing. At Flapen 50 operators carry about 70 brands, close to 1.4 brands each, and that ratio is the reason the fee sits where it does. Ask any agency for their equivalent number and divide their monthly fee by it. If one manager carries fifteen accounts, the fee is buying you a fraction of a person, and your ROI model should assume the attention you are actually paying for.
What most agencies will not tell you
A pretty ROI calculator is a sales instrument. When the inputs are chosen by the party being measured, the output is a foregone conclusion. Ask who picks the baseline, who sets the window, and whether the return rate is included, and you will learn more than the number ever tells you.
The second thing: some accounts should not be advertised harder at all. When conversion is the constraint, more traffic multiplies the leak. A calculator that only has an ad spend input cannot represent that, so it will always recommend more spend. Build the listing half or you have built a machine that agrees with whoever is selling you clicks.
Related answers
- Amazon agency ROI benchmarks 2026
- Calculate payback period for Amazon PPC management
- KPIs an Amazon agency should report weekly
- What does a good Amazon account audit include
- Amazon agency pricing and economics: the complete guide
If you want the model run against your own numbers before you commit to anything, that is what the free audit does at Flapen.
