The Formula
ROAS = conversion value ÷ ad spend
$4,000 ÷ $1,000 = 4.0 → 400% ROAS
Google's own arithmetic, verbatim from their Target ROAS documentation: $5 USD in sales ÷ $1 USD in ad spend x 100% = 500% target ROAS.
In Google Ads the historical figure is the Conv. value/cost column multiplied by 100. In Meta it appears as purchase ROAS. In TikTok it is reported only once you send a value, because the platform's documentation lists the value parameter as required for ROAS and for value-based optimisation.
ROAS Is Not ROI
They are different measurements and the difference is the cost of goods.
ROAS compares revenue to ad spend only. ROI compares profit to total cost. A campaign at 400% ROAS on a product with a 70% cost of goods is losing money, and the ROAS column will keep reporting 400% the entire time it does so.
This matters differently depending on who you are. A brand with a 30% gross margin needs roughly 3.3x just to break even before overhead. An affiliate is in an unusual position here, because an affiliate has almost no cost of goods: the commission arrives net of everything the merchant spent making and delivering the product. For an affiliate, ROAS computed on commission is very close to ROI. Computed on the customer's order total, it is not close to anything.
Where the Numerator Comes From
Here is the part that gets skipped. The conversion value in that ratio is not measured by the ad platform. It is reported to the ad platform, by you.
Google's documentation says it directly: Google Ads "predicts future conversions and associated values using your reported conversion values, which you report through conversion tracking." Meta's parameter reference defines value as "required for purchase events or any events that utilize value optimization" and specifies that it "must represent a monetary amount."
Read that definition again. It says the amount must be monetary. It does not say whose.
The spend side of the ratio is a billed fact; the value side is an assembly of decisions about what to send, which conversions to include and how far back to look.
Four separate choices sit inside that numerator, and none of them is announced when the number appears in a column.
1. Which number you send. For most advertisers there is one obvious candidate. For an affiliate there are two, and both are true. ClickBank's instant notification carries totalOrderAmount, defined as "total the customer was charged", alongside totalAccountAmount, defined as "total you received for the transaction in USD". If the $297 order paid you $89.10, your ROAS is more than three times higher when computed on the wrong one.
2. Which conversion actions are included. Google's Include in "Conversions" setting decides which conversion actions land in the reporting columns that bidding reads. Turn a lead action on and your conversion value goes up without a single extra sale.
3. What lookback window is in force. A longer window credits more conversions to the same spend. The window is a setting, and moving it moves your ROAS without touching the campaign.
4. Whether value rules are rewriting it. Google's conversion value rules apply a multiplication factor for higher-value customer types, devices or locations. The value in the column is then not the value you sent.
Target ROAS Is a Different Thing From Reported ROAS
Reported ROAS is a description of what happened. Target ROAS is a bidding strategy that treats your reported conversion values as training data and bids to hit an average you name.
The eligibility minimums are worth knowing before you set one, because they are published and specific:
| Campaign type | Minimum before Target ROAS is available |
|---|---|
| Search and Shopping | 15 conversions in the past 30 days |
| Display | 15 conversions with valid values in 30 days, across all campaigns |
| App | 10 conversions per day, or 300 in 30 days |
| Demand Gen | 50 in the past 35 days with 10 in the past 7, or 100 across the account |
| Video Action | 30 in the past 30 days |
| Hotel | 50 per week |
| Travel | 50 in the past 7 days |
Google also notes that conversions must have a value greater than zero to count as eligible for Demand Gen. A conversion firing with a blank value is invisible to the thing you are asking to bid on it.
Two dated changes worth carrying: from June 2026 Google relabelled "Maximize conversion value with a Target ROAS" to simply "Target ROAS" with no change to the underlying behaviour, and from 17 August 2026 Google began changing how budget-limited campaigns bid, with a Bid Target Adjustment Tool released on 6 July 2026. If your Target ROAS campaigns behaved oddly in late August, that is the first thing to check.
Reported ROAS and Incremental ROAS Are Not the Same Number
A ROAS figure answers the question of how much attributed revenue followed your spend. It does not answer how much revenue your spend caused.
The best-documented gap between the two remains the eBay field experiment published in Econometrica in 2015. On non-brand paid search, the observational estimate with time and geography controls came out at 1,632% ROI. The randomised experiment on the same spend came out at negative 63%.
That is not a tracking failure and better attribution does not close it. Both estimates used complete data. The distance is selection: the people who clicked were already more likely to buy. A ROAS column is a credit measurement, and credit and cause are different questions. See incrementality for how the causal version is measured and what it costs to run.
The Affiliate's Version of the Problem
Everything above applies to any advertiser. One part applies to affiliates in a way it does not apply to brands.
On a 30% commission the two defensible ROAS figures for one campaign are 356% and 107%, and only one of them predicts whether you can pay for next month's traffic.
The arithmetic is not subtle. Your gross-value ROAS is your true ROAS divided by your commission rate. At 30%, everything reads 3.33x better than reality. At 10%, ten times better. And because value-based bidding maximises the number you supply, sending the order total does not merely misreport the campaign, it steers the campaign toward buyers with large carts rather than buyers who pay you well. Those are the same people only when your commission rate is flat across every product, tier and upsell, which for most affiliates it is not.
Meta already has a parameter for the number an affiliate actually earns. net_revenue is defined in their reference as "the margin value of a conversion event." Almost nobody fills it.
Common Mistakes
- Comparing platform ROAS to platform ROAS across channels. Meta, Google and TikTok each attribute the same sale under their own model and window. Adding their reported revenue together produces a total larger than your bank deposits.
- Treating a ROAS improvement after a window change as performance. Widening a lookback window raises reported ROAS with no change in the campaign. Check whether the setting moved before you celebrate.
- Reporting ROAS on gross order value while paying for traffic out of commission. Covered above. It is the single most common way an affiliate ends up scaling something unprofitable.
- Setting a Target ROAS from a period that includes the conversion delay. Google recommends excluding the most recent conversion delay period, because the latest days are always understated and will make you set the target too low.
- Sending a purchase event with no value at all. It will not train value-based bidding, and for some campaign types it does not even count toward the eligibility minimum.
Related Terms
- Incrementality is the causal version of the question ROAS answers descriptively.
- Attribution window sets how far back a conversion may be credited, and moves reported ROAS directly.
- Offline conversions are the revenue events that arrive after the browser is gone, which for affiliates is most of them.
- Conversion API (CAPI) is the pipe the value parameter travels down.
