Set ROAS targets from lifetime value and payback, not from last month
A method for deriving a ROAS target for a subscription app from predicted lifetime value, margin and the payback window you can finance, with a worked example.
Most ROAS targets are set by looking at last month's ROAS and picking something slightly better. That makes the target a description of the past. For a subscription app it is also the wrong time horizon, because the revenue that justifies an install arrives over months and the platform's optimizer only sees the first few days. A better target starts with the economics of a paying user and works backward.
Start with the cash you can wait for
Two numbers set the frame. The first is gross margin on subscription revenue, after store fees and any other cost of delivering the product. The second is how long you can wait to get the spend back. A team with thin cash may need a three-month payback. A team with funding or a strong balance sheet may accept twelve.
Here is a hypothetical with round numbers. Suppose a subscription app earns $60 in revenue over the first twelve months from each paying user, keeps 70% of it as margin after fees, and sees 5% of installs become payers. Revenue per install over twelve months is then $3, and margin per install is about $2.10. If the goal is to break even at twelve months, the most the app can pay per install is about $2.10. Expressed as ROAS on twelve-month revenue, that is $3 of revenue for each $2.10 spent, or about 143%.
If instead you need payback inside three months, and the same user earns $20 in the first three, the same math gives $1 of revenue per install, $0.70 of margin, and a break-even cost per install of $0.70. The ROAS target in revenue terms is again about 143%, but the cost per install you can afford is a third as high. Short payback windows are mostly a spending cap in disguise.
Translate the target into what the platforms can optimize
This is the part that goes wrong. The target above is based on twelve-month value. The ad platform can see only what arrives in its window. So you need a bridge: a predicted value that you send as the conversion value, or a proxy event that correlates with long-term payers.
- Google Ads. Target ROAS uses the conversion values you report. Google's help page says ROAS is conversion value divided by ad spend, so $5 in sales per $1 spent is a 500% target. It also says to base the target on business goals and historical ROAS, to exclude the most recent conversion delay period when evaluating, and, before switching, to bid to a conversion goal and report values for four weeks or one to two conversion cycles, whichever is longer. Smart Bidding also needs minimum conversion volume, which differs by campaign type, so check the requirement for yours.
- Meta. Meta's Marketing API documentation describes value as an optimization goal, shown in Ads Manager as a highest value bid strategy, and a minimum ROAS floor set through the API. Both depend on you sending purchase values.
- Apple. Apple's attribution framework returns conversion values from 0 to 63 and, at low volumes, only a coarse low, medium or high value. If you buy on Apple surfaces and measure through that framework, your value signal is coarse by design, so keep the mapping simple and tied to a few events that predict payment.
Calibrate the early signal against real cohorts
The proxy only works if it is checked. Take cohorts that are old enough to have paid out, and compare the early event, such as starting a trial or completing a key action, with the revenue those users actually produced. Then express early value as a multiple of eventual value. If users who start a trial go on to produce $12 on average within a year, send $12 as the trial-start value, not the full $60 and not the $0 that a purchase-only count would imply. Revisit the multiple every quarter, because pricing, mix and retention move.
Be honest about uncertainty. A target built on predicted value is a bet, so give it a margin. If break-even is 143%, aim for 175% to 200% so that a modest forecasting error does not put you under water.
Set different targets for different jobs
One number for the whole account blurs real differences. A brand search campaign should clear a much higher ROAS than a prospecting campaign, because it captures demand that already exists. If you hold both to the same target, the platform will pull spend toward the easy wins and starve the channel that creates new buyers. A reasonable approach is to set a blended target for the business and separate floors by campaign role, then check the blend monthly.
Practical takeaway
Write down your margin, payback window and payer rate. Compute break-even cost per install and the ROAS that implies. Then set the platform ROAS targets above that line by your chosen margin of safety, and check them against cohorts once they have matured.