Est. reading time: 7 minutes

Optimizing ROAS for subscription services starts with the right target

Most subscription brands we audit are running Facebook ads against a ROAS target borrowed from one-time-purchase ecommerce. Someone decided 3x was the line, the account gets managed to that line, and every decision downstream inherits the error. Campaigns that recruit high-retention subscribers get paused for “underperforming.” Campaigns that recruit coupon-driven churners get scaled because their day-one number looks good.

ROAS divides immediate revenue by spend. A subscription business does not collect its revenue immediately. It collects a small first payment, then earns the rest over months of renewals. So the metric you are watching in Ads Manager reports the smallest and least representative slice of the transaction, and optimizing hard against it means optimizing against a distortion.

The fix is to define your target from your own economics before touching the account. Take gross margin per subscriber month, average retained months by cohort, and the payback window you can tolerate on cash flow. From those three numbers you can calculate the first-order ROAS at which a new subscriber breaks even. For most subscription brands that number is well below what a one-time-purchase brand would accept, sometimes below 1.0. A campaign showing 1.3x on day one can be your most profitable acquisition channel. A campaign showing 3x can be losing money if its subscribers cancel at the first renewal. Until you know your breakeven, you cannot tell which is which.

Feed Meta a conversion signal that matches subscription revenue

Meta’s delivery system optimizes toward whatever event you hand it. Hand it trial starts, and it will find people who start trials, including the segment that starts trials habitually and never converts to paid. Hand it a generic purchase event, and it treats a $1 trial and a full-price annual plan as the same win. The algorithm is not confused. It is doing exactly what it was told, and it was told the wrong thing.

There is a volume constraint to work within. Meta’s documentation puts the learning phase exit at roughly 50 conversion events per ad set per week. If your deepest meaningful event (paid subscription start, or second successful billing) fires 15 times a week, ad sets optimizing for it will sit in learning indefinitely and delivery stays unstable. So the practical rule is to optimize for the deepest event in your funnel that still clears that volume threshold, then move deeper as spend and conversions grow.

Two adjustments do most of the work here. First, send subscription events back through the Conversions API with real values attached, so a $79 plan and a $19 plan are not identical signals. Second, once purchase volume supports it, test value optimization rather than conversion volume optimization, since it points delivery at the subscribers worth more rather than the subscribers easiest to convert.

Creative that sells month six, not the first click

Discount-led creative is the default for subscription acquisition, and it quietly selects for the wrong customer. “50% off your first box” pulls in people whose motivation is the 50%, and that motivation expires precisely when full-price billing begins. The ad performed. The cohort didn’t.

The alternative is creative that sells the recurring value, because the person who responds to it has already accepted the recurring part:

• Selects for churn: “Get your first month for $5”
• Selects for retention: “The coffee you’ll never have to remember to reorder”

The first version can still have a place at the bottom of the funnel, as a closer for someone already convinced. It fails as the lead message, because it makes the discount the product.

Testimonials and user-generated content are especially effective for subscriptions because the buyer’s real question is not “is this good” but “will this still be good in month four.” A customer on camera saying “I’ve had this for a year and canceled everything else” answers that question in a way brand copy cannot. When you A/B test creative, judge the variants on cost per retained subscriber or per second billing where volume allows, not on CTR. Click-through rate rewards the loudest ad, which is frequently the one recruiting the worst cohort.

Seed your targeting with the customers who stayed

Lookalike audiences replicate whatever you seed them with. The common setup seeds from all purchasers, which means the model learns from your churners in proportion to how many of them you have. You end up paying Meta to find more people who cancel.

Better inputs are easy to build if your billing data is accessible. Seed lookalikes from subscribers who passed their third billing cycle, or from the top quartile by lifetime value. Use custom audiences the same way in reverse, retargeting site visitors and trial abandoners while excluding active subscribers so you stop paying to acquire people you already have. The mechanics of connecting billing data to ad platforms are the same ones covered in our piece on subscription strategies that raise customer lifetime value, and the seed-audience quality matters more than any interest-targeting refinement layered on top.

Match the funnel stage to the buyer’s actual objection

Subscriptions carry a heavier commitment than a one-time purchase, which means the funnel has to do more persuading before the conversion ask makes sense. Cold prospects served a “subscribe now” ad are being asked to commit to a recurring charge from a brand they met four seconds ago.

Structure the campaigns around the objection at each stage. At the top, the job is recognition of the problem your service solves on a recurring basis, the territory we cover in our tips on building brand awareness through social advertising. In the middle, the objection is trust in the ongoing experience, which is where testimonials, unboxings, and “how it works” content belong. At the bottom, retargeting with a specific offer closes people who have already resolved the first two objections. Format follows stage. Video carries the middle, carousels show tiers and plan options, and if your traffic skews mobile, the format considerations in our guide to Facebook ads for mobile users apply directly.

Bidding follows the math you did at the start

Once you know breakeven ROAS and target payback, bidding stops being a guessing game. Cost-per-result goals and bid caps should be derived from what a retained subscriber is worth, not from what feels efficient in the dashboard. If your allowable acquisition cost is $60 against first-cycle revenue of $30, a campaign converting at $45 is healthy even though its first-order ROAS reads as a failure. Automated bidding handles the auction mechanics well. What it cannot do is know your unit economics, and it will faithfully optimize toward whatever ceiling you set, right or wrong.

Retention is a ROAS input, not a separate department

Every month a subscriber stays, the ROAS on the ad that acquired them improves. This is the one advantage subscription brands hold over one-time-purchase brands, and most accounts we audit are not using it. Acquisition and retention are run by different people, measured on different dashboards, and the ad account never learns which campaigns produced the subscribers who stayed.

Close that loop in two places. Operationally, onboarding is the highest-return retention work available, since first-cycle cancellation is usually a failure of the first two weeks, and automated flows like the ones in our breakdown of ActiveCampaign workflows for growth can carry it without headcount. Analytically, report cohort retention by campaign, not just by month. When you can see that campaign A’s subscribers stay seven months and campaign B’s stay two, budget decisions make themselves.

None of this is exotic. Set the target from your own economics, hand Meta a signal that reflects real subscription value, build creative and audiences around the customers who stay, and feed retention data back into budget decisions. If you want a second set of eyes on where your account sits against that list, we’re straightforward to reach.