--- title: "Same budget, higher-LTV users: Signal engineering on Meta Ads" description: "How a fitness app used LTV-weighted trial values in Meta to cut under-25 spend from ~25% to under 10% and triple paid subscriptions from users 35+." url: "https://adapty.io/blog/same-budget-better-users-signal-engineering-on-meta/" language: "en" slug: "same-budget-better-users-signal-engineering-on-meta" category: "grow-your-app" date: "2026-10-08T09:50:18.122Z" date_modified: "2026-10-08T13:11:48.912Z" author: "Murat Menzilci" --- # Same budget, higher-LTV users: Signal engineering on Meta Ads ## TL;DR - **The problem:** campaigns were split by age, yet ~25% of Meta spend still went to users under 25, the group with the lowest LTV. - **The change:** with Adapty Attribution's Revenue Override, under-25 trials were sent to Meta at 20% of the reference value. Older cohorts got higher values based on expected LTV. Campaigns, bidding and spend stayed the same. - **The result:** under-25 share of spend fell below 10%, and paid subscriptions from users 35+ grew 3x at the same spend. - **The caveat:** this was a before-and-after comparison, not a controlled test, and the 3x applies only to the 35+ group. - **The takeaway:** don't just tell Meta who converted. Tell it which conversions are worth more. Meta gives advertisers powerful audience targeting options. But demographic targeting alone doesn't always deliver the users who generate the most value. We recently tested this with a Health & Fitness subscription app. The team already knew that older users had stronger subscription economics. Their historical data showed higher LTV and better long-term monetization among users aged 35 and above. To reach those users, they had structured Meta campaigns around different age segments, including 25–35, 35–45, and older audiences. But campaign reporting revealed a problem. Despite the age-based segmentation, **roughly 25% of campaign spend was still going to users under 25**, a cohort with significantly lower LTV. The team already knew which users were more valuable. The challenge was getting Meta to prioritize them, even when targeting controls couldn't. --- ## The problem: Not every trial has the same value From Meta's perspective, a trial start is a positive conversion event. But a trial from a 22-year-old user and one from a 42-year-old user may have very different expected values. The customer's historical subscription data showed exactly that. Older users generated stronger LTV, while users under 25 were considerably less valuable over time. Yet Meta was receiving similar trial signals from both cohorts. **The algorithm knew who was starting trials, but the signals didn't reflect which trials were worth more to the business.** That became the basis of our experiment. ### Why age targeting wasn't enough The obvious fix was to tell Meta not to reach younger users. The team had already tried that. On top of the age-segmented campaigns, they set a **minimum age of 25** and added **Value Rules** to lower bids for younger audiences. Yet, based on the age users reported inside the app, under-25 users still accounted for roughly a quarter of spend. Value Rules adjust bids, but they don't exclude anyone. When younger users converted cheaply, Meta kept reaching them. **Targeting controls told Meta who to reach. They didn't tell Meta which conversions were worth more.** ## The hypothesis: Change the signal, not the audience Instead of introducing more demographic restrictions or additional campaigns, we tested a different approach. Since the app already knew each user's self-reported age, that first-party data could be sent back to Meta as a conversion value. What if Meta received different conversion values based on the expected LTV of each age cohort? Our hypothesis was that **value-weighted trial signals could steer Meta toward users with stronger subscription economics**, without changing the existing campaign structure. ## The setup: Revenue Override with Adapty Attribution Using **Adapty Attribution's Revenue Override**, we adjusted the value sent to Meta when a user started a trial. The campaign structure, optimization event, bidding configuration, and spend level remained unchanged throughout the comparison. The only deliberate change was the value attached to the trial event. ![how the campaign works](https://adapty.io/uploads/chart-0-how-it-works-1.png) Users under 25 still generated trial events, but those events carried a substantially lower value than trials from older cohorts. These values were not actual revenue recognized at trial start. They were **modeled optimization values**, informed by historical subscription economics. Instead of treating every trial as equally valuable, we gave Meta information about the relative value of those conversions. --- ## The results: Less low-LTV spend, 3x more paid subscriptions from users 35+ The change in campaign delivery was significant. ![Share of Meta spend](https://adapty.io/uploads/chart-1-under25-spend-2.png) After introducing value-weighted trial signals, the share of spend going to users under 25 fell by more than 15 percentage points. More importantly, paid subscription conversions from users aged 35+ increased 3x, while overall campaign spend remained unchanged. ![trial value sent to Meta by Age](https://adapty.io/uploads/chart-3-35plus-paid.png) This was a before-and-after comparison, not a randomized controlled experiment. The 3x result refers specifically to the 35+ cohort, not total paid subscriptions or overall ROAS. We cannot attribute the entire improvement to signal weighting alone, but the delivery shift was consistent with our hypothesis. --- ## Why this matters for subscription apps The interesting part of this experiment wasn't a new targeting strategy. It was using existing subscription data to improve the feedback Meta received. In subscription businesses, the actual value of a user often becomes clear weeks or months after acquisition. A user might start a trial today, convert in seven days, and renew multiple times before their full value becomes apparent. That makes early value signals especially important. When historical data shows meaningful differences between cohorts, **revenue weighting can give Meta a more useful proxy for expected value than an identical signal for every trial**. And that principle extends beyond age. Similar approaches can be explored using plan type, geography, early engagement, or other attributes with demonstrated relationships to LTV. ## Beyond attribution: Better signals, better optimization Attribution tells you where users came from. Signal engineering helps you decide what the ad platform should learn from those users. In this experiment, Adapty Attribution helped translate historical subscription economics into differentiated trial values sent back to Meta. The result was a meaningful shift in delivery toward higher-LTV cohorts, alongside a 3x increase in paid subscription conversions among users aged 35 and above. **The takeaway is simple: Don't just tell Meta who converted. Help it understand which conversions are worth more.** [Learn how Adapty Attribution helps subscription apps optimize Meta campaigns](https://adapty.io/attribution/) ## FAQ ### What is signal engineering in mobile app marketing? Signal engineering means shaping the conversion data you send to an ad platform so it learns what actually matters to your business. Instead of sending the same signal for every trial, you attach values that reflect expected LTV, so the algorithm steers delivery toward higher-value users. ### What is Revenue Override in Adapty Attribution? Revenue Override lets you change the value Adapty sends to an ad network like Meta when a conversion event fires. In this case, trial starts from different age groups were sent with different values based on historical LTV, instead of one flat value for every trial. ### Why didn't age-based targeting alone work on Meta? The campaigns were already split by age, yet about 25% of spend still went to users under 25. Meta saw every trial as an equally good result, so it had no reason to favor older, higher-LTV users within those audiences. ### Are the trial values sent to Meta real revenue? No. They are modeled optimization values based on historical subscription data, not revenue earned at trial start. Their job is to tell Meta how much one trial is worth compared with another.