When Optimizing Campaigns on Free Trial Sign-Ups Isn't Resulting in Revenue

The Misalignment Problem

If you run a subscription-based business and invest in paid ads, there's a good chance you're optimizing for the wrong signals. Most platforms make it easy to optimize for top-funnel actions like free trial starts, sign-ups, or installs. But if those actions don't correlate with actual revenue or retention, you're training the algorithm to bring you more low-value users. Over time, a gap forms between what the ad platform thinks "success" looks like and what actually drives your business forward. When this occurs, we need to course-correct the algorithm through signal engineering.

What Is Signal Engineering?

Signal engineering is the intentional design and feedback of conversion events that reflect real business value. Instead of sending every micro-conversion back to Meta or Google, you define and send signals that correlate with high-value outcomes like retention, Average Revenue Per User (ARPU), or Lifetime Value (LTV). The goal: to train the algorithm not just to get you the cheapest trial users — but to find users more likely to provide long-term revenue.

Signs It Might Be Time to Explore Signal Engineering

• You're getting a high volume of free trials, but low subscriber conversion
• Your CAC is rising even though targeting and creatives haven't changed
• LTV hasn't budged despite increased ad spend
• You're relying on "Trial Started" or "Install" as your main optimization event
• You're hitting scale ceilings or saturation quickly
• You can't explain why some users convert and others don't
• You have strong engagement metrics but poor ROAS
• Your team is building great creatives but performance remains flat
• You have meaningful in-app events but aren't using them in ad optimization
• You've never updated your conversion events since launch

What Makes a Good Signal?

Good signals help your campaigns scale smarter, not just cheaper. Strong signals typically include:

• High correlation to long-term business value — predictive of retention, ARPU, or LTV
• Early enough to fit within conversion windows — preferably Day 0 to Day 2
• Enough volume to optimize — aim for 10+ signals per day
• Low noise — minimizes false positives and negatives
• Clean and transferable — can be passed via SDK, MMP, or CAPI
• Adaptable — can be edited or tuned by thresholds (e.g. quiz score > 80%)
• Monitorable — you can measure how often it fires and its downstream impact

What Signal Engineering Looks Like in Practice

1. Behavioral Indicators

• Completed Onboarding Flows: signals the user has taken time to explore core product features
• Activated Key Features: engagement with your "aha" moment (uploading a file, building a project, creating a playlist)
• Session Depth or Frequency: returning more than once within 24–48 hours can be a strong signal of intent

2. Declarative Indicators

• Quiz Funnel Responses: if you ask users about their use case or goals during sign-up, certain answers may correlate strongly with subscription likelihood
• Self-identification (e.g. SMB owner): if your product serves specific personas, ask qualifying questions directly

3. Predictive Scoring

• Regression Models: use historical data to run regression analysis on which combinations of early behaviors drive LTV
• Lead Scores: assign a score based on a weighted formula (quiz score + time on platform + feature use)

4. Post-Trial Conversion Signals

• Upgraded Without Discount: users who convert to paid without a promo often signal strong intent
• Early Plan Expansion: users who add seats, features, or integrations in the first 7 days can be excellent targets

Expected Outcomes of Signal Engineering

When brands shift from generic signals like "Trial Start" to qualified, behavior-based events like "Onboarding Completed" or "Qualified Trial," the results can be substantial. One test showed a 48% drop in cost per trial after switching the optimization event from "Trial Started" to "Onboarding Completed." With further iteration (no creative changes), cost per trial dropped as much as 64%. In a separate test, optimizing for a Qualified Trial event instead of a standard install drove a 31% improvement in ROAS.

How to Measure Success

Once you begin feeding new signals into your campaigns, your traditional KPIs may not tell the whole story. You might see costs per lead or trial go up — and that's not necessarily a bad thing. Stronger signals usually surface more qualified users who often cost more to reach. The real question is whether those users are delivering more value. Shift your focus to cohort behavior:

• Are your LTV or ARPU numbers rising?
• Are users sticking around beyond the first billing cycle?
• Are you seeing stronger retention from signal-qualified users?

Don't rely solely on modeled conversions. Measure how each signal-qualified cohort actually behaves over time.

Do's & Don'ts of Signal Engineering

DO:

• Look at your own data before choosing a signal
• Focus on signals that happen early in the user journey
• Map your best-quality events to standard conversion types (like Purchase or Lead)
• Keep an eye on signal health and volume
• Run multiple campaign types in parallel

DON'T:

• Assume more data is better
• Let one optimization event define everything
• Delay your signal too far beyond the trial
• Try to be overly clever too early if you're just getting started
• Use signal engineering as a patch for bad creative or an unproven offer

Frequently Asked Questions

What is signal engineering in paid advertising?

Signal engineering is the practice of intentionally designing and selecting conversion events that reflect real business value — and feeding those events to ad platforms for optimization. Instead of optimizing on shallow top-funnel events like trial starts or installs, signal engineering identifies behaviors that predict retention and revenue, and uses those as the optimization target for campaigns.

Why is "Free Trial Started" a bad optimization event?

Free trial starts are easy to measure and plentiful — which is why platforms love to optimize for them. But if trial users don't correlate with paying subscribers, you're training the algorithm to find cheap trial users, not valuable customers. Over time, this drives up CAC without improving LTV. The fix is replacing the optimization event with a signal that better predicts conversion to paid — like onboarding completion or a qualifying behavioral score.

How do I know which signal to use for campaign optimization?

Start by analyzing your existing customer data. Look for behaviors that happened in the first 0–48 hours that are disproportionately common among users who converted to paid and remained subscribers. These behavioral patterns — onboarding milestones, feature activations, session frequency — become your candidate signals. The best signal is early, high-volume, low-noise, and strongly correlated with long-term business value.