When I started out in affiliate marketing, I used to run a lot of mVas offers and always enjoyed the process of squeezing out profits on tiny margins. You also get a lot of data to work with which I enjoy!

Campaign Overview

⚠️  Actual profit is estimated ~$5K higher than postback data reflects, likely due to missing postbacks and recurring push subscription revenue not tracked in real time.

Starting Small and Validating GEOs

The testing phase was deliberately low-risk. New campaigns launched at $5/day using the popunder format: fast to load, quick to generate data, and cheap enough to run across multiple GEOs simultaneously without burning through budget.

The core metric at this stage was simple: what CPM is Propush.me returning for traffic from this GEO? That number tells you whether there’s enough margin to buy traffic profitably. If the earned CPM covers your cost with room to spare, you have a viable GEO. If it doesn’t, you move on.

How does ProPush.me constructor work?

Scaling What Worked

Once a GEO proved profitable, the playbook was straightforward: layer in additional formats. Popunder opened the door; in-page push, banner, video, and interstitial scaled the volume. Each format adds a new traffic pool while keeping you in the same winning GEO.

The portfolio of active campaigns eventually grew to 200–300 running simultaneously. Some of these were micro-earners: generating just a few dollars a day, while others were reliable, consistent performers. The micro-earners had their own value: when CPMs spiked in a geo, they’d pop into meaningful profitability without any manual intervention.

Creative Strategy: Balancing CTR and Quality

Creatives were split across two objectives rather than optimized for a single goal. Some were designed to maximize CTR, driving raw traffic volume. Others were built to improve traffic quality.

Account manager played a useful role here too — sharing insight into what types of campaigns are running on a given traffic source, informing landing page choices and helping align traffic with what advertisers on the Propush side were actually buying.

AI-Powered Bid Management

This is where the setup diverges from a traditional media buying workflow. With hundreds of campaigns active, manual bid management simply isn’t viable for a solo operator.

The solution was an AI-powered optimizer, built using Claude, that adjusts bids in real time based on performance signals. Weekly geo digests from Propush account managers are fed directly into the bidding strategy, flagging where demand was rising before CPMs peaked.

Without AI I might not have been able to be successful. To launch new campaigns in bulk and optimize them without a team, you need a tireless robot working for you, or an insane work ethic.

---

What Didn’t Work

❌  iOS traffic failed to deliver iOS was tested as part of the arbitrage setup but didn’t generate viable returns. The entire campaign ran on Android mobile exclusively. Don’t assume iOS will mirror Android performance in this model – test it, but don’t count on it.

Another successful case study with Constructor

---

Results & Key Takeaways

• Spend: $50,290.42
• Revenue: $75,231.61
• Profit: ~$24,941.19 (likely higher when accounting for untracked recurring revenue)
• ROI: ~50%

The campaign ramped gradually: March served mainly as a testing and learning period, with the bulk of revenue arriving in April once winning GEOs and formats were locked in and scaled.

⚠️ As mentioned above, part of the revenue was not reflected in the tracker/postback data, so while the tracker showed ~40% ROI, the actual final profitability was higher.

---

GEO Breakdown

⚠️  South Africa delivered the highest CPM at $8.65 — nearly double Brazil and Nigeria — despite lower traffic volume. Brazil and Nigeria drove the most impressions and anchor total revenue. Mexico punched above its weight on conversions relative to impressions, suggesting strong traffic quality. Egypt and Peru work best as always-on, AI-managed micro-earners.

---

Lessons Learned

1. The Constructor model rewards patience in testing. The first phase is about data collection, not profit. Launching cheap popunder campaigns across multiple GEOs gives you the CPM benchmarks you need to make smart bids. Don’t skip this step.

2. Traffic quality and volume are not mutually exclusive. Running creatives that balance CTR with quality directly affects your earned CPM. Higher-quality traffic earns more per impression, and changes your entire bidding math in your favour.

3. No offer caps are a genuine structural edge. Unlike traditional affiliate offers that can pause or disappear overnight, Constructor keeps running as long as you’re sending traffic. For affiliates who’ve been burned by offer volatility, this matters more than it might seem.

4. AI is the force multiplier for solo operators. Managing 200–300 campaigns without a team is only possible if something else handles the operational load. AI bid management was the single most important unlock in this setup.

5. Use your account manager – seriously. Weekly GEO digests and traffic source recommendations directly informed bidding decisions and creative strategy. This isn’t courtesy communication – it’s actionable intelligence. Treat it that way.

---

Closing Thought

This case study isn’t about a complex funnel or an exotic traffic angle. It’s about a disciplined, systematic approach to arbitrage — lean testing, quality-aware creatives, AI-managed optimization, and steady scaling into proven GEOs. The margins aren’t enormous, but they’re consistent, and the model scales without the usual affiliate headaches. For anyone comfortable with data and willing to invest in the right tools, the Constructor model is worth a serious look.

If you want to thank our partner for sharing his case study, join ProPush.me at his affiliate link

Interested in trying Propush.me Constructor? Reach out to your account manager for top GEOs and traffic source recommendations before you launch.