Andromeda, a More Expensive Auction, and Storms: A Survival Recipe from ParadoX Team

We are ParadoX Group — an iGaming ecosystem built around three units. ParadoX Team is our in-house media buying operation across FB, native, and in-app. ParadoX Partners is an affiliate network with 1,500+ offers and 400+ direct advertisers. ParadoX Media is our editorial arm, where we publish industry insights. We buy traffic ourselves, compete in auctions ourselves, test offers ourselves — which is exactly why we understand a partner's challenges better than anyone. 

At the start of 2026, the media buying market was shaken up once again. But panic is a poor strategy. We gathered data, adapted our infrastructure, and are now ready to share what helped us not just hold our volumes — but grow them — in conditions where the old approaches simply stopped working. 

What Changed in Meta in 2026

It all started with the Andromeda update. After the rollout, the algorithm seemed to lose its ability to find the target audience. Before, you'd launch a campaign and wait 2–3 days for it to ramp up — now it lives for a day or two, delivers a sharp spike, and then breaks down. Maintaining stable results has become significantly harder.

The targeting logic also changed. We always liked control — precise settings, specific age ranges, placements. In 2026, FB decided to make all of that optional. We had to let go and give the algorithm the right to decide who sees our creatives. For the team, it was a psychologically uncomfortable step — but without it, there were no results. In practice, targeting is now driven by the creative itself: its message, hooks, and triggers.

The auction became unpredictable. Before, we had a rough sense of what CPM to expect. Now CPM can shift dramatically within a single day. Competition for high-quality, paying traffic has intensified, and FB has gotten more expensive overall. The one thing that held — a strong working bundle kept delivering even through the chaos.

A separate note on moderation. FB has moved to neural network-based multimodal analysis — the platform now evaluates the creative, the app, and the landing page as a single unit in real time. One account, one launch — and you're immediately catching rejects and bans. Running without proper infrastructure has simply become impossible.

How We Maintain Volumes

Speaking specifically about our in-house buying operation, ParadoX Team, three pillars are what keep our volumes up today:

• Our stack: FB, native, and in-app. If one source is turbulent, we balance across the others.
• We work across a wide range of GEOs — from Tier-1 to the massive-volume Tier-3 markets. We know how players behave in different parts of the world.
• We take a systematic approach to infrastructure. Our own technical solutions, custom account setups, and work with trusted ad accounts — that's what allows us to withstand any sanctions from traffic sources.

The first thing we revisited was our approach to creatives. Before, we'd launch one creative per campaign. Now we work on a 1-1-5 model: one offer, one campaign, five creatives. 

And not five similar variations with different headlines — genuinely unique materials, localized for each GEO, accounting for local trends and news-based triggers. Attention spans have shortened: where we used to have 3–4 seconds to warm up the viewer, now a creative has to hook in the first 1–2 seconds.

Keeping all spend on FB is a risk we can no longer afford. We moved into in-app traffic — and discovered a whole new world. The entry threshold is lower, the average CPM is more attractive, and across a number of GEOs, ROI and LTV are already outperforming FB. What started as diversification has become a full-fledged part of the strategy.

In parallel, we integrated PWA push notifications via Extendy into our apps. We noticed something interesting: players who receive push notifications through Extendy re-deposit at significantly higher rates. The tool fit organically into our in-app traffic strategy and became part of the retention funnel.

After connecting push notifications and the pixel, the deposit conversion rate grew from 15% to 26%: a properly calibrated pixel started pulling in audiences that actually pay. And the average check rose from ~€110 to ~€340 — a single first deposit doesn't account for that, meaning players were coming back and topping up.

Before

After

A separate point on infrastructure. We saw firsthand how an absolutely identical campaign performs radically differently across different ad accounts. Running everything from a single account no longer works — you need a pool of quality accounts for launching and testing.

Interestingly, the same applies to pixels. The same offer and creative on different pixels produced drastically different ROI results. So we introduced a simple rule: if a pixel underperforms at the start — we move on to the next one. As soon as we find one that actually delivers — we shift the main traffic focus to it.

What You Should Definitely Avoid

Entering FB without real infrastructure and experience. Before, you could come in with a minimal budget, spy on other people's approaches, and somehow make it work. Now FB filters out anyone who isn't ready to operate systematically.

Aggressively scaling budgets without data — we've been burned by this. Our rule: start with small budgets and scale no more than 20–30% per day. On Tier-1 you can afford to be a bit bolder; on Tier-3 — only carefully and gradually.

Panicking after every update and rebuilding everything on the fly. Experience has shown: it's better to wait 3–5 days for the algorithm to stabilize, collect initial data — and only then test new approaches.

Keeping everything in one account and relying on a single source. Not being dependent on the mood of one algorithm isn't paranoia — it's strategy. In some cases, in-app and new GEOs delivered better results than FB. In others, they simply hedged the risk. Either way — it's the stronger long-term approach.

Conclusion

FB in 2026 is not a platform where you can come in with a minimal budget and run experiments. It's getting more expensive, moderation is tightening, and the algorithm demands trust rather than control. Those who accepted this and adapted their infrastructure are continuing to scale. Those waiting for things to "stabilize on their own" are losing time and budgets.