1. Add locales — the easiest CPA lever
If your product can serve more than one market, localization is usually the cheapest win available. Ad auctions are priced per market, and CPMs differ several-fold between countries — the exact same creative, localized, competes in much cheaper auctions. That lowers your blended CPA without a single new concept. The cats already do the heavy lifting: Crea produces native-reading variants with/localize, and Bidder routes each language to the right audience and landing page. Your only job is telling Roy which markets your product actually supports.
❌ keeping everything in English because that’s what setup started with ✅ “we can serve Spanish and German customers too — add those locales and localize our current winners”
Only add locales your product genuinely supports — checkout, onboarding, support. Cheap clicks into an English-only funnel just move the problem down the funnel.
2. Check your targets against reality
You can’t simply tell the team to hit a very low CPA or a high ROAS and expect the auction to comply. Every industry and niche has its own benchmarks — a $5 CPA might be trivial for a casual app and impossible for B2B software. An unrealistic target quietly poisons the whole loop: every review reads as failure, and ads that are actually performing fine for your market get killed. Also judge the overall economics, not just the initial loss. Subscription and LTV-driven businesses routinely run below break-even on first purchase and make it back over the customer’s lifetime — that’s why setup asks for an initial Meta target, not your forever number.❌ “our ROAS target is 4x” (in a niche where 1.5x is a strong result) ✅ “what’s a realistic initial target for our niche and business model? Let’s set that, and revisit once we have our own data”If you’re not sure your targets are sane, ask Roy directly — it knows your business model from setup and can sanity-check the math with you.
3. Widen the search — pain over polish
If locales and targets check out, the problem is usually a search that’s too narrow. Three ways to widen it:- Go painkiller, not vitamin. Ads that sell relief from a concrete, felt pain beat “nice to have” framing almost every time. Ask the cats to go harder on the pain route — name the problem bluntly, then relieve it.
- Test more broadly. More pains, more audiences, more formats. If every ad in the account is a UGC video about the same benefit, you haven’t tested three ads — you’ve tested one idea three times.
- Help each other. The kitties can get stuck in a local maximum, iterating on the same losing direction. You know things about your customers the data doesn’t show yet — bring your own hypotheses, and let the cats bring theirs. It’s a collaboration, not a vending machine.
❌ “the video didn’t work, make a better video” ✅ “let’s go more pain-first: propose 5 concepts, each attacking a different pain, mixed formats — and I’ll add two angles from what customers tell our support team”
Still stuck?
Run/review with Roy for a deep portfolio analysis — it ranks what you’ve tested, finds the patterns, and proposes the next round of concepts against your real numbers.
Working with the Kitties
The collaboration habits that get dramatically better output.
Performance tiers
How the system ranks winners, risers, and fatigued ads.