What went wrong
A modest project budget was split across four narrow audiences. Two micro-audiences received about PKR 3,100–3,400 each and around six thousand impressions — never enough conversions to exit Meta’s learning phase. Their CTRs were the best in the project, so the audience and creative were right. The structure was wrong.
| Ad set | Leads | Cost per lead | Impressions | CTR |
|---|---|---|---|---|
| Family audience | 189 | PKR 1,640 | 1,148,916 | 0.35% |
| Broad interest | 192 | PKR 2,062 | 739,698 | 0.57% |
| Luxury micro-audience | 2 | PKR 1,553 | 6,806 | 0.84% |
| Business micro-audience | 2 | PKR 1,709 | 5,843 | 0.92% |
The fix
We merged the segments into one broader ad set carrying the full budget and moved the segment logic into creative — one ad set, several angles, each written for a segment. Under housing ad rules, that is where segmentation belongs anyway.
Applied forward, the same principle produced the tightest project in the portfolio: three ad sets and a PKR 887–922 cost-per-lead band.
Ad sets that cannot reach optimised delivery do not produce weak data — they produce meaningless data, which still looks like a result on a dashboard.