When we took over TDT's five Meta ad accounts, they were running at roughly 1.2X ROAS. For a ticketed-event business, that isn't a marketing problem — it's an existential one.
The account wasn't broken by a sophisticated attribution bug. It was broken by the accumulated cost of small, default decisions: · A half-configured pixel — automatic tracking off, no audience segments, no retargeting. Every campaign was optimising against a ghost. · Meta's "helpful" defaults left on — Creative AI testing, Advantage+ enhancements, campaign end dates and standard instead of incremental attribution, all quietly bleeding budget. · Targeting wrong on two axes — age stretched to 40+ when nearly all revenue came from 18–34, and ~20% of spend burning on zero-conversion placements like Marketplace and Explore. · Creative under-rotated — Reels barely tested, high-ROI carousels underweighted, and frequency hitting 6–8 (the same person seeing the same ad seven times).
Over February, March and April 2026, across six cities, the numbers moved: ROAS climbed from 1.2X to ~4X average, peaking near 5X on events running the full 12-creative rollout, with 5,000+ tickets sold through Meta in 90 days.
We won by doing a lot of small things correctly. We fixed the infrastructure first — UTMs on every ad, the pixel switched on across all five accounts, six core custom audiences on a 180-day window. Then we turned off everything Meta had turned on, capped targeting at 18–35 across four high-intent placements, and rebuilt the creative pipeline to 12 assets per event. The lesson we keep relearning: the platform is not on your side — it's on its own. Its job is to spend your budget; yours is to sell tickets for Saturday. When those incentives diverge, someone has to overrule the platform. That someone is us.
Outcome
We made five sets of changes across all six cities. 1. We fixed the infrastructure first UTM parameters on every active ad Pixel tracking turned on across all five ad accounts Six core custom audiences built on a 180-day window: PV, VC, ATC, IC, AWV, Purchase Audience segments defined at the campaign level (engaged audience + existing customers) This is the unglamorous part nobody puts in a case study. It is also the reason the next four things worked. 2. We turned off everything Meta had turned on Creative AI testing → off Advantage+ creative enhancements → off End dates on campaigns → removed (we kill ads manually, always) 13 active AI generation features on one account → off Meta makes money when you let it "optimize." You make money when you don't. 3. We tightened the targeting Age capped at 18–35 across every campaign Placements cut to four: Instagram Reels, Instagram Feed, Instagram Stories, Facebook Feed Everything else excluded Locations set per event, not per account default 4. We rebuilt the creative pipeline 12 creatives per event: 4 Reels, 4 Carousels, 2 Static, plus variations Each creative attacked a different angle: the dance floor, the men and women in the room, the DJ, the venue, the FOMO of last year's sellout, and early-bird urgency The Beond Approach Over February, March, and April 2026, across six cities, the numbers moved: ROAS: 1.2X → ~4X average across the 90-day window Peak ROAS: ~5X on events using the full 12-creative rollout Tickets sold: 5,000+ through Meta in 90 days (including walk-ins attributed to Meta) CAC: Dropped as a function of cleaner pixel data, excluded placements, and disabled AI features Frequency: Brought back into the 2 to 3 range on healthy campaigns, sustained by the 12-creative-per-event rotation The reason this worked is not that we did something clever. We did a lot of small things correctly, and one thing, the gendered pricing, that the client had not been asked to consider. Agencies like to sell the clever thing. In our experience, the clever thing is almost always the last 10%. The first 90% is refusing to let Meta run your account for you.
Next projects.
Book a Call
Have a project in mind? Reach out to us, and we’ll discuss the best way to move forward.
Anurag Attri
Founder & Creative Director