Cutting ticket cost by 90% through pixel strategy

The problem

Bethel Music Worship School runs an annual ticketed event. The December 2023 campaign was live, the ads were running, and the numbers looked fine on the surface - 24,277 clicks, $0.09 cost per click, 3.22 million impressions. But only 16 purchases came through, which put the real cost at $130 per ticket.

The brief was straightforward: drive ticket sales for Worship School 2024 through Meta and Google, set up proper tracking, and report on what was working. What we found instead was a pixel attached to 109 different websites, mixing apparel orders, event registrations, and general site traffic into one unusable dataset.

The audit

The initial campaign ran as a traffic objective - optimized for clicks, not conversions. Meta delivered exactly what it was asked for: 24,277 clicks at $0.09 each. The $2,083 spend looked efficient until you measured it against actual purchases.

The pixel diagnostic showed why. The Bethel Music pixel was installed across 109 sites: the main store, event microsites, artist pages, and landing pages dating back years. Meta's algorithm couldn't separate a ticket buyer from someone browsing T-shirts or streaming a live event. The campaign was optimizing for the wrong action.

The rebuild

We built a new pixel specifically for Bethel Music Worship School and installed it on two surfaces: the Brushfire ticketing checkout and the BMWS landing page. Three events went live - add to cart, registration start, and purchase - so Meta could see the full funnel.

The audience strategy started with the data Bethel already had: a Mailchimp list of past attendees going back to 2018, segmented by country. We imported those lists into Meta and built 1% Lookalike audiences for the US, Canada, UK, Australia, Brazil, and South Africa. That gave the algorithm a clean starting point: people who look like past ticket buyers, not people who once clicked on a worship playlist.

Sales campaign (Dec 31 – Jan 10)

The new campaign ran for ten days with a sales objective, and the targeting split into two tracks: cold Lookalikes and warm retargeting of the imported past-attendee lists.

The retargeting track returned ROAS between 4x and 9x across the board, with the top-performing ad sets hitting 91.67x and 39x. The single best ad - a static image with event copy - came in at 186x. These are conservative figures; without server-side Conversions API, iOS purchases are underreported.

What this proved

Traffic campaigns optimize for the thing you ask them to optimize for. If you ask for clicks, Meta will buy you clicks, and the cost per click will look great right up until you divide spend by actual purchases.

A pixel shared across 109 sites is not a pixel; it's a data dump. The algorithm can't optimize for a ticket sale if it thinks a ticket sale, a T-shirt purchase, and a song stream are the same event.

Past-attendee lists are warm retargeting gold if you have the pixel wired to see purchases. The top-performing ad in this campaign was a simple static image - nothing fancy - because it was being shown to people who already knew the event.

What I'd do next

The obvious next step is Conversions API. Meta's browser pixel misses a significant percentage of iOS conversions thanks to App Tracking Transparency. Server-side tracking would close that gap and give the algorithm better data to work with, which would improve ROAS further.

The Lookalike audiences worked, but we only had time to test the 1% tier. A 2%–5% expansion test would show whether there's a larger addressable audience without sacrificing efficiency, which matters if Bethel wants to scale the event.