[ Case study ]
Meta optimized toward lead-form submissions, but free-lead volume didn't fill paid seats — the ads celebrated cost-per-lead while the actual event underfilled. The conversion signal the algorithm needed (paid attendance) never reached it.
CLIENT a professional-events company — FOCUS Send real outcomes back to the algorithm
Representative examplesEvery case study in this library is an illustrative composite of the kind of engagement we deliver — written to show our method and standards, not to name clients.
Conferences and paid professional workshops are the events company's product — ticketed seats, a free community tier that feeds the funnel, and registration running on an external platform the team likes. Cash follows the attendance sheet, not the registration list: refunds and no-shows resolve weeks after sign-ups. Paid social had been run on cost-per-lead targets for years, each event's marketing celebrated at registration close and quietly re-assessed when seats were counted in the room.
Meta optimized toward lead-form submissions, but free-lead volume didn't fill paid seats — the ads celebrated cost-per-lead while the actual event underfilled. The conversion signal the algorithm needed (paid attendance) never reached it.
We proposed sending the algorithm what the business actually sells: paid attendance. Registration-platform outcomes flow back through the Conversions API as value-bearing events, timed after the refund window so 'registered' stops being the target. Free community seats and paid professional seats received separate campaigns, with the free line deliberately feeding the paid line's retargeting audience. Bidding decisions lag attendance data by design — waiting for real outcomes replaced celebrating lead counts.
Just as important is what we ruled out, and why:
Paid-attendance events (post-refund-window) flow back via the Conversions API with values, so optimization targets what the business sells.
Free community seats and paid professional seats got separate campaigns — the free line feeds the paid line's audience, deliberately.
Bidding decisions lag attendance data by design — the discipline of waiting for real outcomes replaced the dopamine of lead counts.
Delivered by the growth pod — paid social specialist over 5 weeks, with working increments reviewed with the client every week.
Obstacle
The registration platform's export was a weekly manual CSV someone downloaded by hand — the feedback loop the whole design depended on was slow and fragile.
Handled: We negotiated a scheduled export from the platform's support, stood up an automated import, and the interim weeks ran on the manual file with its delay documented.
Obstacle
Value-based bidding on paid attendance started with too few recorded events per week to learn from, and delivery drifted while the algorithm searched for signal.
Handled: We seeded optimization with the earlier funnel stages at low weights until event volume accumulated, then shifted the learning signal fully to paid attendance without a delivery collapse.
Obstacle
The free-tier campaign initially outbid the paid line for shared audience overlap, quietly starving paid prospecting in the first weeks.
Handled: We tightened the free campaign's exclusions and re-weighted budgets so the free line fed the paid audience deliberately — the design's actual intent — instead of competing with it.
The headline: paid attendance attributed to paid social per campaign, year over year at similar spend — +55%, read from Attendance records + attribution reconciliation. A second check: cost per paid attendee at −19%.
Post-event reviews now open with the attendance reconciliation instead of the lead dashboard, and the marketing manager stops celebrating registration milestones mid-cycle because everyone knows the real number arrives later. The free tier is discussed as the paid line's audience engine rather than a vanity counter. Registration-platform leads, once the headline metric, became an early indicator — useful, no longer worshipped. The team plans campaigns around the lag, booking learning into the calendar instead of being surprised by it.
The result was read from Attendance records + attribution reconciliation against the pre-engagement baseline over the stated window, with a guardrail check on cost per paid attendee. Where platform-reported numbers and business outcomes differ, this record says which layer it is quoting.
What we would do differently
We would have negotiated the registration platform's data export earlier — weekly manual CSVs delayed the feedback loop the whole design depends on.
[ Related service ]
[ Related builds ]
−33%Cost per qualified enquiry over the quarter versus the prior one, on verified conversion tracking
0.8s 2.3sAverage 3-second hold rate on ads (proxy for hook survival), cohort over cohort
[ Next step ]
Next case study