03Case File 03 · High-Risk Funnel

Pay For Registrations,
Get Registrations.

The team was buying registrations to keep delivery stable. It worked. Delivery stayed stable and produced 7,320 registrations and 198 deposits — a rate of one in thirty-seven.

VerticalHigh-risk, clear deposit event
ChannelMeta
Window10 weeks
Spend$118,500
ChangeEvent hierarchy

What happened

Quick answer

Registrations fell from 7,320 to 3,980 while deposits rose from 198 to 356. The account was not underperforming before — it was performing exactly as instructed, toward the wrong instruction.

Ten weeks, $118,500, one channel. The only structural change was which event the budget was paid to produce.

The numbers

Before And After,
Same Budget

MeasureBeforeAfterNote
Ad spend$118,500$118,500―Same budget, both periods
Registrations7,3203,980▼46% fewer
First-time deposits198356▲80% more
Registration to deposit2.7%8.9%▲3.3x
Wasted volumeHighLow▲Fewer people who were never going to pay

Diagnosis

What Was Broken

ISSUE 01Registrations were bought to protect delivery stabilityA defensible instinct. Deeper events are rarer, and rare events destabilise delivery. The cost of that protection was never calculated.
ISSUE 02No hierarchy between eventsRegistration and deposit sat at the same level in reporting, so the cheaper one won every internal argument.
ISSUE 03Volume was read as healthRising registration counts were presented as growth. The deposit line was flat throughout and nobody plotted the two together.
ISSUE 04Nobody owned the number underneathMedia reported registrations. The desk reported deposits. Two teams looked at the same funnel through different ends of it and neither had a reason to reconcile the two. The gap survived eighteen months because it belonged to no one.

The work

What We Changed

CHANGE 01Established an explicit event hierarchyRegistration became diagnostics. Deposit became the optimisation target. Written down before anything changed, so the volume drop surprised nobody.
CHANGE 02Verified deposit volume could sustain learningChecked first that the deeper event occurred often enough to leave the learning phase. This is the step most teams skip and then blame the change.
CHANGE 03Held the change for three full weeksPerformance dips when the event moves. Reverting during that dip is why most teams conclude it does not work.
CHANGE 04Gave the deposit number a single ownerOne person reporting registration to deposit weekly, in one place, to both teams. Not a dashboard nobody opens. The measurement problem was never technical — it was that no individual was accountable for the figure that decided whether the budget worked.

The uncomfortable part

What Did Not Work

Published because a case file without this section is an advertisement. These cost time and money before they were abandoned.

⚠Abandoned after testing

Switching the event on all campaigns at onceDelivery destabilised across the account simultaneously with nothing left running as a control. We moved half, held half, and compared.
Expecting the deposit rate to improve immediatelyIt got worse for eleven days before it got better. Anybody watching a weekly report would have reversed the decision on day seven.
Keeping the same audience definitionsThey had been built around people who register. Rebuilding them around people who deposit took another fortnight and produced most of the remaining gain.

What it means

The Takeaway

→If the algorithm is paid for registrations, it will buy registrations, efficiently and indefinitely. Change what it is paid for and the entire funnel reorganises itself around the new instruction.

Related

Where the deposits actually wentOpen → High-risk media buying in fullOpen → The same change in casinoOpen → All twelve documented filesOpen →
Discuss your account → All documented files