Running three paid channels without kidding yourself

Overview

DécisionIA trains and advises executives and consultants on artificial intelligence. Its acquisition model rests on a simple principle: a high-value free asset (a live masterclass) that attracts the right audience and then feeds a paid offer. My remit: run all paid campaigns across Meta, Google and LinkedIn, with a weekly review.

The problem

Three channels, several offers, and one permanent question: where does the next euro go? The trap with this kind of setup is judging a channel on a feeling rather than on data, and letting campaigns run because they "generate volume" or because you believe in them. The job was to decide, numbers in hand, what to scale and what to kill.

Objectives

• Bring down the cost per masterclass signup without breaking volume
• Arbitrate between the three channels on real conversions, not clicks
• Make measurement trustworthy before committing any spend
• Reallocate budget away from anything that does not convert

What I did

• Monthly management of three channels with a strict comparison scope: masterclass is compared to masterclass, never to totals mixing different campaigns.
• Creative iteration on Meta: identifying the visuals actually carrying performance and removing those costing money without converting (including, counter-intuitively, a testimonial visual).
• Watching audience spill-over: Meta's automatic expansion sometimes widened into age brackets far more expensive than the core target. Diagnosed, and arbitrated.
• A target cost-per-acquisition cap on Google once the channel started drifting: visible effect within days.
• Stopped a retargeting campaign that never converted in paid acquisition, and redeployed the budget.
• Stopped a LinkedIn campaign before its budget was spent: targeting tightened, cost per click and click-through rate improved, but still no bookings. Signal enough to cut.
• A consistent UTM convention across all three channels, down to the creative identifier, so creatives can be compared against each other.

The real decision: stop before you have spent it all

The LinkedIn campaign targeting consulting firms had an allocated budget. After the targeting was tightened, cost per click and click-through rate improved, but still no bookings. At that point there were two options: let it run until the envelope was consumed, or cut it. I cut it, before the planned budget was spent.

That is where this differs from management that lets things drift. Once the signal is there, spending more only over-confirms something you already know. Every extra euro buys the same answer with more decimal places. That money is far more useful invested in a different hypothesis.

Cutting is not condemning the channel. Two avenues stay open and will be tested with fresh variables: the ad format first (a static visual probably lacks impact on this channel, so video needs testing); then the offer itself, which would likely benefit from being rethought for this audience, because what works with consultants does not transpose as-is to the heads of consulting firms.

Starting again with different variables costs less than proving the same thing three times. It is a management reflex, not excessive caution: budget not spent on a bad lead is budget available for the right one.

Results

On a strictly like-for-like basis month over month, the cost per masterclass signup fell by 15%, with signup volume held and budget down. Meta, previously the more expensive channel, became the better of the two after the creative work. On Google, the cost-per-acquisition cap reversed the drift in under a week.

Business impact: budget is no longer spread across three channels "to see": it is concentrated where the data shows conversion, and pulled everywhere else without sentiment. The two campaigns that were cut are not failures: they are budgets recovered before being burned for nothing, where passive management would have let the envelope run to the very end. That is what now funds the launch of a new offer aimed at company directors.

Acquisition costs are deliberately given as orders of magnitude: the detail is the client's commercial performance.