Marko Cvijić / Paid media oversight
Independent governance for ad budgets large enough that a bad month is a real number. Not another pair of hands on the account: the layer above it that verifies what your bidding is actually learning from, holds the agency to a standard, and answers to the person who funds the spend.
The problem
A large ad budget almost never fails because the bidding was set up badly. It fails because nobody above the bidding is accountable for it. The agency reports the metrics that make the agency look effective. The platform reports the conversions the platform is allowed to claim. Finance sees a cost line. Somewhere between those three sits the actual answer, and it is nobody's job to produce it.
The pattern I find in audits is consistent. Smart Bidding is optimising toward a conversion action that no longer fires. A primary conversion was imported from an analytics event eighteen months ago and the site stopped sending it after a replatform. The match rate between the ad platform and the analytics property is low enough that half the signal is missing, and the algorithm is learning confidently from the half that remains. None of that shows up as a broken campaign. It shows up as spend that stops producing and a report that cannot explain why.
This is not another pair of hands on the account. It is the accountability layer above it - verification of the signal your bidding learns from, and an owner for the number.
What the work covers
01
What the algorithm is actually learning from
Conversion action audit
Every conversion action in the account: what fires it, when it was created, whether it still fires, whether it is counted once or three times, and whether it is a real commercial event or a button click somebody tagged in a hurry.
Match rate verification
How much of the platform's data actually reconciles with your analytics and your back end. Under ninety per cent triggers a full attribution audit before I will comment on performance. For B2B with long cycles the acceptable floor is lower - around forty-six per cent is normal there, and treating that as a fault wastes a month.
Enhanced conversions and consent
First-party data flowing correctly, consent mode implemented so it neither breaks the law nor silently discards conversions, and server-side tagging where the client-side path has become unreliable.
Value, not just volume
Conversion values that reflect margin rather than order total, offline conversions imported for businesses that close on the phone, and lead quality fed back so the algorithm optimises toward customers instead of form fills.
02
For budgets where a bad month is expensive
Account structure review
Whether the structure serves the business or the reporting. Brand separated from non-brand so brand traffic stops inflating everything. Budget caps that are not silently throttling the campaigns that work.
Incrementality
The question nobody wants asked: how much of this revenue would have arrived anyway. Brand search, retargeting and shopping on your own product names are the usual places where paid takes credit for demand you already owned.
Agency oversight
I do not replace your agency. I hold them to a standard, translate their reporting into terms your CFO recognises, and take the questions they cannot be objective about because their fee depends on the answer.
Budget allocation
Across platforms, markets and funnel stages, with a written rationale for each allocation and a stated trigger for changing it. A reallocation nobody can explain is indistinguishable from a guess.
Creative and landing accountability
Where the loss is downstream of the click. Media buying cannot fix a landing page that contradicts the ad, and it is usually cheaper to fix the page than to keep bidding harder.
Paid and organic together
Queries you are paying for that you already rank first on, and categories where organic has collapsed and paid is quietly covering the gap at full price. This is the whole reason to have one owner across both.
Operating rules
01
No performance commentary before the tracking is verified. Opinions about a channel built on unverified data are worth exactly what they cost to produce.
02
A match rate below ninety per cent triggers an attribution audit - with the exception of long-cycle B2B, where the realistic benchmark is far lower and the audit target is consistency instead.
03
Every primary conversion gets re-verified quarterly. Sites change. Events break. Bidding keeps going regardless, which is what makes it dangerous.
04
Brand is reported separately from everything else, permanently. Blending them produces a flattering number that cannot be used to make a decision.
05
No change to bidding, structure or budget without a written reason and an expected outcome. If it did not work, we should be able to say what we thought would happen.
06
The report goes to the person who owns the P&L, in their language. Impressions, CTR and platform-attributed ROAS are diagnostics, not results.
Fit
Ad spend large enough that a month of misallocation is a material number. An agency doing competent work that nobody internally can evaluate. A board asking questions the current reporting cannot answer. Multiple markets or platforms with no consolidated view. Or a new CMO or CFO who inherited a budget and needs an independent read before defending it.
You want someone to take over day-to-day campaign management - that is your agency's job and I am not competing for it. You want ammunition to fire an agency rather than a genuine assessment. Or the spend is small enough that the oversight would cost a meaningful fraction of the media, in which case fix the tracking once and skip the governance layer.