How to Audit Paid Media Spend for Waste Before You Cut the Channel

When a paid media channel underperforms, the discussion usually moves quickly towards the channel itself. Meta is saturated. Programmatic does not work. Search has become too expensive. YouTube is not converting. The natural recommendation is then to reduce the channel or move the budget somewhere else.

Sometimes that is the correct decision. But a channel-level result is an average of everything that happened inside the channel, and that average can conceal very different economics across creative, placements, audiences, campaigns and account configuration.

A channel can be genuinely incremental overall while still containing substantial waste. Equally, a channel can look weak because a large share of the investment was deployed badly rather than because the underlying media opportunity has disappeared.

That distinction should be resolved before the larger allocation question.

This is the purpose of a paid media audit as I think about it. The objective is not to produce another account score or a list of generic optimisation recommendations. It is to locate where the money actually sits inside the channel, determine which parts are producing something commercially useful, and identify the spend that can potentially be removed or redirected without requiring an econometric model to tell us that a bad placement is bad.

At Growth Dynamics, I separate this into three interacting areas: campaign and account mechanics, placement and inventory economics, and creative effectiveness. I do not see those as a hierarchy of importance. Creative can easily be the biggest determinant of whether media works at all, while poor inventory can destroy the economics of good creative and broken account mechanics can prevent both from being deployed effectively.

The audit is therefore less about asking whether a channel works and more about understanding what the channel actually contains.

A channel average can hide radically different economics

Imagine a company spending £10 million a year on paid social with an acceptable blended ROAS. That figure combines every campaign, creative, audience, placement, bidding decision and period into one number.

Some of the investment may be highly productive. Other campaigns may be surviving because their weak performance is averaged together with stronger activity. A small group of creatives may account for most of the useful response while a long tail continues receiving spend. Certain placements may produce large amounts of cheap traffic but almost no commercially meaningful behaviour.

The blended channel result can remain stable throughout.

This is the same problem we encounter elsewhere in marketing measurement. Aggregation makes the system easier to understand, but it also removes information.

If paid social has an average ROAS of 3, I still know very little about where the next pound should go. I need to know how that return is distributed across the activity underneath it.

The same applies when the overall result is poor. A channel-level ROAS of 1.5 does not tell us whether the channel itself has no economic value, whether the creative is weak, whether most of the budget is flowing through poor inventory, whether campaign settings are producing unnecessary cost, or whether attribution is simply failing to capture the role the channel plays.

Those explanations require different decisions.

Cutting the entire channel because one blended metric is disappointing can remove useful media alongside the waste. Increasing the channel because the average looks healthy can do the opposite, allowing additional budget to flow towards the weakest parts of the account.

A useful paid media performance audit should therefore decompose the average before making the larger capital allocation decision.

Campaign mechanics are the easiest waste to inspect

Some media waste does not require sophisticated measurement because the mechanism is directly observable.

Search terms can show spend accumulating against queries that do not fit the commercial intent of the campaign. Audience configuration can reveal unnecessary overlap. Conversion tracking can be incomplete or incorrectly configured. Campaign objectives can encourage an algorithm to optimise towards an event that is easy to produce but weakly connected with business value. Landing pages can create obvious friction after otherwise useful traffic has been acquired.

These are operational problems rather than causal inference problems.

Google's own documentation illustrates this in search advertising. Ad quality and Ad Rank incorporate factors including expected click-through rate, ad relevance and landing-page experience, which means the relationship between the query, advertising and destination is part of the auction mechanics themselves.

I therefore want to understand the account structure before interpreting the channel average. Which campaigns consume most of the budget? What objectives are they optimising towards? Which search terms, audiences or campaign components are responsible for the cost? Is the conversion event genuinely valuable? Does the landing page match the intent created by the advertising?

The point is not to produce an exhaustive checklist of everything somebody could change inside an advertising platform. Most large accounts contain thousands of possible optimisations, and many of them have almost no financial significance.

The useful question is which mechanical issues have enough money behind them to matter.

A campaign spending £2,000 a year inefficiently is different from an account configuration affecting £2 million of media. The audit should therefore translate operational findings into dollars wherever possible, rather than presenting a list of platform settings with no indication of their economic importance.

This first layer is often valuable precisely because it is so traceable. You can point to the spend, show what happened, explain why it deserves investigation and estimate the amount of capital exposed to the problem without making a causal claim the data cannot support.

Placement economics can disappear inside cheap media

Programmatic advertising makes the aggregation problem even more obvious because the same campaign can distribute impressions across an enormous range of sites, apps and supply paths.

The average CPM may look efficient while the quality of the underlying inventory varies considerably.

This is not a theoretical concern. The ANA's 2023 Programmatic Media Supply Chain Transparency Study analysed $123 million of advertiser spend and found that Made for Advertising sites represented 21% of impressions and 15% of spend in the study. The ANA also found that the average campaign ran across 44,000 websites and argued that an excessive focus on cheap CPMs was encouraging advertisers to buy low-value inventory.

The industry has improved since then, but the problem has not disappeared. The ANA's 2024 benchmark reported that 43.9% of every $1,000 entering a DSP reached consumers under its True Ad Spend Efficiency measure. It also found that participating marketers had reduced their MFA exposure substantially, showing that active supply-chain management can improve the economics rather than waste being an unavoidable feature of programmatic buying.

This changes how I want to audit placements.

A cheap click is not automatically a valuable click. A site producing huge numbers of visits is not necessarily an efficient source of demand. I want to connect inventory with what happens after the impression or click, using the strongest business outcome available and meaningful intermediate behaviour when final revenue takes too long to mature.

That means looking beyond CPM and CPC towards placement-level economics. Which inventory absorbs meaningful spend? What kind of traffic follows? Does that traffic progress towards commercially important actions? Are sales or qualified outcomes concentrated elsewhere?

The purpose is not to eliminate every placement that has not converted. Low-volume data is noisy, and cutting inventory after one bad observation is not measurement discipline either. The analysis needs enough exposure or spend to make the finding commercially meaningful.

But sometimes the pattern is difficult to ignore.

In one Growth Dynamics placement audit, 17.9% of the spend in a roughly $401,000 Criteo account was flagged for investigation. That portion of the account carried 63.7% of the clicks and produced only three sales. The point is not that 17.9% is a benchmark companies should expect to find. It is that the channel average concealed a very different economic picture once the spend was decomposed by inventory.

If that analysis had stopped at the overall Criteo result, the decision would have been whether Criteo was good or bad. Looking inside the channel produced a more useful question: which inventory deserved to continue receiving money?

Creative is not decoration around the media investment

Creative needs to be treated differently from a normal optimisation variable because it can materially change the effectiveness of the media itself.

This is where I would be cautious about an audit process that concentrates almost entirely on account settings and audience optimisation. You can make the buying mechanics extremely efficient and still spend a large amount of money distributing advertising that people do not care about.

System1's work on the cost of dull advertising makes this economic relationship explicit. Its research with Adam Morgan and Peter Field examines the additional media investment required for weaker advertising to produce the same long-term market-share effect as stronger creative.

That has an important implication for a paid media audit. Media efficiency cannot be separated completely from creative efficiency.

Suppose two campaigns have similar distribution, bidding and audience conditions, but one contains advertising capable of attracting attention and creating a meaningful response while the other does not. The difference in performance should not automatically be interpreted as evidence that the platform allocated the media badly.

The creative itself changed the treatment.

I therefore want to understand the distribution of spend across creative rather than looking only at the average channel result. Which assets have received enough scale to evaluate properly? Which are earning both scale and efficiency? How concentrated is performance? How much budget is sitting behind a weak creative tail? Are we repeatedly spending behind assets that have enough evidence to suggest they are unlikely to improve?

The exact metric depends on the objective. Short-term conversion metrics alone can favour creative aimed at people already close to purchase, while upper-funnel creative may require other behavioural, brand or experimental evidence. The important point is that the analysis should connect creative with the outcome the advertising is actually supposed to influence.

A creative audit should therefore avoid becoming a league table of CTRs.

The commercial question is whether the portfolio of creative is making the media investment work harder or forcing the business to buy more media to compensate for advertising that is doing very little.

Creative, placement and campaign mechanics interact

It is tempting to turn these three areas into an ordered checklist, but that can create another misleading simplification.

Campaign mechanics, placement quality and creative effectiveness interact with each other.

Strong creative placed in poor inventory may never receive useful attention from the right people. Excellent inventory carrying weak advertising can mean the company is efficiently distributing something ineffective. Both can be undermined by a campaign configured around the wrong objective or conversion signal.

That means the audit should preserve context when decomposing the account.

If one creative appears to perform badly, I want to know where it was delivered and to whom before declaring the asset weak. If one placement produces poor economics, I want to understand whether it consistently performs badly across different creative and campaigns or whether the result belongs to one particular combination.

This is particularly important as advertising platforms automate more of the allocation process. The advertiser may define the available creative, objectives, audiences and constraints, but the platform determines a large amount of the final delivery.

The result we observe is produced by the interaction between all of those components.

A channel is therefore not a single homogeneous intervention. Saying that "Meta produced a ROAS of 2.4" compresses a large portfolio of different treatments into one number.

That level of aggregation can be useful for a board conversation, but it is a poor place to stop if you are trying to find waste.

Look for concentration before chasing averages

One of the most useful things an audit can do is show how concentrated the economics are.

Average metrics can conceal a long tail because they combine high and low performers into a single ratio. A channel might contain a small group of placements responsible for most of the useful outcomes, another group with uncertain performance because they have not received enough volume, and a third group consuming meaningful spend with very little evidence of value.

Creative can exhibit the same pattern.

The purpose of analysing concentration is not to remove everything below the average. Doing so mechanically would eventually leave you with an increasingly narrow portfolio and little ability to discover new opportunities.

The purpose is to understand how much money is being spent in parts of the distribution that have accumulated enough evidence to deserve a decision.

This is where I prefer a spend-first analysis.

Begin with where the money actually sits, then work out what evidence exists around that spend. There is little value in finding a spectacularly inefficient placement that received £40 while ignoring a moderately inefficient cluster consuming £500,000.

The same logic applies to creative. An asset with a terrible conversion rate but almost no spend may simply be noise. A group of creatives with poor economics collectively absorbing a substantial part of the budget deserves much more attention.

This sounds obvious, but many audits are still organised around the number of issues discovered rather than the value of the money attached to them.

A ten-page list of optimisation opportunities is less useful than identifying three areas where the economics suggest significant capital can be protected or redeployed.

A paid media audit cannot tell you whether the entire channel is incremental

This is an important boundary.

Finding waste inside a channel does not prove whether the channel as a whole creates incremental revenue.

The first four layers of the Growth Dynamics measurement approach are deliberately observational. They are designed to show what can be seen directly before introducing a counterfactual. A paid media audit belongs in that observed layer.

If a placement receives significant spend, generates poor-quality traffic and produces almost no downstream commercial behaviour, that is useful evidence. It may be strong enough to justify an operational decision depending on the circumstances.

It is not the same as proving that the entire channel would produce zero incremental revenue if it disappeared.

That broader question may require a designed experiment.

The distinction matters because companies can make the opposite mistakes. Some jump directly from a weak channel-level result to cutting the entire channel. Others believe that nothing can be changed until an incrementality test or MMM has established causality.

Both positions are too rigid.

There is often a large amount of observable waste that can be investigated before paying the assumption and operational cost of a causal design. Once that waste has been removed, the remaining channel is a cleaner treatment to evaluate anyway.

If the final decision is whether to remove another £10 million from the channel entirely, the evidence requirement should increase accordingly.

Measurement should become heavier as the decision becomes harder.

The audit should end in money, not recommendations

A good paid media audit should not finish with a slide containing forty optimisation opportunities.

It should finish with an economic map of the channel.

Which spend looks productive enough to protect? Which areas need to be changed before more money is added? Which placements, campaign components or creative clusters have accumulated enough evidence to justify reducing investment? Which findings are observational and which would require a causal test before a larger decision is made?

That creates a much cleaner distinction between optimisation and allocation.

An account team can act on campaign mechanics.

A media team can change inventory and supply paths.

Creative teams can see where weak work is forcing the media budget to work harder.

Leadership can see which amount of money is genuinely exposed and which larger questions remain unresolved.

The objective is not to make the channel look more efficient by removing every weak-looking line from a report. It is to improve the economics of the actual media investment.

This is also why I would not use ROAS alone to prioritise the audit. If the attribution system itself overvalues activity close to conversion, optimising exclusively towards attributed ROAS can make the account look better while moving it further towards demand capture.

Where possible, I want the audit connected with broader business outcomes, meaningful mid-funnel behaviour and the other evidence available around the customer journey.

The channel should not be allowed to grade its own homework.

Audit the execution before making the channel verdict

There are cases where a paid media channel should be reduced dramatically or removed. No amount of account optimisation changes the fact that some investments have reached diminishing returns or are failing to create enough value.

But the evidence for that decision should be stronger than a blended dashboard number.

Before deciding that the channel itself is the problem, understand what was bought with the money.

Look at the campaign mechanics. Understand the inventory. Examine where the creative budget sits and whether the work being distributed is capable of producing the response expected from it. Find out how concentrated the economics are and which parts of the channel are dragging the average down.

Then decide whether the remaining question is still an execution question or has become a causal allocation question.

That sequence matters because the two decisions are different.

Removing obvious waste inside a channel can often be supported by observable evidence. Deciding whether the cleaned-up channel deserves another £5 million, or whether the whole investment should move somewhere else, may justify an incrementality test or a broader model.

The mistake is using the most aggregated number available to answer both questions.

Growth Dynamics audits paid media at the level where the money is actually being deployed, across campaign mechanics, inventory and creative, before escalating to heavier measurement. The output is an evidenced invest, divest or hold decision showing where spend appears productive, where it is being wasted, how confident we are in that conclusion and which larger questions still require causal evidence.

A channel average tells you how the portfolio looks from a distance.

A useful paid media audit tells you where the money is going.