The principle
Anyone can open an account and find things they'd change. The job is working out which problem is actually costing money, proving it with evidence rather than opinion, and putting the fixes in an order the business can realistically act on.
Three rules govern the work.
Fix the measurement before judging the performance. If the numbers can't be trusted, every conclusion drawn from them is guesswork wearing a suit. Measurement and attribution get reviewed first, always.
Separate what is observed from what is inferred. A finding is checkable. An implication is a hypothesis about what it means. Those are different things and they get labelled differently.
The channel is rarely the problem. Most businesses ask about paid media, because it has a bill attached. The leak is often elsewhere — a persona nobody validated, a CRM that doesn't segment, a form failing silently, a page nobody can find, or an agency relationship with no accountability in it. The audit goes where the evidence points, not where the invoice is.
The sequence
Every audit runs in the same order, because each stage depends on the one before it.
1. Access and orientation. Into every relevant system before forming an opinion — analytics, tag manager, ad platforms, CRM and email, CMS, search console, call and lead tracking, and whatever commercial data sits outside all of it. The gap between what the platforms report and what the business actually banked is usually where the real story is.
2. Measurement and attribution. What is each conversion actually counting, and has its definition changed? Have settings moved mid-period? Are there outages or duplicate tags? Does platform reporting reconcile with CRM and with sales? Anything that breaks period-on-period comparability is documented here.
3. Diagnosis. Across whichever modules are in scope — listed below.
4. Prioritisation. Findings scored and sequenced. Two or three priorities, not a list of twenty.
5. Baseline and readout. Agree the number future work is measured against. Without one set at the start, there's no honest way to demonstrate improvement later.
What gets audited
Scoped per engagement. A focused audit uses three of these. A full marketing function review uses all seven.
Audience and persona. Whether the business is targeting who it thinks it is. Personas built from real engagement evidence rather than a workshop guess, mapped through See / Think / Do / Care with the channel and content named at every step. An unvalidated persona is the most expensive assumption in most marketing plans, because everything downstream inherits it.
Search and organic. Technical health, indexation, and whether the site is discoverable for terms that carry commercial intent rather than vanity volume. Increasingly this includes how AI systems represent the brand, since that's now a discovery surface in its own right.
Paid media. Where spend is concentrated versus where returns are. Trend direction rather than headline totals — a strong year-on-year figure can hide six months of decline. Competitive position, unmanaged search terms, and anything paused without a recorded reason.
Conversion and journey. The actual path from first touch to enquiry, walked rather than assumed. Where the friction is, what objection the page is really being asked to overcome, and whether it answers that before asking for commitment.
CRM, retention and lifecycle. What happens after the first conversion, which is where most businesses leak value invisibly. Segmentation quality, lifecycle coverage, and whether data collected at signup is ever actually used.
Technology and data stack. What's installed, what's actually used, what's duplicated, and what's silently broken. Tag governance, data flow between systems, consent handling, and whether the stack can answer the questions being asked of it.
People, agency and accountability. Who owns which outcome, whether agency performance is reviewed against agreed measures or against a relationship, and whether reporting exists to inform decisions or to justify activity.
How findings are written
Every finding uses the same structure, whichever module it came from. This is what makes an audit credible rather than a list of criticisms.
Observed — what is genuinely, checkably true, sourced.
Possible implication — what it might mean commercially, labelled as a hypothesis.
Alternative explanation — the credible reason the signal may be benign.
Smallest useful test — the cheapest way to find out which is right, before committing to a large fix.
What to check internally — the questions only the business can answer.
The alternative explanation isn't politeness. It's the discipline that stops an audit confusing "I can see this" with "this is broken". Plenty of things that look wrong from outside are deliberate decisions with good reasons behind them, and an auditor who can't tell the difference is expensive to have around.
You can see this structure applied to real, public examples in the Outside In series — every one of them uses the same logic on a brand I have no commercial relationship with.
What an audit does not do
It doesn't rebuild anything before the diagnosis is agreed. It doesn't recommend more spend before proving the current spend works. It doesn't judge brand activity on direct-response metrics, or direct-response on reach. It doesn't present a hypothesis as a finding. And it doesn't confine itself to the channel you asked about if the evidence points somewhere else.
What you get
A written diagnosis separating verified findings from hypotheses. A measurement health check stating what can and cannot currently be trusted, and why. An agreed baseline to measure future work against. Two or three prioritised actions with the reasoning shown. And the questions to take to your own team.
Method built from enterprise demand generation at Tata Communications, performance planning across the Mars Petcare portfolio, consulting delivery at Palladium, and high-net-worth acquisition at Bupa Global. Full engagement models are on the services page, or book a 15 minute call.