What a real digital marketing audit covers

The sequence is not stylistic. Each phase depends on the one before it.

1. Measurement. Is the data true?
2. Economics. What can you afford to pay for a customer?
3. Demand capture. Are you harvesting existing demand efficiently?
4. Demand creation. Are you creating any?
5. Site and conversion path. Does the traffic you buy have somewhere good to land?
6. Lifecycle. Does a customer buy twice?
7. Operations. Who does the work, and how long does a change take?

Audit phase three before phase one and you will optimize toward a number that is wrong. That single mistake accounts for most of the audits that recommend confidently in the wrong direction.

This is the longest phase and the one clients least expect to need.

• One analytics tag per page, firing once. Duplicate containers are common after a site rebuild and silently halve bounce-related metrics and double sessions.
• Conversion events defined once, with one primary action per outcome. Check for a purchase event and a thank-you page view both counted as conversions.
• Transaction and lead values populated, in one currency, with tax and shipping handled consistently.
• Timezone and currency consistent between the ad platforms, the analytics property and the commerce or CRM backend. Mismatches here are why two reports never agree.
• Consent handling in place and correctly configured for the regions you serve. For advertisers running Google's ad services against traffic from the European Economic Area, consent signals have been a requirement since early 2024, and misconfiguration shows up as unexplained data loss rather than an error message.
• Server-side or API-based conversion forwarding, where used, deduplicated against the browser events. Check that a shared event identifier is actually being sent.
• Click identifiers captured and stored on the order or lead record, so offline outcomes can be matched back later.
• UTM conventions documented and followed. Inconsistent case and naming fragments channel reporting into near-duplicates.
• Lead quality feedback for lead-gen businesses. If the CRM does not return qualified-lead status to the ad platforms, bidding is optimizing toward form fills.
• Reconcile a single week: platform-reported conversions, analytics conversions, and the actual orders or qualified leads in the system of record. Write down the gaps and their causes.

Output of this phase is a one-page statement: which numbers can be trusted, for what, and which cannot.

• Contribution margin per order or per customer, after all variable costs including returns.
• The acquisition ceiling that follows from it, and whether current targets were derived from it or inherited.
• Payback period, and the cash constraint that sets how fast spend can grow.
• Price architecture and offer: shipping thresholds, bundles, entry products, discount dependency.
• Mix: which products or service lines carry the margin, and whether marketing is currently pushing them.

• Account structure against the reporting questions the business needs answered.
• Brand and non-brand separated, with cross-negation.
• Share of spend and share of conversions going to brand. A program that is mostly brand is harvesting, not growing.
• Search terms report: wasted spend, and queries with demand you are not covering.
• Bid strategies matched to data volume, and targets matched to margin.
• Catalog and feed quality for shopping: titles, attributes, availability, disapprovals.
• Remarketing audience definitions, exclusions, and frequency.
• Landing page match: does each ad group point somewhere that answers the query.

• Who the audience is beyond "people who already searched for us."
• Creative volume and variety, and whether a testing cadence exists at all.
• Branded search volume trend over time. It is the most accessible proxy for whether anyone new is hearing about you.
• Share of voice and message distinctiveness against the two or three competitors that actually come up in sales conversations.
• Channels present versus channels where the audience actually is.

• Page performance on the templates that carry traffic, measured on a mid-range mobile device rather than a developer laptop.
• The funnel step by step: landing, product or service page, cart or form, checkout or submission, confirmation. Find the step with the largest drop.
• Mobile checkout or form completion specifically, separated from desktop.
• Clarity of the primary action on every template that receives paid traffic.
• Trust and risk reduction: delivery expectations, returns terms, pricing transparency, proof.
• Internal search and navigation, if the catalog is large enough for people to use them.
• Forms: field count, validation behavior, error messaging, and whether anything required is genuinely required.

• Welcome, abandonment and post-purchase sequences: do they exist, do they send, do they say anything useful.
• Segmentation beyond "everyone" and "recent buyers."
• Repeat purchase rate at a fixed window, by cohort.
• Discount dependency in retention messaging.
• Deliverability: authentication records, list hygiene, complaint and bounce rates.
• Where automation would remove a manual step that is currently being skipped under pressure.

• Who owns each channel, and whether that person has access and authority.
• How long a landing page change takes from request to live. If it is weeks, no conversion program will work regardless of findings.
• Approval path for creative and copy.
• Reporting cadence, and whether anyone makes a decision from it.
• Documentation: is there a record of what was tried and what happened.

A finding nobody can act on is not a finding. Every item in the output carries the same five columns.

Then rank. Sort by effect divided by effort, group into "this fortnight," "this quarter" and "needs a decision first," and keep the list to the ten items that matter. Everything else goes in an appendix.

• The audit is run on platform data without reconciling against the system of record, so the findings inherit the measurement errors they should have caught.
• Benchmarks are quoted from somewhere unspecified. Compare the account against itself over time instead.
• Recommendations assume capacity that does not exist. Three changes that will happen beat thirty that will not.
• Nobody is named as owner, so the document becomes a reading exercise.

We run the audit as the opening phase, before any media plan or build scope, because the sequence above determines what the work should even be. It ends in a ranked plan with owners and a measurement baseline, which becomes the reference point for everything after. That sits inside digital strategy and advisory; if you want to know what it would look like against your own stack, get in touch.