Marketing Operations Audit: The Checklist I Run in Week One

Every engagement I take starts the same way, whatever the brief says: a marketing operations audit. Not because audits are billable — because opinions formed before the audit are guesses, and I have watched too many strategies built on guessed foundations. The audit below is the one I actually run in week one. It takes three to five working days with cooperative access, and it fits on a page.

The method matters as much as the checklist: for every item, I look at the system itself, not the description of it. Not "do you have lead scoring?" but "show me the last twenty leads and their scores." The gap between how a company describes its marketing operations and what the systems reveal is, reliably, where the money is.

Block 1: Data and definitions

• Lifecycle definitions exist in writing. Ask three people what a "qualified lead" is. Three matching answers is a pass. I have yet to see it on a first audit.
• The CRM is the system of record. If any revenue-relevant list lives in a spreadsheet that is more current than the CRM, fail.
• Attribution is captured at entry. Source, medium, campaign and landing page recorded on the contact at creation — not reconstructed later from analytics.
• Duplicate rate below ~5%. Export contacts, dedupe by email, compare counts. Above that, every downstream number is soft.
• Field hygiene. Count the custom fields; count the ones filled on more than half of records. The ratio tells you how much of the CRM is aspiration.

Block 2: Funnel mechanics

• Stage conversion is measurable. Can you produce lead→opportunity→closed rates for last quarter in under ten minutes? The time-to-answer is itself the finding.
• Speed to first touch. Median time from form submit to human contact. Under an hour is good; under a day is survivable. I have audited companies proud of their lead volume where the median was six days.
• Every form leads somewhere. Submit each public form yourself. Confirm the record appears, routing fires, and a notification reaches a person. On most first audits, at least one form leads nowhere.
• Closed-lost reasons exist and get read. If the data is captured but nobody has looked at it in a quarter, mark it half — the sensor works, the loop is open.

Block 3: Tooling and integration

• Tool census against invoices. List every marketing tool from the accounting system, not from memory. The delta is your zombie stack. Cancelling unused seats has funded a meaningful part of more than one of my engagements.
• Integrations are live, not nominal. "It's integrated with the CRM" means data flowed both directions this week — check the sync logs, not the settings page.
• One automation layer, known to more than one person. Workflows nobody can explain are liabilities with a login. I document each automation's trigger, action and owner as I go; the audit doubles as the missing documentation.
• Tracking survives a real visit. Open the site with a fresh profile, complete a conversion, verify the event lands in analytics and the CRM with attribution intact.

Block 4: Process and cadence

• A weekly metrics review exists — 30 minutes, from live systems, producing at least one decision. A monthly slide review does not count; the feedback loop is too slow to steer anything.
• Campaigns have owners and end dates. Scan the last quarter's campaigns for any still "running" that nobody can name an owner for.
• Sales-marketing SLA in force. Qualification definition, routing rules, follow-up windows — written, and violated less often than followed.
• Someone can ship without the founder. Trace the last landing page or email from idea to live: count the approval touches. More than two is a bottleneck wearing a quality-control costume.

Block 5: Spend and yield

• Cost per acquisition by channel, trailing 90 days. Not blended — by channel. Blended CPA is where bad channels hide behind good ones.
• Agency reporting reconciles with your systems. Compare the agency's reported conversions with what your CRM shows for the same period. Discrepancy over 15% means you are steering by someone else's instruments.
• Spend follows the funnel's actual constraint. If conversion collapses post-lead, more top-of-funnel budget is expensive noise. This finding alone justified reallocation in the DTC turnaround I described in the cost-per-sale case, where the fix was funnel reassembly, not more traffic.

Scoring and what to do with it

I score each item pass / half / fail, weight nothing, and resist composite scores — a single "68/100" hides exactly the information you need, which is where the system is weakest. The output that matters is a ranked fix list: each finding with its estimated revenue impact, effort, and a named owner. Ranked by impact-over-effort, the top five items become the next quarter's operations backlog. In my experience the first two fixes are almost always from Block 1 or Block 2 — definitions and funnel mechanics — because everything downstream inherits their defects. Tooling findings feel most urgent and matter least.

Two closing habits. First, rerun the audit quarterly; twenty items, one afternoon once the first pass exists, and drift is caught while it is cheap. The audit is a sensor, not a ceremony — I treat it as the instrumentation layer of the operations-first marketing framework. Second, automate the evidence collection where you can: the weekly numbers, the form tests, the sync checks. Reporting that assembles itself turns the audit from an event into a standing capability — which is the actual goal. An audited marketing operation is nice; a self-auditing one compounds.