How to Run a RevOps Audit (Without Hiring Anyone)

Most RevOps audits produce a slide deck that gets praised in the meeting and ignored afterward. The findings are true. They're also unattached to any decision anyone was about to make, which is the actual reason nothing happens.

A useful audit does something narrower. It answers seven questions with evidence pulled from your own systems, and it ends with a ranked list where each item names the person who has to decide. You can run one yourself in about a week. Here's how we do it.

Before you start: pick the window

Use the last four completed quarters, or the last two if the business changed shape recently. Write the window down and hold to it. The most common way an audit falls apart is that different findings quietly use different time ranges, and then someone in the meeting notices two numbers that can't both be true and the whole thing loses credibility.

Question 1: Where does a deal actually come from?

Pull every Closed Won opportunity in the window with its source field. Then pull the same set and count how many have that field blank or set to a default like "Other" or "Web."

The second number is the finding. We routinely see companies where 30 to 60 percent of closed revenue has no usable source attribution, which means every marketing spend conversation for the past year has been an argument about a coin flip.

Don't try to fix attribution in the audit. Just measure how much of it is missing, because that number decides whether the next twelve months of marketing reporting is worth building.

Question 2: Do stages mean anything?

For each pipeline stage, calculate the median days a deal sits there and the conversion rate to the next stage. Then look for two specific patterns.

A stage with near-100% conversion isn't a stage. It's a formality someone clicks through on the way to the next one, and it inflates your funnel math by adding a step that never filters anything.

A stage with a huge spread in duration, where the median is 8 days and the top quartile is 60, usually means two different sales motions are sharing one pipeline. That's worth knowing before anyone builds a forecast model on top of it.

Then do the qualitative half. Ask three reps, separately, to define stage three in one sentence. If you get three sentences, you don't have a process problem in the CRM. You have an agreement problem among humans, and no configuration change fixes it.

Question 3: How many places is the same fact written down?

Pick three facts: contract value, renewal date, and primary contact. For each one, list every system that stores it. CRM, billing, the CS platform, that spreadsheet finance maintains, the PDF in Drive.

Then sample twenty accounts and check whether the copies agree.

This is the single highest-yield hour in the whole audit and almost nobody does it. When we run it, the disagreement rate is rarely under 10 percent, and the interesting part is never the error rate. It's that in most companies no rule exists for which copy wins, so both departments are correct and both are working from a different number.

Question 4: What breaks silently?

Open your automation platform, your CRM's workflow logs, and your integration tools. For each active automation, answer two things: when did it last run, and what happens when it fails.

The ones that alert a human on failure are fine. The ones that log an error nobody reads are the ones that will cost you, because a workflow that has been silently failing for five weeks has been producing a specific, wrong outcome five weeks running, and the damage is already downstream. We wrote up the patterns that prevent this separately; for the audit, you just need the count.

Also look for automations that fire on the same trigger. Two workflows both listening for Closed Won, built eight months apart by different people, is one of the most common findings we log, and the second one usually does something the first one already did.

Question 5: Where does a human retype something?

Sit with three people for twenty minutes each: someone in sales ops, someone in finance, someone in delivery or CS. Ask them to walk through their week and stop them every time they read a value off one screen and type it into another.

You're not looking for the biggest task. You're looking for the most frequent one, because frequency is what makes automation pay and because retyping is where the disagreements in question 3 are manufactured.

Write down the volume, not just the task. "Invoices created by hand" is not a finding. "About 45 invoices a month, roughly four minutes each, and the amount comes from a field two people can edit" is a finding, and it prices itself.

Question 6: Can you answer the forecast question from a system?

Ask for pipeline coverage for the current quarter and forecast accuracy for the last four. Time how long it takes to get an answer, and count how many people had to be involved.

If the answer requires an export and a spreadsheet, you have a reporting gap. If the answer arrives fast but nobody in leadership believes it, you have a definitions gap, and those are fixed in very different places. The dashboard architecture handles the first. Only a decision from a founder or CRO handles the second.

Question 7: Who owns the definitions?

Name the person who can say what "qualified" means and have it hold across marketing, sales, and CS. Not the person who administers the CRM. The person whose ruling ends the argument.

If no such person exists, put that at the top of your findings, because it caps the value of everything else on the list. This is the failure mode described at length in our RevOps primer, and we keep repeating it because it's the one that quietly wastes the most money: a company hires or contracts for RevOps, gets a competent operator, and gives them no authority to settle anything.

The deliverable

One page. Not a deck.

For each finding, four lines: what you observed, the number that supports it, what it costs in hours or risk, and who has to decide. That last line is what makes an audit act like a decision document rather than an assessment.

Then rank by a simple rule. Anything that produces a wrong number that people act on goes above anything that merely wastes time, because wasted time is expensive and wrong numbers are dangerous. Within each group, sort by frequency.

Where a self-run audit falls down

Two places, honestly.

The first is question 3. Checking whether copies of a fact agree means going into systems you may not have access to, and asking finance to justify a spreadsheet is a political act in most companies, not a technical one. If you can't get to the billing data, say so in the findings rather than skipping the question.

The second is that you're inside the org. Question 7 asks who owns the definitions, and if the honest answer implicates the person you report to, a self-run audit will find a gentler way to say it. That's not a character flaw. It's structural, and it's most of why outside audits get commissioned at all.

If either of those is your situation, an outside pass runs three to six weeks and comes back with the same seven questions answered by someone with nothing at stake in the answers. And if you run it yourself and get stuck on one finding, send us that one. A single question is a fine reason to book half an hour.