The three questions before every ad account audit

Question 1. What is the account trying to do?

The single objective the account is optimised for. Not the marketing team's quarterly targets, not the CMO's slide, not the board's north-star metric. The account's operating instructions.

The right answer is boring: "acquire qualified leads under ₹1,200 CAC," or "drive first purchases at 4.2x ROAS," or "book demos with prospects above ₹5 lakh ARR potential." One sentence, one number, one boundary. If the answer requires reading between the lines of an internal deck, the account is unclear before the audit begins.

An account without a single operating objective develops a common pathology. Different campaigns optimise for different things: some for conversions, some for clicks, some for a legacy target set in a previous quarter and never updated. Media budget flows toward whichever campaign the automation deems "successful" this week, which often is not the campaign that supports the business. On paper the account looks busy. In reality it is drifting.

Ask the question the way a new head of marketing would ask it. If the account owner needs more than one sentence, or hedges, or names three things, note it. That is the audit's first finding.

Question 2. What is the evidence it is doing that?

The measurement architecture. Which conversions are counted, which windows, which attribution model, which primary metric on the reporting layer. What is imported from the CRM, on what cadence, joined on what identifier.

This is not a place to be diplomatic. Either the evidence chain from click to counted outcome is intact, or it isn't. Common breakdowns worth naming plainly: • The account's stated target is CAC on qualified leads, but the counted conversion is "all form submissions," including newsletter signups and support tickets. • Attribution is set to last-click inside the account, but the reporting deck cites Data-Driven numbers copied from a different environment. • View-through conversions are on by default and quietly inflate the numbers no one questions. • Offline conversion imports run weekly, but bidding operates on the seven-day click window; the loop closes after the automation has already made its decisions. • Conversion tags fire twice on the thank-you page because the redesign in April kept the old tag alongside the new one.

Each of these is invisible at the reporting layer and produces a systematically wrong picture of what the account is doing. An account with a broken measurement chain cannot be audited for performance; it has to be audited for measurement first. The order matters.

Question 3. What would we change first?

The account owner's honest answer, before the audit begins. If the operator cannot name their top-priority change without the audit, the audit has an outsized role to play, and probably the wrong one. It is now expected to surface an unknown, which is a hard thing to ask of any audit and an easy thing to get wrong.

If the operator can name it, the audit's job is different and much better defined. It becomes: validate this instinct, or contest it, in that order. Most of the time the audit ends up validating. The operator is closer to the account than anyone else and has usually already sensed where the leverage is. The value of the audit is confirmation, sequencing, and one or two adjacent findings the operator was not close enough to see.

The other value of asking this question up front is that it flushes out the account's political geography. When the operator names their top-priority change and it is different from the change the CMO wants named, the audit is now walking into a decision the audit was not authorised to make. Better to know that before writing the report than after.

What the three questions change

The three questions turn an audit from a list of things wrong into a prioritised set of decisions with an owner. That is the whole difference. A long list of findings without an owner is a document; a shorter list of decisions with an owner is a plan.

For a marketing lead handed a mid-size Google Ads account, the questions land as a standing pre-audit checklist and the front end of the Google Ads audit framework I use across every engagement. Send them to the account manager three working days before the audit call. Ask for one-sentence answers. Read the answers before opening the interface. If any answer is missing or hedged, that is the finding, and the audit either changes shape or is deferred until the underlying decision is made.

Across the ad accounts refined through MagicWorks IT Solutions Pvt. Ltd., the AI-first digital marketing agency I founded in 2009, the pattern has held. Accounts where all three answers arrive cleanly are audited in a day; the audit becomes a decision brief. Accounts where the answers arrive hedged or missing produce a longer audit, but that audit's first section is now about the operating decisions that need to be made before the media questions can be addressed sensibly.

What the AI-first tools do here

Automated audits sharpen this rather than replace it. The current generation of tools can produce more findings in an hour than any team can act on in a quarter. That volume is a feature only if the findings can be filtered by relevance to a stated objective. Without the three questions answered, "relevance" has no definition, and the tool's output becomes noise dressed in charts.

With the three questions answered, the automation earns its place. It flags the anomalies that touch the stated objective, sequences them by likely impact, and hands the shortlist to a human who can now audit with the confidence that every item on the list connects to something the account is actually trying to do.

The full audit framework, including the four-phase structure that follows once the three questions are answered, is Chapter 3 of The AI-Powered Google Ads System, co-authored with Mohan Chute and published September 2026.