What a growth audit should contain, and what most leave out

An audit is the only part of an agency engagement where the incentives are clean. Nothing has been sold yet. So it is the part worth doing properly, and the part most agencies do worst.

Ask five agencies for a growth audit and you will get five sales decks. Each will have a slide of screenshots, a spreadsheet of keywords, a few red boxes drawn around your homepage, and a recommendation that happens to match the service the agency sells. That is not an audit. It is a proposal wearing an audit's clothes.

A growth audit has one job: to find where growth is actually constrained, with enough evidence that you could hand the document to a sceptical CFO and defend every line. Below is what that requires, in the order we do it.

1. Start from the outside, not from your analytics

Your analytics tell you what happened to people who already found you. They say nothing about the people who found a competitor instead. So the audit has to begin where the market begins: search results, AI answers, review sites, app stores, job boards, pricing pages, and your competitors' equivalents of all of those. Reading a brand outside-in is the discipline; it means every finding is one a prospect could reproduce.

2. Define the peer cluster before you score anything

A number without a comparison is an anecdote. A 2.1-second load time is fine in one category and a liability in another. Every metric in the audit should be scored against a named set of structural peers, chosen by business model and buyer, not by whoever the founder considers a rival. We resolve that cluster from public evidence, not from a list the founder writes down, so the comparison set is defensible.

3. Cover all seven dimensions, not just the one you sell

This is where most audits fail. An SEO agency audits search. A paid agency audits ad accounts. A growth constraint does not respect service lines. A real audit covers market position, operational health, strategic posture, audience intelligence, measurement integrity, business and financial health, and whether the brand learns from any of it. A brand with excellent search visibility and a broken checkout is not a search problem, and an audit that only looks at search will spend your money in the wrong place.

4. Tag every number with its evidence tier

Measured, real-user, sampled, modeled, inferred, observed, lab. A lab page-speed score and a real-user field score are different things and should never sit in the same sentence unlabelled. If a figure is modeled, the audit says so. Nothing erodes trust faster than a confident number that turns out to be an estimate.

5. Size the measurement gap

Before recommending spend, the audit should say how much of your current spend is unmeasured or misattributed. In our experience it is rarely under a fifth of the budget. The measurement gap is usually the cheapest thing to fix and the most valuable, because it makes everything after it provable.

6. Rank the plan by expected movement, with the evidence attached

The output is not a list of recommendations. It is a ranked 90-day plan where each line says what it should move, by roughly how much, against which baseline, and what evidence supports the expectation. Some lines will be uncertain, and the plan should say that too.

7. State how the audit could be wrong

Every audit rests on assumptions about the buyer, the category, and the competitive set. The good ones say what would change the conclusion. That is the section a board actually reads.

What it should not contain

• A recommendation that matches only what the agency sells. If the constraint is your pricing page, an honest audit says so even when there is no retainer in it.
• Unlabelled estimates. Every "industry benchmark" needs a source.
• Screenshots as evidence. A screenshot shows that something exists. It does not show that it matters.
• A timeline that starts on signature. The audit should already be dated, baselined, and re-runnable.

An audit built this way takes about two weeks and is worth keeping even if you never hire the agency that wrote it. That is the test: would you pay for the audit on its own? If the answer is no, you are reading a proposal.