For agencies · part of audit servicesWhat belongs in a tracking audit deliverable
Most audits are delivered as a list of problems. A list of problems is not a deliverable, it is homework.
The only test that matters is whether work happens afterwards. By that test most tracking audits fail, and they rarely fail because the findings were wrong. They fail because the document did not say what to do first, who should do it, or how long it would take, so it joined the pile of things that are important and never urgent. Here is what has to be in it for that not to happen.
The seven things it must contain
- 1
A verdict in one line
The person you hand it to has to repeat it to someone else, usually within a day and usually without the document open. A grade, a score, a single sentence: something they can say out loud. Without it they will summarise it themselves, and their summary will be either "it's fine" or "it's a disaster", neither of which is what you wrote.
- 2
Findings ranked by what they cost
Not by severity label, and certainly not by the order the interface presented them. A consent gate that blocks every tag outranks a naming convention, and the ranking should be obvious without reading the detail. If the first three items are not the three that matter, the rest will not be read.
- 3
Evidence attached to each finding
Name the tag, the event, the container, the number. "Duplicate conversion tracking" is an assertion. "These two tags both send purchase to the same measurement ID, and here they are" is a fact the client's developer can verify in a minute without trusting you. Audits get rejected far more often for being unverifiable than for being wrong.
- 4
A fix written for whoever has to do it
Most findings have exactly one correct audience, and it is rarely the person who commissioned the audit. Write the fix for the developer, the marketer or the agency who will carry it out, in their language, and say which one it is. A fix addressed to nobody in particular is addressed to nobody.
- 5
An effort estimate
Quick, medium or involved is enough. Without it, nothing can be scheduled, and work that cannot be scheduled does not happen. This is the single most commonly missing element, and the one that most reliably converts a read document into booked time.
- 6
What is working
Two reasons, and neither is politeness. It stops somebody rebuilding a thing that was already correct, which is a real risk when a client reacts to a bad grade by tearing everything out. And a document that only contains criticism reads as a sales pitch, which is exactly how it will be discounted.
- 7
What you did not check
Say which systems you had access to and which you did not, and what that leaves unknown. An audit that ran without the container connected cannot speak to consent, and saying so protects the client from acting on a gap and protects you from being blamed for it later. Nobody has ever lost a client by being clear about scope.
The shape that survives a busy reader
Assume the person opening this has fifteen minutes and three other fires. Three sections, in this order.
1. One page of summary
Verdict, the three things that matter, what they cost. This is the page that gets forwarded, and for many readers it is the only page. Write it last, and write it as though the rest will not be read, because often it will not.
2. The ranked findings
One block each: what it is, the evidence, the fix, who does it, how long it takes. Consistent shape throughout, so the reader learns it once and can then skim.
3. The appendix of everything checked
Every check, its status, what was found. Almost nobody reads it and it is still the most important section, because it is what makes the first two pages credible. It is the difference between an opinion and an audit. Put it at the back.
Deliver two formats, not one
A document and a spreadsheet, because they do different jobs and neither does both. The document is what gets read, approved and forwarded to whoever signs the work off. It should look like something the client would be comfortable putting in front of their own board, which for an agency means it carries your branding rather than a tool’s.
The spreadsheet is where the work lives afterwards. One row per finding, a status column, an owner column. It is the artefact that turns your audit into a project, and it is the reason somebody is still looking at your findings six weeks later. Deliver only the document and the findings have nowhere to go once work starts.
What to leave out
- ·Raw API output and screenshots of interfaces the client already has access to. It pads the page count and lowers the signal.
- ·Severity inflation. If everything is critical, the ranking carries no information and the reader falls back to guessing.
- ·Findings you cannot write a fix for. If you do not know what to do about it, it is an observation, and observations belong in the conversation rather than the document.
- ·Jargon that only survives inside analytics. Nobody outside the discipline knows what a consent gate or a lookback window is, and the people who approve the budget are outside the discipline.
The handover is part of the deliverable
Send the document ahead of a short call rather than presenting it cold. People do not argue with a document they have already read, they ask questions about it, and the questions are where the next piece of work usually comes from.
End the call by agreeing who owns the top three findings and when they will be looked at. If that conversation does not happen, the audit was an expense. If it does, the spreadsheet you delivered becomes a project plan and the audit pays for itself twice.
Tracking Auditor produces both formats from one run: a Google Doc with the verdict, the ranked findings and the fix plan, and a Google Sheet to track the work, both under your own branding. The appendix of every check performed comes with it. £79 an audit, no subscription.
Run a free audit →FAQs
What should a tracking audit deliverable include?
A one line verdict, findings ranked by what they cost rather than by severity label, evidence attached to each one so it can be verified independently, a fix written for whoever will carry it out, an effort estimate so it can be scheduled, a note of what is already working, and a clear statement of what was not checked and why.
How long should a tracking audit report be?
As long as the appendix needs to be, and one page for everything that matters. The summary is what gets read and forwarded; the detail exists to make the summary credible and to be referred back to during the fixes. Judging an audit by page count is how padded audits happen.
Should the deliverable be a document or a spreadsheet?
Both, because they do different jobs. The document is what gets read, approved and forwarded to whoever signs things off. The spreadsheet is what the work gets tracked in, with a status column that someone updates as fixes land. Delivering only the document means the findings have nowhere to live once work starts.
Why do clients not act on audits?
Rarely because they disagree. Usually because the document did not tell them what to do first, who should do it, or how long it would take, so it joined a list of things that are important and never urgent. Ranking by consequence and attaching an effort estimate to every finding is what turns a read document into scheduled work.
Should I include findings I cannot prove?
Not as findings. Anything you could not verify belongs in the scope note, stated as a limitation rather than a problem. The credibility of everything else depends on the reader being able to check any individual claim and find it holds.