The marketing activity keeps piling up: a website rebuilt two years ago, LinkedIn posts, a bit of Google Ads, a newsletter that goes out when someone has the time. Nobody really knows what brings in business. Before you relaunch a plan or pick a supplier, you need a marketing audit, in other words a review that lets you decide.
My projects almost always start there: a few days to understand the context, put numbers on what can be measured and propose a realistic plan. I have already described how I turn that into a marketing plan from six months to five years. Here I focus on the audit itself: the data I collect, what I check before trusting it, the people I talk to and the shape of the findings.
01What a marketing audit should produce
A classic marketing audit often ends with a thick document and a long list of recommendations, all valid, none ranked. The document goes in a drawer and nothing changes.
So I set the output before I start. At the end you should have three things: one page of findings, each with its source; a handful of decisions to take, ranked; and a list of what was looked at without reaching a conclusion. That last part matters. In my content machines, a check that did not run never gives a green light. An audit works the same way: a tool I could not get access to is marked "not verified", not "fine".
The audit does not rewrite the strategy. It tells you where you stand, precisely enough for the next discussion to be about facts rather than impressions.
02Collect the data, then check you can trust it
The first day is mostly about getting access: Search Console, Google Analytics, the ad account, the CRM or the plain spreadsheet where enquiries land. It always takes longer than expected, so I ask for access before the kick-off meeting.
Take as much history as you can
The Search Console performance report shows the past three months by default. That is too short to see a trend. I switch to the longest period available: Google's help notes that Search Console keeps sixteen months of data, no more. If nobody has ever exported it, anything older is gone. So I export everything on day one.
To read those figures, I follow Daniel Waisberg's advice in a Google Search Central video on analysing performance: compare with the same month of the previous year, and separate branded queries from the rest. The first step cancels out seasonality. The second shows whether the site attracts people who did not know you, or only those already typing your name.
Check what is actually measured
This is the step I most often see skipped. In Google Analytics, a key event is an action that matters to the business, and it is what lets you compare channels. Yet according to Google's help, only purchases are marked as key events by default on a website. On a B2B site with no shop, a contact request is therefore not counted until someone sets it up.
So I run a simple test: I fill in the form myself, then look for that enquiry in Analytics and in the CRM. If it shows up nowhere, every conversion figure in the audit becomes "not verified", and fixing the measurement moves to the top of the decisions.
I also look at the site as a visitor would: logged out, on a phone, in a private window. On one project, the site looked normal to the whole team, who were logged in, and was broken for visitors because of the cache. Since then I no longer trust what I see with a session open.
03Listen to the people who sell and the people who buy
Figures tell you what is happening, rarely why. So I talk to two groups: those who sell (sales staff, customer service, often the managing director) and a few recent customers, ideally one who bought and one who walked away.
I ask few questions, and the same ones to everyone. How did you hear about us? What made you hesitate? What did you compare? What did you ask before signing? I write the answers down as they were said, with who said them. In my content machines, the FAQ comes from questions people actually asked, each with where it came from; a question with no source is rejected. I apply the same rule here.
There is a limit to keep in mind. Nielsen Norman Group points out that interviews collect reported behaviour, not observed behaviour: memory fails, details go missing. A customer who says "I found you on Google" may well have seen you on LinkedIn first. That is why I always cross what people tell me with the data, and why I distrust a finding that rests on interviews alone.
04Review the channels, one at a time
For each channel I ask the same questions, in the same order, so they can be compared:
- How many enquiries or sales come from it, and how do we know?
- How much time and money goes into it each month, and who actually looks after it?
- What has changed in the past year?
- What, concretely, is blocking it?
This grid brings out things no dashboard shows. A channel can look weak when all it lacks is someone to run it. On one of my content projects, the problem was not the writing but the publishing, done by hand. Producing more would have made the bottleneck worse. The publishing had to be automated first.
For organic search, I also look at search volumes and keyword difficulty for the terms that matter to the business. I read them in Semrush, in my browser, rather than through its API: it is a rule I set for all my machines. If the site itself raises questions (slow pages, errors, a tool nobody knows how to update), I note it and point to a dedicated technical audit, which is a separate piece of work.
05Write findings that can be checked
A finding, for me, has three parts: the fact, its source and what it implies. For example: "Contact requests are not measured in Analytics (form test on the 3rd of the month). Today we cannot say which channel brings in customers." Anyone can repeat the test and check.
I hold myself to the rule I use for my articles: no figure without its source in the same paragraph. "Traffic is falling" with no period and no tool does not pass. "Non-branded clicks fell compared with the same quarter last year, according to Search Console" does.
Only then come the decisions. I keep a handful, and I put first the ones that unblock the others, such as fixing the measurement or finding who will run a channel. The rest goes on a "not now" list, with the reason.
Finally, I suggest a rhythm for following up those decisions. A monthly report that goes out on its own and a committee that decides are worth more than a dashboard nobody opens. I have described the automated report I use and how a monthly marketing committee runs.
An honest limit: a short audit does not replace market research, and it says nothing solid about a market you are not in yet. It is also of little use if nobody will have time to act on the decisions. And if you only want a first look, public support may be enough. In France, France Num explains that chambers of commerce and trade offer a digital diagnosis, often free depending on the region, covering online visibility, customer relationships and online sales.
06Where to start this week
You can do a good part of this work yourself. This week, list your channels on one page, with the person who looks after each one. Export the sixteen months of Search Console data. Fill in your own contact form and look for the enquiry in your tools. Ask two recent customers "what did you compare before choosing us?".
With those four steps you will already know whether your figures are reliable and where things get stuck. If you want to go further, the method for scoping a project in one week adapts well to an audit. And if you would rather do it together, my marketing strategy and project leadership services describe how I work, and the portfolio shows the rest. Happy to talk it through.