Marketing attribution is breaking, and pretending otherwise is costing businesses real money.
Not because the tools got worse. Because the buying journey moved into places no tool can see. Private conversations with AI assistants. Group chats. Direct messages. Podcasts. A recommendation from a mate at a barbecue.
The result is a dashboard that confidently attributes a sale to the last Google click, when the actual decision was made three weeks earlier in a chat window.
When discovery goes invisible, you need a scoreboard that measures the shadow rather than the object.
There is one, it is free, and you already have access to it. It is branded search: the number of people typing your business name into Google. When you can no longer see where people hear about you, counting how many of them come looking for you by name is the most honest measure of whether your marketing is working.
Branded search is the honest number
Every invisible recommendation ends the same way. Someone hears about you somewhere you cannot see, and then types your business name into a search box.
That is branded search, and it is the closest thing marketing has to a truth serum right now.
It cannot be gamed by a channel taking credit for work it did not do. It rises when your reputation grows and falls when it stalls. And in our view it is one of the most useful leading indicators available to a small business.
Open Search Console. Filter queries containing your business name and its common misspellings. Chart impressions and clicks monthly.
That line is your brand health, measured weekly, for free, by the largest measurement system ever built.
Traffic tells you how many people you reached. Branded search tells you how many people decided you were worth remembering. Only one of those predicts next quarter.
How to build the branded search chart
This takes about twenty minutes and you only do it once. If you have never opened Search Console, it is Google's free tool for site owners, and it records the actual queries people used to find you.
- Open the Performance report. Set the date range to the last sixteen months, the longest it holds (Google's own guidance points to the "Last 16 months" view), so you can compare the same months year on year.
- Add a query filter. Use "Queries containing" and enter the core of your business name. If your name is two words, try each word on its own as well as together.
- Catch the misspellings. Look through the full query list for garbled versions of your name. People mistype, drop spaces and spell phonetically. Note each one.
- Use a custom regex filter for the full set. Search Console accepts a custom regular expression filter, so one filter can capture your name and every misspelling at once.
- Export monthly figures. Paste clicks and impressions into a simple sheet, one row per month. That sheet becomes your scoreboard.
Watch out for branded queries that are not really about you. If your name is also a common word or shares it with a well known brand, you will need to exclude those queries or the line will be noise.
The four number scoreboard
Branded search is the headline. Three more numbers complete the picture, and together they replace an attribution model you can no longer trust.
One. Branded search volume, monthly. The demand you have created. Watch the trend over quarters, not weeks. It moves slowly and it does not lie.
Two. Direct traffic landing on deep pages. Nobody types a service page URL from memory. Direct visitors arriving on anything other than your homepage were often sent by something invisible, such as an assistant or a private share. Segment it and watch it grow. Some AI referrals do pass a source you can see, and traffic from AI tends to convert well, which is another reason to stop lumping it in with everything else.
Three. Revenue per session. The metric that survives the collapse of traffic volume. If sessions fall while revenue per session climbs, you are winning and every volume chart in the building is lying to you. For a lead based business, swap revenue for enquiries or booked jobs per session. The logic is the same.
Four. Self reported attribution. One open text field on your enquiry form. "How did you hear about us." Not a dropdown, which only returns the options you already imagined. An open field returns sentences like "ChatGPT recommended you twice" and "my builder said you did his mother's place."
That field is the single most valuable analytics upgrade available to a small business, and it costs ten minutes.
A worked example: the physio who almost cut the wrong thing
Say you run a physiotherapy clinic in Perth. Over the last year you have posted short videos twice a week, asked every patient for a review, and spent a small amount on Google Ads. Your website sessions are down. The Google Ads report says it drove most of your new bookings. The obvious move is to cut the video and put the money into ads.
Now look at the scoreboard. Branded searches for the clinic name have climbed steadily all year. Direct visits to your sports injury page have grown. Bookings per session are up even though sessions are down. And when you read back through the "how did you hear about us" answers, patients keep writing things like "saw your videos" and "my teammate sent me your page."
Here is what actually happened. The videos and reviews made people remember your name. Many of them then searched for you, clicked your own branded ad because it sat at the top, and the ad took the credit. Cut the video and the branded searches dry up a few months later, and the ads have nothing left to harvest.
The four numbers do not tell you everything. They tell you enough to avoid the most expensive mistake in small business marketing: cutting the thing that creates demand because a different thing sits closer to the sale.
What to stop measuring
Some of what is on your current report is actively misleading and deserves to be deleted. We made the longer case in vanity metrics are quietly bankrupting your marketing, but here is the short list.
- Last click attribution as a decision tool. Fine as a record. Disastrous as a budget allocator, because it systematically overpays whatever channel sits closest to the sale and starves everything that created the demand.
- Total sessions as a headline. It is under structural pressure for many sites as more answers get delivered without a click. Judging your work on it will lead you to cut things that are working.
- Follower counts. In our experience they are a poor guide to revenue.
- Impressions and reach on their own. Useful as a diagnostic, meaningless as a result.
The test that settles arguments
When attribution fails, there is one method that still works and almost nobody uses it. Turn something off.
Pick a channel. Pause it for four weeks. Watch branded search, enquiry volume and revenue. Then turn it back on and watch again.
It is blunt, it is a little scary, and it produces a genuine causal answer that no attribution model can give you. Businesses that run one of these tests a quarter learn more about what actually works than businesses that spend the same time arguing about the dashboard.
A few rules keep the test honest. Change one thing at a time, or you will not know which change caused the result. Avoid running it across a holiday or your seasonal peak, when demand moves on its own. Write down what you expect to happen before you start, so you cannot rewrite the story afterwards. And start with a channel you genuinely doubt, not the one you would be terrified to lose.
The conversation this fixes
Every business has the same argument each year. Somebody asks what the content, the reviews, the video and the reputation work are actually returning, and the honest answer has historically been a shrug and a story.
With this scoreboard you can say something better. Branded search is up 40% over twelve months. Direct traffic to deep pages has tripled. Revenue per session has doubled while sessions fell. And one in five enquiries now mentions an AI assistant or a referral by name.
That is not a story. That is a case, and it protects the budget that is doing the actual work.
The trap of measuring what is easy
There is a reason businesses cling to bad metrics. They update daily, they produce a chart, and they feel like control.
Branded search moves over quarters. Revenue per session needs context. Self reported attribution arrives as messy sentences rather than a tidy percentage. None of it is satisfying to look at on a Monday morning.
But the comfortable numbers are describing a world that no longer exists, and comfort is expensive when it leads you to defund the work that is actually creating demand. Plenty of businesses that cut their reputation spending because the traffic chart looked flat end up paying to rebuild it.
Pick the honest numbers even though they are slower. Slow and true beats fast and wrong every single quarter. And if you want a sense of where that invisible discovery is happening, ask the AI assistants what they say about you. It takes five minutes and it is usually revealing.
Set it up this week
Add the open text field to your enquiry form. Build the branded search chart in Search Console. Build the direct to deep page segment in your analytics. Add revenue per session to your monthly review.
Then read the self reported answers once a month and tag each one by source: AI assistant, referral, social, search, other. Within a quarter you will have a picture of where your customers actually come from that no dashboard could give you.
An hour, once. Then stop looking at the traffic chart and start looking at the four numbers that still tell the truth. If you would like a reporting setup built around these numbers instead of vanity charts, that is part of how we run digital marketing at LIQID NOISE: measure what is true, then spend where it points.
Frequently asked questions
What is branded search and why does it matter?
Branded search is any search that includes your business name, including misspellings. It matters because it captures demand you created in places you cannot track, like AI chats, group messages and word of mouth. When someone hears about you anywhere, they usually end up searching your name, so a rising branded search trend is strong evidence your reputation is growing.
How do I track branded search for free?
Use Google Search Console. In the Performance report, filter queries containing your business name, add a regex filter to include common misspellings, and set the widest date range available. Export clicks and impressions monthly into a sheet. Review the trend each quarter rather than week to week, because branded demand moves slowly.
Should I stop using last click attribution?
Keep it as a record of what happened, but stop using it to decide budgets. Last click hands all the credit to whichever channel sat closest to the sale, often a branded search ad, and ignores whatever created the demand. Pair it with branded search trends, revenue per session and an open "how did you hear about us" field.
How do I know if a marketing channel is actually working?
Run a holdout test. Pause one channel for four weeks, outside your seasonal peaks, and watch branded search, enquiry volume and revenue. Then switch it back on and watch again. It is blunt, but it gives you a genuine cause and effect answer that no attribution report can.