How Do You Measure Marketing When the Click Disappears?

A marketing director in Kuala Lumpur recently described a problem that is becoming common. Organic traffic to her company’s site is down about a fifth year on year. Enquiries are flat. Deal quality has, if anything, improved. Nobody can explain any of it, and the board wants to know why the content budget is not producing traffic growth.

The explanation is that a growing share of the work marketing does now happens somewhere she cannot see.

Where the click went

Three things have changed at once, and they compound.

Search results answer more questions directly. A user asking how long a website takes to build gets an answer at the top of the page. They never visit the article that supplied it, but they did read it, and the company that wrote it shaped their expectation.

AI assistants have become a research step. Buyers ask an assistant to explain a category, compare approaches, or suggest vendors before they open a search engine. That conversation is entirely invisible to analytics. If your brand was mentioned in it, you will never know from your dashboard.

Referrals arrive stripped. Traffic from AI assistants frequently shows up as direct, because the referrer is lost. A large direct-traffic segment that nobody can explain is now a common finding, and a portion of it is people who were sent by a machine.

The net effect is that the click was never the value. It was the measurement. The value was influence on a buying decision, and influence is still happening — it has just stopped generating a row in the analytics table.

Stop defending the wrong number

The first thing to change is internal, not technical. If marketing is reporting sessions as its headline metric, it is going to keep losing an argument it should not be having.

Sessions were always a proxy. They were a good proxy for a long time because clicks and influence moved together. They have decoupled. Continuing to report the proxy when it has stopped tracking the thing it proxied for is how good programmes get cut.

The replacement is not one number. It is three.

What to measure instead

Impressions and position, not clicks. Search Console still reports how often you appeared and where. If impressions are stable and clicks are falling, you have not lost visibility — the result page changed. That distinction is the single most useful thing to put in front of a board, because it separates “our content stopped working” from “the click-through environment changed”, which require completely different responses.

Presence in AI answers. This is measurable, just not automatically. Build a list of thirty to fifty questions a buyer would actually ask, run them across the main assistants on a fixed schedule, and record whether your brand appears and in what terms. It is tedious the first time and largely automatable after that. Tracked over months it produces a trend line, and a trend line is what a board responds to. Systematic AI search visibility tracking of this kind is how the influence that no longer shows up as traffic becomes visible again.

Self-reported attribution. Add one optional field to enquiry forms: “How did you hear about us?” Free text. It is imperfect, biased, and incomplete, and it will still be the most informative data you collect this year — because it is the only source that captures a conversation with an assistant, a recommendation from a colleague, or a podcast mention.

Brand search is the leading indicator

One number deserves separate attention: searches for your company name.

Branded search volume is what happens after someone encounters you somewhere you could not track. They heard the name, they remembered it, they looked you up. It is the cleanest available signal that unmeasurable channels are working.

If branded search is rising while non-brand traffic falls, the programme is working and the measurement is broken. If both are falling, the programme has a problem. Most companies do not separate the two in their reporting, which means they cannot tell these two very different situations apart.

Separating branded from non-brand in Search Console takes about ten minutes and changes what the data says more than almost any other reporting change available.

What this means for content

If a meaningful share of your audience now reads your content inside someone else’s interface, the content has a different job.

It needs to be quotable in isolation — a paragraph that makes sense lifted out of the page, because that is how it will be encountered. It needs the brand name in it, not just in the byline, because a summary drops the byline. And it needs to say something specific enough to be worth quoting; generic content is not selected, and being unselected is the new version of ranking on page two.

That is a change in craft rather than a change in volume. Publishing more of what was not being quoted does not help.

Reporting it without losing the room

The presentation matters as much as the measurement, because the people receiving the report have spent a decade being told that traffic is the number.

The structure that works is three lines. First, visibility: impressions and average position, stated plainly, with the note that this is how often we appeared. Second, capture: clicks and click-through rate, framed as the share of that visibility we converted into a visit. Third, outcome: enquiries, branded search, and whatever self-reported attribution has come in.

Presenting them in that order makes the mechanism legible. When clicks fall while impressions hold, the chart itself explains what happened and no one has to take marketing’s word for it.

It also changes what the conversation is about. A board looking at a single traffic line asks why it is down. A board looking at three lines asks which of them we should be working on, which is a considerably more productive meeting.

One practical caution: set the baseline before you change the reporting, not after. Switching metrics in the same quarter that the old metric fell looks like moving the goalposts, whatever the merits. Run both for a period, show that they tell the same story historically, and then retire the old one.

The honest position

Attribution was never as solid as the dashboards implied. Last-click was always a convenient fiction that gave credit to whichever channel happened to be standing nearest the finish line.

What has changed is that the fiction has become too obviously fictional to keep using. That is uncomfortable, and it is also an opportunity to start measuring something closer to what actually matters: whether the market knows who you are, whether you are present when buyers are deciding, and whether the people who arrive are the right ones.

Those questions were always the real ones. It is just that nobody could avoid them any more.

This article was contributed by the team at Trinergy Digital, a software and creative technology company based in Kuala Lumpur, working with clients across Malaysia, Singapore and Australia.