Measuring Marketing When the Click Disappears

The gap is widening, not stable

This isn't a new problem — dark social has been discussed for years — but the proportion is shifting. Platforms increasingly reward content that keeps users on the platform, so publishing natively outperforms linking out. Search increasingly answers in place. Both trends are commercially rational for the platforms and both reduce the share of your influence that your analytics can see. Expect the visible portion to keep shrinking.

Measure outcomes, not paths

When the path is invisible, stop trying to reconstruct it and measure what happened instead. Branded search volume tells you whether more people know your name. Direct traffic tells you whether people are seeking you out. Self-reported source on forms tells you what they'd credit. Win rate and cycle length tell you whether arriving prospects are better prepared. None trace a journey; together they tell you whether the work is landing.

• Branded search volume — the cleanest proxy for demand you've created
• Direct traffic and its conversion rate, tracked quarterly
• Self-reported source, free text, read monthly by both marketing and sales
• Sales cycle length and win rate — better-informed buyers close differently

Use holdouts, deliberately

The most reliable instrument available is also the least popular: turn something off and see what happens. Stop a channel in one region for six weeks. Pause a programme for a quarter. The results are unambiguous in a way no attribution model achieves. It requires accepting a period of deliberately reduced activity, which is why almost nobody does it, and why the teams that do have much better information than their competitors.

Report the uncertainty honestly

There's a temptation to present a precise-looking number because precision is reassuring to a board. Resist it. Say plainly which portion of your influence you can measure, which you can't, and what evidence you're using for the invisible part. In my experience this builds more credibility than a confident dashboard, because most experienced board members already suspect that the confident dashboard is wrong.