Measuring the Channels That Have No Click

Click-based measurement systematically undervalues everything that does not produce a click. Television, radio, podcast, outdoor, sponsorship, most video, and much of social do not generate a trackable action at the moment of exposure. Some responses are real but difficult to observe directly; this glossary entry on limbic resonance illustrates the measurement problem.

So they get little credit, budget moves toward the channels that harvest the demand, and total sales stop growing. This is one of the most predictable failures in marketing measurement, and it follows directly from the tooling rather than from anyone's judgement.

Why the problem is structural

Attribution records observed touchpoints. A television advertisement is not observed by your analytics. A search click is. So the search click gets the conversion.

The person searched your brand name because of the advertisement. Attribution cannot see that, and it never could. See what attribution actually measures.

The consequence is a ratchet. Each reallocation toward measurable channels looks like an improvement in the reports, and the reports get better while the business does not. Because demand creation was cut, the harvesting channels eventually have less to harvest, and the decline appears later and is attributed to something else.

What does not solve it

View-through attribution. It attempted to credit impressions without clicks, and it was always weak — it counted an impression served, not one seen or remembered. With third-party cookie restrictions it has largely stopped working anyway. See third-party cookies.

Post-exposure surveys alone. Asking people whether they saw the advertisement measures recall, which correlates with attention and not reliably with purchase.

Vanity reach numbers. Impressions delivered tells you what you bought, not what it did.

Correlating spend with sales without controls. Both rise together at Christmas.

What actually works

Three methods, in order of rigour.

Geo experiments

The strongest available for broadcast and outdoor. Spend in some regions, withhold in matched regions, compare sales.

This directly estimates the incremental effect, and it works for exactly the channels that clicks cannot measure. See geo experiments and holdout tests.

The main constraint is contamination — media crossing regional boundaries, national coverage, people travelling. It biases toward zero, so a null result may be a broken test rather than an ineffective channel.

Marketing mix modelling

Estimates contribution from aggregate spend and outcome data over time. Covers online and offline in one framework, and it does not need to observe anyone.

Correlational, so it depends on specification, and it needs several years of data with genuine variation in spend. See mix modelling.

Best used calibrated against a geo experiment on at least one channel.

Matched market and switchback designs

Where geography is impractical, alternating periods — on for two weeks, off for two, repeated — provides a within-market comparison. Works where the effect decays quickly enough that the off periods are genuinely off. It does not work for channels with long carryover, which is most brand advertising.

The intermediate signals worth tracking

None of these prove effect. All of them move earlier than sales and are worth watching alongside a proper test.

Branded search volume. The most useful single indicator for broadcast activity. If a television campaign works, people search the brand name. It is observable, it is free, and it responds within days.

Direct and organic traffic. Same logic, more noise.

Share of search — your branded search volume as a proportion of the category's. Corrects for category-wide seasonality, which raw volume does not.

Brand tracking surveys, if run consistently over time. Expensive, slow, and the only direct measure of what people think.

The honest framing for all of these: they are leading indicators, not outcomes. Branded search rising after a campaign is evidence the campaign was noticed. It is not evidence of incremental sales, and treating it as such reproduces the original problem with a different metric.

The branded search trap

Worth its own section, because it produces an expensive circular error.

A brand campaign increases branded search. Paid search captures those searches. Attribution credits paid search. The brand campaign gets nothing and the search campaign gets a rising number.

And most of those clicks were unnecessary. The user was looking for you specifically; the organic result was there. A holdout on branded search terms typically shows a substantial share of those conversions would have happened without paying.

Two tests worth running, in this order:

Hold out branded search in some regions and see how many of those conversions arrive anyway. This usually recovers budget immediately.

Then hold out the brand channel and watch what happens to branded search volume. That tells you what the brand spend is producing before any of it reaches a conversion.

Making the case internally

The measurement problem is also a political one, because the channel that cannot prove itself loses budget arguments to the channel that can.

Name the mechanism, not the outcome. "Click-based reporting cannot see this channel by construction, so its absence from the report is not evidence about its effect" is a stronger opening than defending the channel's performance.

Point at the ratchet. If budget has moved toward measurable channels over several years and total sales have not improved proportionally, that is evidence worth putting on a chart.

Propose the test rather than the argument. A geo test on the brand channel costs a defined amount and produces a number. This converts an unwinnable debate about measurement philosophy into a two-month plan.

Agree beforehand what result changes what. Including the case where the test shows a small effect — that is a real possible outcome and pretending otherwise makes the test look rigged.

The summary

Click-based reporting cannot see these channels, so their absence from attribution is not information about their effect.

Budget drifts toward demand capture and away from demand creation, and the reports improve while the business does not.

Geo experiments and mix modelling are what actually measure them. Branded search and share of search are useful leading indicators and are not outcomes.

Test branded search first. It is usually the fastest recovery of wasted budget available, and it demonstrates the whole argument with your own data. For industry context on comparable cross-media measurement, see the Nielsen cross-media measurement overview.