Two "Mobile Overtakes Desktop" Headlines, One Real Story: What comScore and StatCounter Actually Measured

Two headlines, two years, two different things

Search "when did mobile overtake desktop" and both the 2015 comScore milestone and the 2016 StatCounter milestone surface, often cited interchangeably as though they document the same event. They do not. They measure different populations, on different geographies, using different units, eighteen months apart. Reading them as one crossover is the single most common misreading of this piece of internet history.

March 2015: comScore's user-composition milestone

In March 2015, comScore reported that the number of US internet users who accessed the internet exclusively through mobile devices, mobile-only users, first exceeded the number of users who accessed it exclusively through desktop, desktop-only users. Mobile-only users stood at 10.8 percent of the measured population, against a desktop-only base that had been shrinking.

The key word here is "only." This metric does not count the large middle group of people who use both mobile and desktop, which was, and remained, the majority of internet users at the time. It measures a specific edge case: how the balance between two exclusive-access categories shifted. It says something real about mobile becoming a viable sole access point for a growing share of US users; it does not say mobile overtook desktop in total usage, total time, or total traffic.

October 2016: StatCounter's usage-share milestone

Eighteen months later, in October 2016, StatCounter reported a different milestone on a different metric: mobile-plus-tablet usage share, measured by page views across its global tracking network, exceeded desktop worldwide for the first time. This is a usage-share metric, not a user-composition metric, and it covers a global population, not the US-only population comScore had measured.

This is closer to what most people mean when they picture "mobile overtaking desktop" in a casual sense, actual traffic or usage tipping toward mobile devices. But it measures page views specifically, on a global sample, using StatCounter's own tracking-tag methodology, which is itself a specific and disclosed but not universal way of counting web traffic.

Why conflating the two misleads channel planning

The practical risk is straightforward. If a business or agency cites "mobile overtook desktop in 2015" while actually relying on the logic of the 2016 StatCounter usage-share finding, or vice versa, the underlying claim about audience behavior can be off by eighteen months and by an entirely different metric, on an entirely different population.

That gap matters when it is used to justify a specific decision, such as reallocating a media budget toward mobile-first creative, or deprioritizing a desktop-optimized experience. The comScore finding supports a narrower claim: a meaningful and growing share of US users now access the internet exclusively via mobile. The StatCounter finding supports a broader, later, global claim: mobile and tablet devices generate more page views worldwide than desktop. Neither, on its own, proves the other, and a decision built on the wrong one is built on a statistic that does not actually say what it is being used to say.

A case study in measurement-definition sensitivity

This pair of headlines is a useful, compact example of a discipline that matters well beyond this one topic: any "crossover" statistic needs its population, geography, unit of measurement, and source method stated before it can be used responsibly. The same caution applies elsewhere in this history. The American Time Use Survey measures only a respondent's stated primary activity in a 24-hour recall diary and therefore undercounts multi-device usage relative to metered panel data, and eMarketer's widely cited "time spent with media" figures are a modeled analyst estimate, not a raw panel or diary reading. In each case, two data sets can both be accurate and still not be directly comparable, because they were never measuring the same thing.

The comScore and StatCounter pair is simply the clearest, most citable version of this general problem, because the two milestones are close enough in time and topic to be mistaken for each other, while different enough in method to genuinely mislead if merged.

A short checklist before repricing a channel on a headline statistic

Before a "such-and-such overtook such-and-such" statistic is used to justify moving budget between channels, four questions are worth asking of it directly.

• What population does it measure? (US only, or global; all users, or only users exclusive to one channel)
• What is the unit? (page views, time spent, unique users, sessions)
• What is the source method? (a disclosed panel, a tracking-tag network, a modeled estimate, a government diary)
• What date, specifically, and does a related-sounding statistic from a different date and source actually confirm it, or just sound like it does?

What happened to mobile's share after the headlines faded

Regardless of which specific milestone gets cited, the underlying direction both were pointing at held up over the following decade. By 2024, mobile alone accounted for roughly 4 hours a day of US adults' media time, the single largest access point in the overall media mix, within a total digital-media share that had grown to 63.7 percent of all US adults' daily media time. The specific crossover date matters less, a decade on, than the fact that both 2015-era headlines were correctly reading the same underlying migration, even though they were reading two different instruments pointed at it.