Magnite's Hong Kong survey puts mobile streaming reach at 83% against CTV's 67%

Magnite published a study called "Streams & Screens: Hong Kong" on 27 August 2026 that puts mobile reach at 83% of the surveyed streaming audience against 67% for connected television, with respondents saying they will accept about seven minutes of advertising per hour of streaming. That device order is the reverse of the US edition Magnite released seven days earlier, on 20 August 2026, where connected television penetration for live streaming reached 92% against 82% for mobile.

Both numbers come from the same series, run with the same research partner, which is what makes the reversal worth a Thai planner's attention. It is also a survey published by a company that sells the advertising inventory it is measuring, and every person in the sample was already a streamer before they were asked a single question. Those two facts sit at the centre of this piece rather than at the bottom of it.

What Magnite published on 27 August 2026

Magnite released "Streams & Screens: Hong Kong" on 27 August 2026. The study surveyed 920 Hong Kong adults aged 18 and over who watch ad-supported streaming. Fieldwork ran in the first quarter of 2026 and was carried out with Bounce Insights, the research technology firm that also handled Magnite's US, Australian and New Zealand editions of the same series.

The shared partner matters more than it sounds. When a vendor runs the same instrument through the same fieldwork house in several markets, the cross-market comparison is at least internally consistent, because the question wording, the screening and the panel mechanics do not change between editions. A comparison between two studies from two different research houses would be far weaker. So the Hong Kong result can be read against the US result without first arguing about methodology, which is unusual for regional streaming data and is the single most useful property of this release.

Three headline figures come out of it. Mobile reaches 83% of the surveyed audience and connected television reaches 67%. Stated advertising tolerance is seven minutes per hour of streaming. And 20% of respondents said they have purchased from a vertical video ad.

Hong Kong reverses the device order Magnite reported in the US a week earlier

Magnite released the US edition of the series on 20 August 2026, a week before the Hong Kong one. In that US edition, connected television penetration for live streaming reached 92% against 82% for mobile. In Hong Kong, the ordering flips: mobile at 83%, connected television at 67%, a gap of sixteen percentage points running the other way.

The two sets of figures are not measuring an identical thing, and the table below says so rather than hiding it. The US number Magnite quoted is penetration for live streaming specifically. The Hong Kong numbers are reach among the surveyed ad-supported streaming audience. Reading them as a like-for-like race between two identical metrics would overstate the case. Reading them as two markets whose leading screen is a different screen is supported by what was published.

What moves between the two markets is almost entirely the television line. Mobile sits at 83% in Hong Kong and 82% in the US, which is close to a flat line across two very different economies. Connected television is the variable, at 92% in the US and 67% in Hong Kong. That is a useful way to hold the finding in your head: mobile streaming looks like a constant, and the big screen is what changes by market.

The daypart curve says the two screens own different hours

Magnite also published viewing by time of day, and the shape of that curve carries more planning value than the headline reach numbers. Mobile viewing runs above connected television through early morning, mid-morning and the afternoon. The two lines cross only in the evening block between 5pm and 9pm, where connected television peaks at roughly 65% against mobile's 50%. After 9pm, connected television keeps the lead.

So the honest summary is not that Hong Kong prefers phones. It is that the phone owns most of the clock and the television owns the evening, and the evening happens to be the block that traditional media planning treats as the day. A plan built on the assumption that the big screen carries the campaign will underdeliver against everything happening before 5pm, which by this curve is most of the viewing day. A plan that treats the phone as the whole answer will lose the block where the biggest single-hour concentration sits.

One point of care with these figures: the reach percentages and the daypart percentages are not the same measure. An 83% reach figure describes how much of the surveyed audience uses that screen at all. A 65% evening peak describes viewing within that block. They should not be subtracted from one another.

Seven minutes an hour is what people say, not what a test measured

The seven-minutes-per-hour figure is stated advertising tolerance. Respondents were asked what they would accept, and seven minutes is the answer they gave. Nobody in this study sat through an escalating ad load until they quit, which is what a behavioural measure of tolerance would look like. Stated tolerance and revealed tolerance are different quantities and they are routinely different numbers, usually in the direction of people saying they will accept less than they actually sit through, and sometimes the reverse when the content is compelling enough.

Converted to something a media plan can hold, seven minutes is 420 seconds, or fourteen thirty-second slots, or twenty-eight fifteen-second slots per hour of viewing. That arithmetic is worth doing because "seven minutes" sounds modest and "fourteen spots" does not. Magnite did not publish a break of that tolerance by device, by content type, or by whether respondents were told what format the advertising would take, so the number should be treated as a single blunt average rather than as a per-placement ceiling.

The 20% purchase figure from vertical video advertising is the other number worth carrying. One in five of the surveyed streamers said they have bought something after seeing a vertical video ad. Magnite did not publish what they bought, on which platform, or how long after seeing the ad, so this supports an argument that vertical creative can close a sale in this audience and supports nothing more specific than that.

Who the 920 respondents actually are

Sample composition decides how far a survey travels, and this one skews to the middle age brackets. The four age bands Magnite quoted are set out below, along with the arithmetic remainder.

Two thirds of the sample sits between 25 and 44. Respondents aged 55 and over make up 5%. That composition is not a flaw, because the study screened for people who watch ad-supported streaming and older audiences are less represented in that group in most markets, but it does mean the seven-minute tolerance figure and the 83% mobile reach figure are substantially statements about people in their late twenties, thirties and early forties. Anyone planning against an older audience is reading someone else's data.

The rest of the composition: the sample splits 52% female to 48% male, and geographically across New Territories at 35%, Hong Kong Island at 32%, Kowloon at 32% and Lantau Island at 1%.

Two limits that decide how much weight these numbers carry

The first limit is the screen on entry. Every respondent in this study already watches ad-supported streaming, because that was the condition for being surveyed. The 83% mobile figure therefore describes 83% of ad-supported streamers in Hong Kong, not 83% of Hong Kong adults. This is the single most common way a survey like this gets misquoted in a pitch deck, and the misquote inflates the number by an unknown amount, because the study says nothing about how large the streaming population is relative to the general public.

The practical consequence is specific. You cannot use these figures to model incremental reach or to argue what share of the country a streaming buy will touch, because the study does not measure penetration across the general population. You can use them to allocate a budget that has already been committed to streaming, because within that established audience the device split and the daypart curve are exactly what the study set out to measure.

The second limit is who paid for the research. Magnite sells the advertising inventory that these numbers describe, and PPC Land flags that conflict directly in its coverage rather than leaving readers to work it out. Vendor research is not automatically wrong, and treating it as automatically wrong would throw away most of the published data that exists on streaming audiences anywhere in Asia. It does need a different reading procedure.

That procedure is short. Check who ran the fieldwork, which here is Bounce Insights rather than the vendor itself. Check whether the sample was screened, which it was, and adjust every headline number accordingly. Then check whether the finding points in the direction of the seller's commercial interest. On that last test the Hong Kong edition does better than most vendor research, because it places connected television second in the market it surveyed, which is not the flattering answer for a company whose business includes selling that inventory. A survey engineered to produce a house-favourable result would not have published a 67% figure next to an 83% one.

What a planner can do with this on Monday morning

Treat the Hong Kong device split as a hypothesis to test against your own measurement rather than a number to copy into a plan. The check is straightforward in GA4 and your own reporting: pull device category against hour of day for the traffic your video campaigns actually deliver, and see whether the crossover in your data lands near the evening block or somewhere else entirely. If your own numbers say the phone leads until late evening, the Hong Kong curve is a reasonable regional cross-check. If they say something different, your own data wins and the survey is context.

On creative, the finding that pushes hardest is the daypart split rather than the reach split. Vertical creative built for a phone held in one hand is doing the heavy lifting through the morning and afternoon, and a sixteen-by-nine cut made for a television screen is competing for a narrower window than a standard plan assumes. That has a direct bearing on how video campaign assets get commissioned, since the cost difference between shooting for both crops at the same time and re-cutting later is considerable.

On measurement, the seven-minute tolerance figure is the wrong shape for an optimisation target and the right shape for a conversation with a client who wants to increase frequency. It is a stated average from a screened sample of 920 people in a different market. It is not a frequency cap.

What this means for Thai marketers

Magnite did not survey Thailand and published no Thai figure in this study. Everything in this section is reasoning about why a Hong Kong data point is worth reading in Bangkok, and it should be labelled that way in any deck it ends up in.

The reasoning is this. Thailand is a mobile-first streaming market, and published English-language streaming research covering Southeast Asia is thin enough that Thai planners routinely end up quoting US studies for want of anything closer. Hong Kong is a mobile-first Asian market with a study that used the same instrument as the American edition, which makes it the closest published benchmark available for the argument that the American device ordering does not travel. If a client asks why a Thai video plan should not be built around the big screen the way an American one is, the Hong Kong edition is a citable data point for the shape of the answer, without pretending it is a Thai number.

What it does not license is a specific figure. Nobody should write 83% into a Thai media plan on the strength of a Hong Kong survey. The transferable finding is directional, that in at least one mobile-first Asian market the phone leads on reach and holds most of the daypart curve, and the correct next step is to check whether your own Thai account data agrees. The same caution applies to the seven-minute tolerance figure and the 20% vertical video purchase figure, both of which describe Hong Kong respondents and nobody else. For campaigns running across social video platforms in Thailand, your own placement-level reporting is a better guide than any survey from another market.

Questions people are asking

Did Magnite survey Thailand?

No. The study covered 920 Hong Kong adults and published no Thailand figure. Magnite has run the same series in the US, Australia and New Zealand, all with Bounce Insights, and the coverage does not mention a Thai edition. Any Thai application of these numbers is inference, not a published finding.

Does 83% mobile reach mean 83% of people in Hong Kong watch streaming on a phone?

No, and this is the most important correction to make before the number goes into a deck. Every respondent already watched ad-supported streaming, so 83% describes the share of that streaming audience, not the share of the Hong Kong population. The study did not measure how big the streaming audience is relative to everyone else.

Can I run fourteen thirty-second ads an hour because tolerance is seven minutes?

No. Seven minutes per hour is a stated preference collected in a survey, not a tested tolerance threshold, and Magnite did not break it out by device, format or content type. It tells you roughly where the surveyed audience says the line is. It does not tell you what happens to completion or abandonment as you approach that line.

Why is Hong Kong the opposite of the United States on this?

The published data shows the difference without explaining its cause. Between the two editions, the mobile figure barely moves, at 83% in Hong Kong and 82% in the US, while the connected television figure moves a long way, at 67% against 92%. Magnite did not publish an explanation for the gap, so anything about housing, device ownership or broadcast history is speculation rather than a finding.

Should I trust a survey published by a company that sells the advertising?

Read it with the conflict in mind rather than discarding it. The fieldwork was run by Bounce Insights, the sample size and composition were disclosed, and the headline result puts connected television second in the surveyed market, which is not the most commercially convenient answer available. Those are the marks of research worth reading carefully. The screened sample remains the bigger limitation, not the sponsorship.

What should I check in my own account before acting on this?

Device category by hour of day for your video traffic, and whether vertical creative is being served where the phone dominates. If your own reporting shows a different crossover point from the Hong Kong curve, follow your own data. The survey is a regional reference, not a substitute for measurement you already own.

If you want a view of how device and daypart actually behave in your own Thai campaigns rather than in a Hong Kong survey, that starts with measurement that is set up to answer the question. Get in touch and we will look at what your own numbers say.