From: Robert Harnaga
As you know, I was at the IAB meeting in Phoenix last week and wandered into a session on something the online folks call "viewability." Christ, what a shitstorm. Apparently in the world of Internet advertising, you can buy an ad and even if no one sees it, you still have to pay for it.
In Phoenix, on the one side you had advertisers telling Web publishers to prove that the ads they’re selling appear on users’ screens -- as opposed to parts of Web pages that people never actually see. On the other side, you have the head IAB guy (from CBS -- figures, right?) saying that if 70% of ads are viewable, everybody should relax and have a Coke. Meanwhile, you have these agency guys saying they are not happy unless all 100% of the ads they buy can be seen. Over in the corner, the ad-tech vendors are saying that 100% ain't possible.
It was GREAT. Almost made me forget about Brian for a few minutes.
On the other hand, it made me stop and ponder the notion of paying for ads that are not seen. You know and I know that since we've been packing in more and more ads into every break, fewer and fewer people are sticking around to watch them. Back in the old days, we gave viewers about 30 seconds to get to the bathroom and back. These days, not only can they get to the bathroom, they can bake a soufflé, walk the dog and pick up all the gloves on the floor of the mudroom -- and still have time to get back on the couch. Technically, these ads are not "viewable.” But thank God, we're getting paid for them.
You know I can't keep up with the C3 vs C7 nonsense. But if we can convince advertisers that people who record shows really don't fast-forward past the commercials, we might as well tell them that the Pope ain't Catholic (although this new guy's not acting like one).
I don't know what the big deal is anyway with that Internet crowd. When you listen to the car radio and an ad block comes on, you switch stations. Those ads aren't "viewed" either. You buy a billboard in Times Square. How many people even see it? Do you get to charge for everyone who walks past it in a month? What a racket that would be. The fact is that nobody sees every ad in any medium. But the Internet wove that sackcloth coat they're wearing with all that talk about accountability. Lol, as they say.
But I'll leave it to the online boys and girls to sort it all out. As long as no one leaves that baby at our doorstep, right?
Bob, the C3 stuff isn't nonsense. When a C3 rating is reported by Nielsen, it includes all set usage that was attained by a show's average commercial minute from the time it was aired "live" plus any usage that transpired up to the next three days, including, obviously, delayed exposures. The numbers are only for those sets that did not "zap" the commercials----"live" or "delayed". In other words, Nielsen is counting only those "audiences" that were 100% exposed, in the sense that these viewers, if they were so inclined, could have watched the ads in their entirety. Whether or not every member of the "audience" actually stays in the room or attentively watches the commercials is another matter. In "TV Dimensions 2015", which has just been published, we estimate that 50-60% of the reported commercial minute viewing"audience"sees all of part of an average "30", with variations by program type, time of day, amount of clutter, etc. This is understood by every advertiser who uses TV----you can't guarantee that every viewer who is confronted by a commercial is going to sit there and watch it. But, with TV, unlike digital, they could---if they wanted to.
"Whether or not every member of the "audience" actually stays in the room or attentively watches the commercials is another matter." Ladies and gentlemen of the jury, I rest my case.
George, if an advertiser pays for digital ad "impressions" where the "audience" never even had an opportunity to see the ad, that is not the same thing as a TV advertiser paying for viewers, all of whom had the option to watch the entire commercial---if they wanted to. I'm sorry----you have no case.
If a tree falls in the woods and no one hears it...Ed you are right there us a difference but my point is simply that all advertisers pay for ads that no one sees.
A "viewability" metric based on 50% of an ad in view for one second pretty much sums up the futility and foolishness of this ridiculous debate.
George, I understand very well what you are saying. You and others are suggesting that the huge problem facing digital advertising regarding viewability is really no big deal as it applies, one way or another, to all media. It doesn't. But rather than go back and forth on this, I'd like to see what some advertisers and ad agency types think about this. Am I wrong----- is viewability in digital advertising a bothersome issue but not really a problem, compared to the ad avoidance situation in TV, radio and print media? What say you, guys and gals?
This is a bit "inside baseball," but the way Nielsen calculates commercial minute viewing, it's possible for a viewer to see as little as one second of a TV commercial yet they would count it as seeing the whole ad.
Darrin, Nielsen has no way to determine whether a person who claimed to be watching a program sometime earlier actually watched any given second of the ensuing program or commercial content. So long as the set remains tuned to the same channel when the commercial airs, it is assumed that the earlier identified viewer "watched" the ad. As for the duration, I'd like to see an explanation from Nielsen. I assume that the set , including those DVRs tuned in on a delayed basis, must have a certain amount of exposure---say more than five seconds---- to qualify for an average minute rating, but I'm not certain about this.
Great piece !
So long as media measurement is based on OTS (Opportunity To See) and CPM (Cost per Maybe) we will continue to be embroiled in the story of the Emperor's clothes. The digital folks fell for it too!
It's the modern age, fricking 2015, time for a wake up call guys, our system is broken, it always has been. We are busy measuring the audience of content instead of counting what counts....the number of the right people engaging with our advertisers messages.
Albert Einstein said it best, not everything that can be counted counts....and not everything that counts can be counted.
When I turn on my TV and go away for the weekend did I have the opportunity to see the ads? Every media has are pros and cons. The folks who say one is perfect are selling that media. Anyone using a diary...well, do we even have to comment? No other media is as measurable as the digital world. Technology is changing the landscape faster than most people are comfortable with. Next year get ready for the technology that will cause even more debate. Remember "eyes don't lie"
Would anyone accept paying for ads on paper getting 90 blank pages and 10 pages with the ads - but paying for 100 pages. Well, I guess not. Why accepting this in the digital world? Fortunately, you are able to track this down to the individual media for your ad campaign. Getting the measurements and facts will put you in a much better position where you will be able to negotiate the CPM with the selected medias - or you can just select other medias with better performance. Instead of throwing your money in the black hole, then react on the measured facts before you have burned your budget. I have seen clients having ad display campaigns with as small as 3% of the display impressions in viewable area - this money burning for real. There are many reason for having these numbers. But it does not need to be so. It starts with the measurement facts on your campaign, before you can change. This is a very easy to do and you will raise your ROI significantly if you act.
Further inside baseball and commercial minute viewing: If a commercial begins at the 30th second of a "national commercial clock minute" and the viewer sees one second of that commercial and also saw the prior 30 seconds (whether it is program or commercial content) the whole minute counts as "viewed" under Nielsen's plurality of viewing rule. Aren't ya glad you didn't ask?
I just learned (by word of mouth) how to turn off Flash, On some sites, the page simply takes too much memory and takes too long to load with all the automatic-play videos and other stuff going on (all at once, so how does anyone imagine everyone sees it all?). Just another variation on saying you're "cutting through the clutter" when you're really only adding to it. So now I see nice restful black squares in their places. The only alternative is for me to stop visiting the site, and that wouldn't help viewing very much either, would it?
Ed, I am unsure of the US parameters, but here in Australia we use audio matching, so the TV set tuning I accurate - that is it doesn't rely on the meter 'clock time'. However, the people 'viewing statements' does. There is generally some drift in these, but the clocks are re-calibrated with every poll. When the data is processed and released it is at a minute-by-minute level. So the question then asked is which channel/programme 'won the minute'. We used to use a 'dominant channel' model, but now use the 'middle of the minute' model. Either way, both are vanities of precision. I think Darrin has explained the US definition above and it looks like 'middle of the minute'. I stress that this is to determine WHO is viewing not WHAT is being viewed. While 'middle of the minute' sounds awfully convenient, what is needed is the analysis to determine what the prior and following 30-seconds of viewing was. Does it correctly attribute (say) 50% of the people to the entirety of the as that was on in the middle-minute? Is it 60%, 70%, 80%, 90% etc.
There is also another missing element in this online 'viewability' discussion. Viewability is based on pixel counts and durations. You can argue all you like as to 50%, 70%, 100%, and 1-second, 2-seconds etc., but these parameters can be pretty easily changed.
These are parameters of an algorithm that aim to decide what proportion of pages/ads served would be physically possible for a person to view. But there is also the issue of VISIBILITY. I'm not talking about multiple invisible frames etc (i.e. fraud), but the simple fact that just because a 'viewable page/ad' was served does not necessarily make the page/ad visible. For example, there are ads being served on other MediaPost pages that I currently have open in my IE browser. They may be viewable, but they aren't visible. Heaven knows what has been served to my minimised Firefox while I have been reading the various MediaPost tabs. Can someone from the IAB definitively say whether their rules for viewability take into account the browser and browser/tab focus status to ensure that only visible and viewable ads are counted. Also, if there are any other interactive researchers out there who have data on the incidence of concurrent multiple browser usage, and average tabs open per browser session, that would help quantify the magnitude of the issue.
John and Darrin, I suppose that the "mid point" definition, if that's what Nielsen is using, isn't all that bad considering that set usage variations, like defections to other channels or new tune ins, can't be that large a factor in a given sixty seconds time frame----perhaps amounting to 2-3%. We keep referring to "viewing"----- which is quite a different matter. Nielsen assumes that viewing is taking place every second----unless otherwise indicated by the panel member---- because it has no other logical assumption to make; there is no practical way to determine if audiences are even in the room or their eyes are diverted during an average minute or portions thereof. So we are stuck with a system which is trying to provide more "granular" data than it is capable of to cater to the industry's incessant demand for more and more detail. In any event, I maintain----and will continue to do so--- that the current "measurement" of how many people "watch" TV commercials ----while subject to considerable inflation due to the methodology utilized-----is infinitely less suspect than digital's phantom ad audiences. As I have said, TV advertisers, who are buying what they think are ad exposures, are coming fairly close to the mark where opportunity to see is concerned. The pro-digital counter, made repeatedly by some, that nobody watches TV commercials simply doesn't hold water as a rebuttal.
Again we agree Ed. One thing to remember with PeopleMeters, is that if you see channel switching (and we do during ad breaks) it means that at least one person is correct in the system! The biggest concern is 'lazy viewing' i.e. logging in and forgetting to log out. This is partially controlled for by having edit rules that set maxima on what is considered not atypical viewing patterns - which is generally twofold based on maximum duration and that person's viewing day. For example, if we see one person watch 4 hours straight, and that is all they watch, and they do that without any button pushing you can be pretty sure that they forgot to log out so can safely be removed from that day's sample. Where it is trickier is within that ad-break pod. Around 10 years ago I analysed every commercial break in Sydney for a week or prime-time. I did this primarily at an aggregate level. What you could see was a lot of switching in breaks - especially between programmes. What surprised me was how ratings often went UP in a break. This was caused by people watching a programme that was over-running and checking whether the next programme had started. But even in breaks with virtually unchanged HH ratings (i.e. tuning) I could see changes in 'viewer' levels. By breaking this down into small demographic groups, and by calculating the average weight of a person in that demo, I could see that people were logging off and back in again during breaks (much to my surprise!). Of course, I could not see people who forgot completely. But I was pleased with the compliance. What I would have liked was the raw coded change-line data, and then I could have calculated the lot. Here in Australia we do a number of panellist checks including 'coincidental surveys' where we telephone the household a couple of times a year and ask that at the time of the call, which TVs were on, and to which channel and programme, and who was watching them. We can then check the change-line data when it arrives the following day. Every study I have seen has had compliance in the 90+% range. We also do 'outlier' analyses - the heaviest and lightest 10% of 'viewers' to try and pick up both 'lazy' viewers and household member churn, to ensure we reflect the population as closely and accurately as possible.
Interesting stuff, John. I do not have direct access to Nielsen peoplemeter data but I have seen the results of several old Nielsen coincidental "validation studies" of the sort you mentioned, which showed, an average discrepancy between what the phone call yielded and what the same people reported via their peoplemeters of plus or minus 10-15%. In other words, the implied "understatement" of viewing was of the same magnitude as the "overstatement". As far as I know, this type of research has not been conducted to differentiate between program content and commercials. In my opinion, such a study ----to be meaningful----would require some "proof" ----like content recall----not just the respondent's claims to have "watched" what was on the TV screen. I would also like to see any of these studies use a sample size that permitted an analysis of pluses and minuses by demo and by program type. If it is true that the average discrepancy is plus or minus 10-15%, imagine what this might look like for early AM news/talk formats versus primetime police dramas, or, in the case of commercials, for highly cluttered ad breaks versus normal ones.
The studies we do are "point in time". It is not about recall, programmes, content etc. We ask whoever answers the phone (I) which TVs are on (ii) which channel each set is tuned to, and what programme would that be, (iii) who was watching one each TV. The objective is to quantify compliance. Yes, we do get claimed tuning with no TV on , and vice versa, and they cancel each other out generally within a percentage point. Viewing compliance would be mid to high single digits. In essence we are checking whether the panellists are pushing buttons at the right time.
There are technologies which guarantee human views. Not just police the ads to find unviewable deployments-- tech that assures the ads only show in a context that is in view and seen by a human user. (My company makes one such tech, and there are other companies doing the same.) But this method can't apply to all ads yet; these are specialized ad units. Is this a way forward? Or is the problem too widespread for specialized ad units to make enough of a difference?
Matthew I am intrigued. In my past decade of trying to cajole the Australian interactive industry to move to people-based audience metrics I have seen many attempts that came up short in some aspect. At the moment we are focussing on producing audience estimates on the back of verified traffic data - however, we have not found a way to "guarantee human views". I have seen technology that allows that in test-bed laboratory conditions, but were undeployable in real-life due to privacy concerns and Australia's National Privacy Principles. Please contact via my MediaPost and enlighten me.
Most of you all are forgetting that TV measurement stats are projected based on algorithms/weights, and a lot of people bicker about on how to best accurately project the raw data. Display media visibility measurement for the most part (99.9%, I know only of a vendor that project data from one particular browser stats, i.e. this is not a large part of the data set), uses non projected number for their reporting. Why should one media source be held to a higher standard of scrutiny? We should be comparing apples to apples, yeah that will be the day, I know what some of you are thinking. Display media does not even receive the largest share of the advertising budget, but its scrutiny on visibility is un-proportionally intense. So if do establish that we cannot compare apples to apples and should not at least for near future until we can establish a consensus on a new method, what is going on here, are there any conspiracy theorist here? Is this just politics, agencies trying to look good in front of their clients, and/or setting boundaries on slippery slope? Whatever the case is, I can tell you this. You cannot have 100% visibility rate even if 100% of your banner pixels are above the fold, the inventory is non fraudulent, and it is 100% of the traffic comes from organic traffic (meaning not supplemented with any publisher media buying activities). There quite a few reasons why this can happen, most of it is technical reason, if you speak in details to all the visibility vendors out there, you can find out what these are. In the future you can probably buy based on visible impressions, when the industry have come to a uniform standard on visibility, and have the necessary tools to record and invoice this. But this is quite different from saying I want to have 100% visibility rate.
Benny, the fact that TV audiences are "estimated" based on projections from a survey, is not the point. While the TV advertiser does not know, specifically, what homes were tuned in when his commercial aired, they do feel confident in the general validity of the audience measurement. So, it's not really a question of comparing apples to oranges, when digital's CPMs are lined up against TV's. In both cases, the advertiser thinks he is paying for a given amount of audience that could be exposed to his ad. It doesn't matter that digital can identify the exact computer that received the ad message, though that might be an edge, in terms of targeting. As for the increasingly heard demand for 100% visibility, I understand that, technically, this may be impossible to deliver. Fine. But the digital ad seller should be able to tell an advertiser which computers displayed his ad message to the same extent as TV ( 100% ) and price accordingly. The reason so many digital ad sellers are upset is their current pricing system, in many cases, includes all of the bogus "exposures" as if they were worthwhile to the advertiser. Based on these vastly inflated audience tallies, a typical "untargeted" digital video ad's CPM is around $24, which is considerably higher than TV's all daypart/all network type norm ( roughly $10 ). If, say, 75% of the digital "audience is discounted due to zero or partial visibility, digital video's CPM jumps to the vicinity of $75 and this is unsupportable, even if better targeting can be attained. To bring such CPMs down to a more reasonable level, relative to TV, you have to grant a huge amount of "make goods"----or significantly reduce your charge per user. Neither of these is a desirable solution. Advertisers---branding advertisers, that is---are concerned about reach not just tonnage. They don't want to have their ad messages bombarding the same audiences repeatedly, just to fulfill a seller's CPM promises. And sellers, certainly aren't thrilled with the idea of slashing their CPMs, either. This is why the visibility issue is such a huge problem.
Again well said Ed. Though I have to quibble about "that digital can identify the exact computer that received the ad message". Majority of online traffic tracking relies on cookies. Cookies are typically deleted regularly (e.g. by Norton, Spy-Bot etc.) so after a relatively short period of time the 'device inflation' becomes dramatic. It wasn't until our monthly unique browsers got to 133m here in Australia - with a population of just 21m at the time - that the publishers realised how much of a problem cookie tracking was.
Now, Benny. There is no conspiracy here, merely a never ending quest to provide continually better metrics for ALL media that advertisers use. But it saddens me that you are in our industry and chose to denigrate probability sampling. So, I'd like to share an allegory with you, of the statistician who only accepted 100% surety and insisted on census complete enumeration in his work. Well, he was feeling rather unwell one day. Headaches. Sweating. Blurry vision. Nausea. Swollen limb extremities. So, he went to his doctor who was very puzzled. The doctor decided that the statistician needed a blood test. The doctor took a small vial of blood, labelled it and was preparing to send it off to pathology for testing. The statistician baulked. "What?!" he cried. "There can't be much more than 5ml of blood in there and I have 5 litres of blood in my body. That is only around one-thousandth, you won't get the right result with that!". The doctor complied. The statistician died.
Yes Ed, using projected stats or not the issue here, I am just saying that visibility report in the display media world is a more accurate measurement, assuming that everyone can agree a standard (MRC allows a few ways to measure at the moment thus creating variances in the stats from various vendors). The scrutiny on display media visibility are reaching insanity level, creating lots of polarization of views, and lots of preconceive notions. By the way most of the discussion involved are regarding banners and not video ads. Visibility measurements are used as pawns and being kicked around, the technology we use now is more quicker and more accurate than ever before and it will be even better.
Hi Benny. You are correct they are not your words. You wrote "Most of you all are forgetting that TV measurement stats are projected based on algorithms/weights, and a lot of people bicker about on how to best accurately project the raw data" ... which sounds a fair bit like a crack at probability sampling. I agree that there are inherent underlying assumptions in most media audience measurement, some acceptable, some not. No we did not sack the supplier who was providing the 133m monthly unique browsers. We sacked the metric and moved to monthly unique audience (and allowed average daily unique browsers as a slightly inflated proxy for daily audience). It took at least 2-3 years to get enough publishers to get on board, and many small publishers still refuse to. I also remain adamant that you can NOT provide "visibility" with server-side metrics that reflect the real world because the underlying assumption then becomes that every page (and let's be honest, we're really talking ads and not pages here) served is to a browser that is in focus and also to the tab that has the focus for that browser. The appear to accurately measure "viewability" (the capacity to be able to view the whole ad/page served) but they don't take into account "visibility" (the capacity to be seen by a human during its duration). To draw a TV parallel. A 30-second ad is served to 100% of the screen to 100% of the TVs that are tuned to that channel. If you know (or can estimate) the sets tuned you have a pretty good handle on "viewability". TVs inherent assumption is that the ad has "visibility" (by a human) tempered only by the number of panellists who either tune away from the ad, or who correctly push the PeopleMeter button if they leave the room, thus reducing the viewing estimate. The issue of 'doing something else/not paying full attention' is the same as on a PC - recall of digital display ads show that, as does the click-off rate of autoplay video. Matthew and I have been discussing his solution. Yes it works well, but then the publisher or advertiser have to deploy it. It is not universal, and it is not passive - but for advertisers you will get a read on the situation.
Just to be clear about CPMs, Benny, when I cite a $10 figure for TV this is not a low ball number based on a few buys in marginal time periods on marginal channels. Rather, it's an average of primetime, early and late fringe, daytime and early AM, across the broadcast network, syndication and cable spectrum and includes 15-second as well as 30-second spots. For a long time digital proponents have been comparing their CPMs with those of the broadcast TV networks in prime, which are the highest in TV by a wade margin, except for sports on the same networks. Since the majority of national TV branding dollars are not spent in prime buys on the broadcast TV networks, this is an unfair comparison, creating the illusion that digital, on average, is competitively priced or, in some cases, that its cheaper. Turning to digital CPMs, the $24 figure for video, again, is an approximate average for non-targeted ads while targeted CPMs come in around $35, typically, and some go for double or triple that rate. As for non-video display ads, a fairly good norm---with exceptions, of course----is in the vicinity of $12. Of course, you can get far cheaper CPMs---down to only $1, in some cases---but these are hardly representative and, yes, they are probably "crap". As I see it, unless the viewability issue is resolved with more than unacceptable "standards" set up by industry committees whose members may have self-serving agendas, branding advertisers will not divert substantial amounts of their TV budgets to digital ----with some exceptions, of course, such as youth targeted mobile buys. I don't know whether it is technically possible, but one solution----- at least for video----may be to create truly standard breaks, like TV, that the user becomes accustomed to and, finally accepts as the price for watching the video. These would provide 100% visibility for the ads, though, obviously, you want to minimize the number of ads per break. A standard break might have no more than 30-seconds of ad content, for example, and it would "freeze" on the user's page to provide 100% visibility. Naturally, there is no guarantee the the viewer would pay attention to the ad----but that is also true for TV, isn't it.
Only one company that I am aware off offers the following viewability standard:
- Guaranteed Viewability.
- Transparent, Actionable Data reported in Real Time.
- Maximize ROI lift.
Ad 100% imaged.
1 - 5 seconds in view
5 + seconds in view