Commentary
How Marketers Can Save Millions Of Dollars In CTV: 3 Lessons From Paid Search, Programmatic
- by Jason Fairchild , Op-Ed Contributor, July 21, 2021
We’re in the midst of yet another massive change in the digital advertising industry and I’m not talking about cookies, iOS14 or ad targeting. I’m talking about the massive ad budget shift to connected TV (CTV) advertising we’re all witnessing.
Over the past two decades there have been three significant digital advertising methods that have reshaped the advertising industry: pay-per-click, real-time bidding, and CTV advertising. But unlike the first two events, CTV advertising will develop much more rapidly than search or display as they had to be proven.
Advertising on TV is a format widely accepted by advertisers (and has been for 80 years) and as TV shifted from a box to a smart device, and content changed from linear to streaming/on demand the audience has grown exponentially.
Currently, 63% of all TV viewing is already done on streaming services, which equates to 90 million+ ad-supported CTV viewing households.
For the last 25 years, I've worked for and co-founded startups in digital and marketplace platforms including paid search, programmatic, and now CTV. As marketers of all sizes look toward the brave world of CTV advertising, there are lessons they can take with them from their experiences in digital advertising -- particularly search. Here are some lessons I've learned along the way that can save any marketer millions of dollars.
Follow the money
Understand exactly where your money goes and what percentage of your CTV budget goes to working media.
Ask for and only work with 100% transparency. There are lots of leading CTV sales houses claiming to be "platforms" that collect 30%-50% margins in non-transparent ways.
The easiest way to solve this is to have them disclose the actual costs of media and/or data, along with their take rate. If they won't disclose, you have your answer. If they disclose and have been "taking some liberties" with your money, you also have your answer and need to find a new, more reputable partner.
Warning: if the answer is too complicated, it's a smokescreen. The truth here is radically simple, so you should be able to “trust but verify” by drawing out the supply path with a clearly defined percentage of revenue for each link in the value chain.
Embrace the Scientific Method
Rather than committing $175,000 to a CTV buy before seeing any results — a legacy of linear TV buying that I refer to as the "bet it all on black" method — I encourage marketers to approach CTV using the Scientific Method applied to advertising. That is Test, Learn, Iterate and only then Scale.
If you have ever marketed on Facebook or Search, this approach will sound familiar. Similarly, in CTV, you can start with a few hundred dollars per day and quickly adjust the creative, targeting or other variables to dial in ROAS.
Then, once you have dialed in ROAS, incrementally scale your campaign — carefully maintaining positive ROAS — against the 200 million consumers that are viewing CTV programming today. This is actually not that difficult if you really think about it, as long as you have access to the right tools, platforms and the right data.
Work with platforms, not networks disguised as platforms
This is closely related to the first point, but understand the fundamental difference between a "network" and a "platform." Networks operate on an opaque "buy low, sell high" basis. Platforms operate on a fully disclosed fee structure.
Generally speaking, arbitrage networks are prevalent in new "wild west" categories like CTV and get squeezed out over time by platforms offering total transparency and control.
We've all seen this movie play out over and over. The only way to work with a network is if you are OK with paying a CPA regardless of the network's margin — that assumes the media risk is on them.
With all the changes occurring across digital and traditional media, marketers should anchor themselves in a clear plan, centered on collecting the data that will guide their marketing teams to drive profitable CTV outcomes at scale.
CTV delivers massive reach with the advanced targeting capabilities to connect with your ideal audience, has trackable ad exposure with 1:1 attribution measured through household IP addresses, facilitates “hands on” media buying with no minimum advertising spend, and has the ability to track the measurable “Halo Effect” across channels within a household and across channels.
When it comes to performance CTV advertising, the only question marketers need to ask is “Why the hell am I not doing it?”
While I agree with the premise and recommendations, it is worth pointing out that Nielsen Research reports that "In first-quarter 2021, video streaming in the U.S. accounted for 25% of total TV use per day" which is far less than the unsourced related stat stated in the MediaPost article. It is also worth noting that Nielsen also reports that only 34% of CTV viewing time is ad supported, a critical factor for advertisers (Source: https://www.nielsen.com/us/en/insights/article/2021/getting-ahead-of-the-curve-in-ctv-advertising/).
The writer claims that streaming services now account for 63% of all TV viewing---an impossibly high percentage---then adds that this "equates to 90 million + ad-supported CTV households". Is this a case of confusing reach with frequency? If 63% of all TV homes use CTV per week---per some unspecified study---that does not mean that streaming accounts for 63% of all viewing. Also, the just released ANA Innovid study found that the average brand---of 20 studied---attained a 13% CTV reach by buying 60 CTV Grps. Is that "massive reach"?
Thank you Marco for the POV! I always appreciate dialogue and even an occasional debate! :-)
The 61% data point comes from Samsung research (https://www.samsung.com/us/business/samsungads/insights/), based on ~10s of millions of CTVs in the US, measured by ACR chip technology that identifies exactly what is displayed on the actual OEM TV glass. And while reasonable minds and reasonable research may disagree, I personally place a lot of weight in ACR tech (and massive ACR panel sizes), having been close to the original development of the category. And I agree that ad supported CTV/Streaming is the key metric, but here again research mileage varies considerably (other than most research points to AVOD being the fastest growing CTV category due to "subscription fatigue" among other factors). I think, however, we are in agreement on the broader points: a) ad supported CTV is a large and rapidly growing category; and b) technologies now enable marketers to buy and measure CTV ad efficacy in a digital-first way, based on clicks, conversions, CPI and, ultimately, ROAS.
Thank you, Claudio for the POV! I always appreciate the dialogue and even an occasional debate! :-). The 60%+ data point comes from Samsung research (https://www.samsung.com/us/business/samsungads/insights/), based on ~10s of millions of CTVs in the US, measured by ACR chip technology that identifies exactly what is displayed on the actual OEM TV glass. And while reasonable minds and reasonable research may disagree, I personally place a lot of weight in ACR tech (and massive ACR panel sizes), having been close to the original development of the category. And I agree that ad-supported CTV/streaming is the key metric, but here again research mileage varies considerably (other than most research points to AVOD being the fastest-growing CTV category due to "subscription fatigue" among other factors). I think, however, we are in agreement on the broader points: a) ad-supported CTV is a large and rapidly growing category; and b) technologies now enable marketers to buy and measure CTV ad efficacy in a digital-first way, based on clicks, conversions, CPI and, ultimately, ROAS.
If CTV expects major TV advertisers to invest significant amounts of branding dollars in this form of TV, it's going to need measurements which are the same as are utilized for TV whether it be "linear TV" ---which isn't going away just getting smaller in terms of time spent tonnage---or AVOD. Those measurements need to focus on who is watching and, ideally, to what extent the viewers are attentive to content and commercials. It's all well and good to talk about CTRs, ROAS, conversions, etc. but those metrics are not the basic ones for branding campaigns as these advertisers do not, for the most part, sell direct to consumers. So long as CTV remains a digital specialist media buy, it will encounter resistence or indifference ---needlessly---from branding advertisers---and be handled separately---like a specialist buy as opposedto being a basic part of normal media planning and buying process when TV is the basic element in a media plan.
Regarding the Samsung findings that 60-65% of all viewing is now done via streaming, I'm a reasonable person and I consider that to be a very hard to swallow statistic---no matter how many millions of smartsets are being monitored. It just doesn't make sense. For example,how do these smartsets know how many people are watching? And to what extent is this panel representative of all TV sets in the country?How does it account for those who are not connected to the internet? How does it compare to known population profiles by demographics? What does it report as the average TV usage rate per day---7 hours per home,10 hour per home? etc. What happens if two sets are in usa at the same time---and tuned to the same channel or program? Is that one or two exposures?
Hmmmm.
I am having trouble with the veracity of the claim/POV that "Currently, 63% of allTV viewing is already done on streaming services, which equates to 90 million+ ad-supported CTV viewing households."
The Samsung data is interesting (variously described as 63%, 61% and 60+%) being the proportion of the total TV usage time that was CTV content. The claim is based on ACR which I have no quibbles with. It must also be remebered that this is for Samsung TV's and not all CTVs but I doubt that CTV usage would vary dramatically between CTV brands, so no quibbles there.
The Samsung data was then equated to ~90m households. The US has 128.5m households , so that is a 70% CTV penetration rate. That is, 30% of houesholds don't have the technology to watch any CTV.
The data also reveals that within the 70% CTV penetration rate, the CTV usage rate is 63%.
Taking the data provided as bona fide, this means that the 30% of the homes with no CTV watch watch nil CTV, and that in the remaining 70%, 63% of the content they view is CTV and 37% is not CTV content.
If we can make the heroic assumption that the nett viewing time in CTV capable homes is basically the same as in non-CTV homes, and that the number of viewers of the TV is basically the same in CTV homes and non-CTV homes, then we can calculate the proportion of CTV viewing as follows:
~ in the 30% of homes without CTV the CTV share is 0%
~ in the 70% of homes with CTV the CTV share is 63%
~ then (30% * 0%) + (70% * 63%) = 44%.
Given the above analysis of the data provided in the article surely it should be "Currently, 44% of all TV viewing is already done on streaming services..."
QED.
That's probably what happened John. As I suspected, it would be a confusion between reach and frequency with the former assumed to equal the latter. However, even so, we need to question whether the ACR panel is representative, how it measures viewing as opposed to tuning, whether these are really average minute or time spent figures, etc.
Agreed Ed. Yes, an indicator as to what is meant by "massive panel sizes" would be handy. But, assuming randomness, once you have a panel of five figures it rarely does great violence to reality at a macro level. Of course if you drill down (e.g. to state level) it can get wonky.
And of course we need to take into account that ACR is 100% tuning, with little or no evidence of the actual size of the audience. You can pretty safely assume that there is an audience if you see the content change frequently and/or in short to medium time frames (e.g. a channel change - but also taking into account that ad breaks can appear as though they are a channel change) that it is valid viewing but you don't know by how many. If you see (say) three hours without a change that tuning should be removed. You could impute an audience by applying 'average people per set' data based on something like a Nielsen panel, or you can play it safe and just apply an audience of n=1 in the knowledge that would be a low-ball estimate.