Several trade groups representing parties involved in online advertising have sent an open letter to the Coalition for Better Ads (of which they are themselves among the largest members) pushing for faster implementation of self-regulatory moves intended to stave off the threat of browser-based ad blocking. The context here is moves by browser makers – notably Google and Apple – to get tougher on bad ads and cookie-based tracking respectively, both of which threaten the online ad industry. The industry would therefore like to put in place self regulatory measures which have been discussed for some time but not implemented as a way to try to stave off more of this stuff, though the Apple changes have already gone into force and Google’s are likely to do so as well. The online ad industry only has itself to blame for failing to self-regulate sooner and more effectively and thereby maintaining an environment in which such moves by tech companies are deemed necessary. Poor online advertising really serves no-one well in the long term but the industry’s short-termism in allowing it to continue unchecked is now leading to nasty long-term consequences which it is essentially powerless to reverse.
via Marketing Land
Facebook’s COO Sheryl Sandberg was interviewed today by Axios’s Mike Allen on the subject of Russian election interference and other topics, while Facebook also issued some data about the effectiveness of its program to flag fake news on the platform. At the same time, the Washington Post reports that Facebook has removed a set of data from its site and tools which allowed for analysis of Russian-related postings.
The Sandberg interview shed little additional light on the topic, with the only real news that Facebook is sharing both the ads bought by Russian-backed entities and additional details associated with them with Congress, which in turn may share them publicly. However, she was also asked whether Facebook was a media company, a characterization she pushed back on, leading to several articles from actual media companies arguing that she’s wrong. There continues to be something of an ulterior motive for these stories given the tense relationship between Facebook and the media, but I continue to believe that these characterizations are wrong. To my mind, Facebook is much more like a cable TV company than a TV programmer, putting together a package of content for users but not mostly doing that programming itself, while selling ads that appear within the programming. I don’t think most would argue that cable TV companies are media companies or that they’re responsible for the specific content of the programming, while they are responsible for establishing general policies and rules about what will run on their platforms.
The data Facebook shared on its fake news flagging effort suggests that a fake news label applied after fact checking from third parties effectively reduces sharing and views once it’s applied, but the problem has always been that it takes several days for this to happen, which means most of the views have already happened by the time it takes effect. It shared the data with its fact checking partners as a way to incentivize them to do better (something they’ve been asking for) but without massive new resources from Facebook or elsewhere, it’s not clear how those organizations will be able to work faster or cover more ground. That, in turn, will continue to limit the effectiveness of the program.
Lastly, Facebook says the data it has pulled from its site with regard to Russian accounts should never have been available in the first place, and its disappearance therefore reflects the squashing of a bug rather than a decision to pull otherwise public information. Whether you believe that or not likely depends on your view of Facebook’s overall level of transparency in relation to the Russia story, which has clearly been limited. It appears Facebook at a corporate level is still desperate to control the flow of information about Russian influence on the platform, which likely isn’t helping its PR effort here – better to be as transparent as possible so that all possible bad news can come out quickly rather than continuing to trickle out.
Google has apparently now, like Facebook and Twitter, found at least some spending by actors tied to the Russian government on its platforms, including YouTube and Gmail, and the Washington Post says the amounts spent were in the tens of thousands of dollars. However, the New York Times reports that the actual amount definitely spent by entities connected to the Kremlin was much smaller, at around $4,700, while there is another $53,000 that was spent by Russian entities which have not yet been proven to have a connection to the government. Unlike the money spent on Facebook, of course, ads on Google’s platforms have far less potential to drive viral activity, meaning that the direct reach of the ads was likely much of the total reach, and that amount of money wouldn’t have bought much of that. Google doesn’t seem to have commented on the record about any of this yet, but my guess is that the Times story was pushed by Google PR to provide context on the Post one. But this does draw Google further into the mire that’s already engulfing Facebook and to a lesser extent Twitter, something of which we saw further evidence over the weekend.
I’m actually tying three stories together here, only two of them referenced in the headline. The first is news that Facebook is tightening the review process for ads that seek to target by politics, religion, ethnicity, or social issues, requiring human approval before these ads can be shown to users. Secondly, Facebook’s Chief Security Officer, Alex Stamos, went on something of a Twitter ant on Saturday in which he complained about what he described as overly-simplistic coverage of complex issues by the media. And thirdly, CBS had an interview on Sunday with the Trump campaign’s digital director, who claims that it worked in very direct and sophisticated ways with Facebook to do micro-targeting of its ads, including having Trump-sympathetic members of the Facebook staff working directly with the campaign in its offices.
The ad review change is a sensible one in response to recent revelations about how these tools were used in the past, but is likely to catch lots of entirely innocent activity too – e.g. someone targeting members of a particular religion with products or services relevant to them – and will likely slow down the approval process for those ads. It will also slow down the approval process for political ads during campaigns, when the volume of ads tends to rise dramatically, and the review team will need to be augmented significantly. That delay could prove costly as campaigns become more nimble in responding to news in real time and want to target ads immediately. We won’t know the impact of that until next year, as mid-term campaigns ramp up.
The Stamos rant garners some sympathy from me, because I agree that some of what’s been in the press has assumed that Facebook should have been aware of these attempts to game its systems at a time when the US government and security agencies hadn’t yet addressed the issues at all in public. But the rant is also indicative of what appears to be a split between the security and engineering teams at Facebook, which clearly want to speak out more, and the PR and broader senior management team, which seem to want to say as little as possible – several reporters I follow on Twitter responded to the thread with frustration over the fact that Facebook hasn’t made people available to talk about the details here.
Lastly, the CBS story doesn’t seem to have been picked up widely and may be partly exaggeration on the part of the source, but there’s no doubt that the Trump campaign did use the tools Facebook offers extremely effectively during the campaign, and that it played an important role in the outcome. What’s important here is that its uses were all legitimate, in contrast to the use of Facebook by Russian actors claiming to represent US interests, but the effects and even techniques used were in many ways similar. Even as Facebook clamps down on one type of influence, the broad patterns will remain similar, and as long as foreign actors can find US-based channels willing to act as fronts, it’s going to be extremely difficult to shut down this type of activity entirely.
Facebook still hasn’t shared all of the details of the ads bought by Russian agents on Facebook over the last few years with Congress, and hasn’t really shared any of the details with the general public. However, some of the details have emerged regardless, and one researcher has used that information to do some analysis of the reach of some of the posts on the accounts controlled by entities tied to the Kremlin. What he found is that the organic reach of those posts has been enormous, much larger than the numbers reached by the ads themselves alone as reported by Facebook, suggesting that Facebook is using the narrowest possible definitions of reach in its reporting and thereby downplaying the impact.
Until Facebook does release the full details of the Russian operations, we can’t know the true reach for sure, and this analysis is merely indicative of organic reach achieved by half a dozen of the biggest accounts we do know about. But it’s clear that the operation was both sophisticated and very effective in reaching large numbers of people, leveraging many of the same techniques used by legitimate news organizations and others on Facebook. Given that these techniques are all available to anyone who uses Facebook, the only way they could have been stopped is if there was clear evidence that the accounts behind them were “inauthentic” (to use Facebook’s terminology) way earlier in the process. And given that neither it nor the US government were actively investigating that possibility during the election, that was never likely to happen. It’s also not clear how Facebook would go about policing this kind of thing going forward.
Google has created a pair of machine-learning-powered filters for website owners which will allow them to avoid hosting racy or seedy ads on their sites. Neither of these categories are banned by Google’s policies today, and anyone who’s visited many news sites online has seen the types of manipulative, low-class ads caught by the second set of filters, while the first set will use an algorithm for skin detection to weed out ads tending towards the racier side. This set of moves is on the flip side of the controversy around YouTube and Google’s broader ad platforms earlier this year, when it was the advertiser brands that didn’t want to be associated with undesirable content, while this deals with the opposite problem of undesirable ads showing up next to high-quality content. Both are part of the challenge for Google and other online ad platforms in a highly automated, entirely price-driven system for placing ads, and it’s good to see Google offering site owners some more quality controls, just as it improved controls for advertisers following the boycott earlier this year.
Third party social media metrics company Delmondo says that across a selection of Facebook Watch videos it measured, average watch time was 23 seconds. That’s a little higher than the 17 second average for videos in the News Feed, but not much. It’s notable, too, that 20 seconds is the minimum amount of time a video must run before a mid-roll ad can be shown to the user, and I wouldn’t be surprised if those mid-roll ads are a big reason why average watch times are around that level. I continue to believe that Facebook’s mid-roll focus is going to harm viewing and ultimately ad revenue for its video platform, and it badly needs to re-think that approach. It’s still early for Watch in particular, and it’s clearly more of a destination for video rather than something users stumble across accidentally as with the News Feed, but it needs to grow well beyond 23 seconds if it’s going to be worthwhile either for Facebook or its content creators longer term.
Digiday has done some digging on the CPMs (payout rates for advertising) on Facebook’s video platform, and has found that on the whole they’re pretty low. One publisher suggested an average CPM of 15 cents, which is indeed low by industry standards, and that theme if not the exact amount was confirmed by others Digiday spoke to. One big challenge, though, in measuring CPMs or other industry-standard ad metrics is that many publishers publish videos which don’t use the mid-roll ad format alongside those that do, so the denominator may be skewing the results a little. But what seems clearer is that mid-roll ads perform far less well than the pre-roll ads YouTube has used very successfully, something I predicted in this piece a few weeks ago on Facebook’s big video pivot. I suspect Facebook will struggle to compensate creators adequately as long as mid-roll remains essentially the only way to monetize videos, and when it likely drives a high abandonment rate.
eMarketer, a research firm which offers projections of ad revenue by company, has lowered its 2017 US forecast for Snapchat for the second time this year. It’s now projecting a total of $642m for the year, down from the most recent forecast of $770m, and the original forecast of $800m. Projecting Snap’s ad revenue is difficult for several reasons, not least that the company itself doesn’t provide any guidance, but also because its shift to serious revenue generation began so recently that there’s no reliable growth rate to base future projections on – year on year percentage growth has slowed from over 400% to 286% to 153% in the last three quarters even as dollar growth has been pretty strong. Snap’s North American revenue in the first half of 2017 was $277 million, meaning that eMarketer is projecting roughly 30-40% growth in the second half over the first half. That’s fairly modest, but as we’ve seen recently Snapchat’s user growth has been modest too, and although ARPU is rising fast, it may not continue to do so at the same pace. There’s growing skepticism about Snap’s business overall, and this report feeds into that overall skepticism, but I suspect it may be a little too pessimistic based on Snap’s strong second-half revenue performance last year. But we’ll know soon enough what Q3 looks like, at least.
via Business Insider
Facebook has made yet another announcement in what’s rapidly becoming the saga of Russian ad-buying on the platform and the ongoing fallout from it. This time around, it says it’s going to share the details of the 3000 suspicious ads placed on the platform with the US Congress, and it’s also going to hire a thousand additional people for its ad review team to ensure that inappropriate ads don’t get through. The rest of the announcement focuses mostly on fleshing out promises made over the last couple of weeks, though there’s still relatively little transparency on what’s actually going to change and/or when in some cases. Over the weekend, Mark Zuckerberg also personally apologized for any role Facebook may have had in sowing divisions in the world and promised to work to make things better in future, as part of a post relating to the Yom Kippur Jewish holiday. It’s clear that he’s taken all of this far more seriously and increasingly personally as well over recent months, though many still want him and Facebook to do far more to increase transparency over how Facebook has been used for ill and how it will change as a result.
Roku IPO Sees Stock Rise Over 50% on First Day (Sep 28, 2017)
Roku went public on the NASDAQ today after filing its S-1 earlier this month with the SEC, and saw its stock pop on its first day of trading, rising from its $14 opening price to as much as $23 in the middle of the day before falling a little, settling in at around $21 at the time I’m writing, about a half hour before markets close. That’s a great start for Roku, which was far from a shoo-in as a consumer tech IPO given its big business model pivot, its losses, and the fact that it competes against three of the biggest names in consumer tech in Amazon, Apple, and Google. Other big consumer tech IPOs this year haven’t fared well, notably Snap Inc’s, but the main reasons for the poor stock performance have been grounded in poor company performance, so we’ll have to see how Roku fares in its first couple of quarterly reports, with the first one likely coming a month or so from now. To commemorate the Roku IPO, I added a Roku deck to the Jackdaw Research Quarterly Decks subscription service today, and will be recording a video voiceover for the deck shortly (a discount is available for those who buy both subscriptions, so contact me if you’d like the discount code for that).
NBC Universal is going to start allowing big advertisers to automate the placement of some of their ads through the use of an API. What this means in practice is that brands will be able to use data they have on which shows their target customers are likely to watch to select exactly when their ads will play on the various NBCU channel. Despite the obvious similarity to automated ad buying online, this is going to be far from a free-for-all: it’s starting with one partner (Target) and this will presumably be limited to big, established brands, and will certainly involve pre-approval of all the actual creative to be shown to viewers. But it’s yet another front in the ongoing war between online ad platforms and the TV companies, with the latter constantly aping the former’s techniques while claiming to be superior in other ways. The reality is that every big brand is going to advertise in both places, and Facebook is now actively pursuing a strategy of trying to tie those two channels together at a measurement level, which feels more realistic than the TV companies’ denial that the online platforms offer anything of value to advertisers.
via Business Insider
Google has announced several new tools for advertisers using its platform to reach users with video ads, and they highlight just how sophisticated the YouTube ad platform is becoming, at a time when Facebook is still struggling with basic formats and helping creators tweak their video formats to work with its ad limitations. There are four parts to the YouTube announcement: custom affinity audiences, which allow advertisers to reach users based on profile-based interests including recent Google searches; customizing video ads by context on the fly using automation; stringing together multiple ads to tell a story or react to user responses; better online-to-offline attribution. To my mind, the custom video ads are the most interesting thing here – they allow advertisers to upload a set of assets and have the system automatically mix and match them to create ads that feel like they’re customized based on the video the user is watching. As this TechCrunch article points out, that’s likely to make the videos more memorable, but it may also cross the “creepy” line for some viewers, and that’s the risk all highly-targeted advertising takes. Various elements of what Google is announcing take advantage of its increasingly strong AI and machine learning techniques as well as the breadth of its tracking of users (for better or worse) across the various properties it owns, and the latter may in future be hampered by increasing limits on this kind of targeting which will come into effect in Europe soon.
Twitter Sells Enough Ads to Launch All Planned Live TV Shows (Sep 25, 2017)
It certainly wasn’t clear at the time Twitter made its big blitz of announcements around its live TV plans that some of the shows weren’t guaranteed to air if they didn’t get sufficient ad backing, but now that they have that backing, Twitter is apparently trumpeting that fact. Since many of the shows Twitter is hosting are existing properties which will come with ads from the original sources, Twitter likely didn’t have to sell that many ad slots itself in many cases. There certainly are some unique-to-Twitter shows, so it’s impressive that it’s sold enough ads on those too, but in many cases I’m guessing that spend is experimental – no-one really knows what kind of audiences most of these shows will attract, and the level of spending involved is likely small enough to fit into niche budgets (as Snapchat long did). The big question is whether, following the first few months of this experiment, those advertisers want to re-up and commit to additional shows and seasons. That will depend largely on a combination of viewership and engagement with the ads viewers see. We don’t have many figures for individual Twitter shows to go by, but we do know that just 55 million or 17% of monthly active users spent any time watching any live video on Twitter in Q2 of this year, so Twitter and its advertisers are clearly hoping that that translates into more committed audiences for specific shows in order to justify continued investment.
Facebook has announced that it’s partnering with Nielsen to provide advertisers with a combined measurement of brand lift for campaigns that run across both Facebook and TV. That provides a consistent set of metrics for advertisers that use both platforms, but more importantly it puts a big dent in the idea that Facebook and TV are at war, a narrative the media seems keen to perpetuate but which Facebook itself has repeatedly downplayed. While it’s certainly the case that Facebook is chasing some of the same ad dollars as TV, and Facebook has even made the case that TV ads are less effective than Facebook ads, it’s also pushed back against the idea that it’s trying to kill TV advertising. This partnership suggests that Facebook is realistic about the fact that most advertisers are going to continue to run ads both online (including on its platform) and on TV, and that it can best support those advertisers by making it easier to measure the performance of campaigns in both media. It’s also making the argument that campaigns that run in this way actually see better results than those which only run in one place.
Last month, the Wall Street Journal reported that Google was preparing to offer refunds to advertisers whose ads had appeared on sites with no legitimate traffic but who were nonetheless charged as if there had been traffic. The problem with the refunds was that Google itself could only refund a small portion of the total because much of the money from the advertisers went to ad exchanges also involved in the transactions. Google has now announced the refund program publicly, and said that several major exchanges will also join in offering refunds, which should return a much more substantial part of the total payment. As also reported earlier by the Journal, Google is also going to be participating in a program called ads.txt, which makes it much easier to validate sites claiming to have large amounts of traffic and therefore cuts down on ad fraud, which should be a big help in this specific type of fraud, even as others persist.
A key part of the Advertising Sustainability narrative on the site is the issue of two companies’ dominance of online advertising in the US and to a lesser extent other western markets. New data from eMarketer is a useful checkpoint in measuring that dominance – it says that the two companies will suck up 63% of US online ad spend in 2017, an increase from its earlier forecast of 60%. Microsoft comes in third place way behind the top two, with Verizon in fourth for now and Amazon projected to take its place over the next couple of years. Google and Facebook’s dominance is neither surprise nor mystery at this point – the former has the unique combination of timeliness and relevance that search offers, while the latter has created the most powerful combination of audience and native advertising, dominating their respective categories and leaving the dregs for smaller competitors and less effective forms of advertising. Importantly, though, eMarketer doesn’t see the two companies’ share rising dramatically over the next couple of years – it projects just 68% share in 2019, meaning that other companies will still capture nearly a third of the market, and their dollar share of the total will actually rise since the market is still growing rapidly. eMarketer’s blog post with all the numbers is here.
It’s been increasingly clear in recent months that Spotify has big ambitions for its advertising operation, even going so far as to pitch itself as a threat to Google and Facebook in time. One thing essentially every big ad platform has in common, though, is self-serve tools to enable the long tail of small and medium-sized businesses to buy ads, and that’s an area Spotify hasn’t emphasized enormously just yet. One big challenge is that audio ads are rather tougher to create for small businesses than display ads, and that’s one of the things Spotify is now looking to solve with what it calls the Spotify Ad Studio. The tool will allow advertisers to upload a script and choose music and other options while text-to-speech technology creates the voiceover. That sounds like it could be terrible if it’s anything like other robotic sounding TTS software, but the key is that it dramatically expands the range of advertisers that could run audio ads on Spotify, which now has a large and rapidly growing audience in the US. Given how poorly ad-based streaming music is monetized today, anything which boosts the demand side of the ad market should help raise prices and therefore improve the overall economics, though it’s never likely to rival paid streaming in terms of revenue per user.
Facebook Announces Further Changes to Ad Targeting Options (Sep 20, 2017)
Facebook COO Sheryl Sandberg posted to the site today to address the issue of offensive terms appearing in the targeting options for would-be advertisers, a problem that emerged last week and which Facebook issued a temporary fix for later in the week. As I said in commenting on Facebook’s initial tweaks, those didn’t feel like a permanent solution and I predicted that it would slowly dial the temporary limits back as it found more long-term answers. Sandberg’s post today both serves as a mea culpa for not detecting and fixing the issues more proactively, and as a confirmation of my prediction: Facebook has begun allowing some of the most common user-specified interests back into its ad targeting tool and will continue broadening those that can be used over time with more human curation. It will also be clamping down more (though in unspecified ways) on ensuring the actual content served up through ads is appropriate. I’ve felt since this all first came to light that the response to it was overblown, and that the criticism Facebook has faced over it was far too harsh, and we discussed this in some depth on the Beyond Devices Podcast last week, in which my co-host Aaron Miller took the opposite view. A piece in Slate today is particularly hard on the company on this front, arguing that the company’s pursuit of profits has somehow blinded it to these issues. The reality here is that at Facebook’s scale almost any potential misuse of its platform will squeeze through somehow simply because Facebook can’t possibly police it thoroughly enough to eliminate it entirely without also generating lots of false positives. The scale of the problem identified last week and its likely impact were so minimal as to be almost insignificant, and in general Facebook is making good progress on this front and on others in taking more responsibility for policing its platform and minimizing its potential for harm. I’m therefore more inclined than others to cut it a break.
Uber is suing ad agency Fetch over what it says is fraudulent reporting of ads placed and clicked on, and resulting downloads of its apps by those who never actually saw ads placed by Fetch. Uber is withholding some of the money it owes Fetch while pursuing the lawsuit for a rather larger sum of $40 million, a little over half of what it’s paid Fetch in total over the last three years. Presumably Uber feels it has decent evidence to support its claims, given that – as Bloomberg points out in the article linked below – it’s not a particularly litigious company despite being a target of others’ lawsuits frequently. Fetch, meanwhile, has spoken in the past about the issue of ad fraud and the challenge of identifying and reducing it, something that’s by no means unique to the company in the broader world of online advertising. Ad fraud continues to be one of several big issues facing ad-based companies and complicating their relationships with brands and buyers.
Update: On September 27, 2017, Phunware, one of the mobile ad firms Fetch used to place ads, is now suing Uber over non-payment, as the latter is withholding payment from Fetch during the lawsuit. Uber says it feels Phunware is one of the parties which engaged in the fraud and will present evidence of this in court.