This is an edited transcript of a talk that The Drum’s Promotion Fix columnist, Samuel Scott, recently gave at The CMO Network in the UK.
The marketing industry had so many expectations of the digital world. For example, everything would be real, measured and trackable. But so many of those expectations turned out to be wrong.
In short, much of what we see online has turned out to be completely and utterly fake. We hear so many people talking about being ‘digital first’ or ‘digital only’. But is that just completely and utterly wrong?
To begin, let’s go back to communications channels before the internet. Radio reports listener figures that represent real human beings. Print media does the same for readership and uses surveys to calculate how many people on average read a single copy. TV does the same for viewership and we can even take into account the average percentage of people who do not go to the loo and do watch the ads.
And all of these numbers are audited by independent third parties. Sure, there have been instances of – shall we say – ‘creative accounting’. But by and large, marketers have a good idea of how many human beings we are reaching when we use traditional media.
But I argue that we do not have good numbers for most internet-based channels. We do not know whether the numbers in analytics dashboards mean anything real. And companies such as Facebook, Google, and ad tech platforms do not let independent third parties audit their numbers. We just accept whatever they tell us.
So, today, after 20 years of Google, Facebook and Amazon ‘disrupting’ everything, we now must disrupt the idea that we see online is real – and discuss what the implications might mean.
Let’s start with web traffic. You can open platforms such as Google Analytics and see visitors, conversions, sales and a lot of other information. But do you know where those numbers come from?
Analytics tools put a piece of code on a website. When a browser loads a page, that code places a cookie in the browser and records what the browser is doing. But if you use an ad blocker, it stops the code from working at all. And roughly 22% of people here in the UK use ad blockers. So, 22% of your website visitors don’t even show up in analytics, and you have no idea what those 22% are doing.
It gets worse. Most analysts have found that around 50% of website traffic is bots – not human beings. I could pay money to send a bot to any website and do almost whatever I want. More on that later. So, at least half of the numbers in your traffic analytics platforms are fake.
Here is how Ellen Pao, the former chief executive of Reddit, put it on Twitter:
“It's all true: Everything is fake. Also, mobile user counts are fake. No one has figured out how to count logged-out mobile users, as I learned at Reddit. Every time someone switches cell towers, it looks like another user and inflates company user metrics.”
On regular radio, getting a number one hit actually means something. But the New York Times recently found that you can buy 5,000 fake YouTube views of videos for as low as $15.
And in addition to those who use bot traffic, there are pay click farms in China that literally consist of people spending their whole workdays rewatching the same videos, revisiting the same websites, and downloading the same apps over and over again. You can find videos of these workers online. Seriously, it is factory farms full of workers who just do those things all day long.
In 2017, the Spanish pop hit Despacito became the most popular song on YouTube with more than 3.4 billion hits. But does that number really mean anything? How many of those “hits” were bots or people paid to watch the video over and over again? We have no idea. But the New York Times found someone in Canada who made $200,000 in 2018 from selling 15 million fake YouTube views.
There are hundreds of Facebook Groups with tens of thousands of members who want to buy and sell reviews on Amazon. In one example that BuzzFeed News found, you can get $36 for writing a positive review of a pair of headphones. Does a positive review on Amazon by someone you don’t know not using a real name – even if it is a “verified purchase” – actually mean anything?
I increasingly think “no.” And I have the same opinion whenever I see comments on Airbnb, Yelp, and every other website. After all, The Washington Post found that 61% of the Amazon reviews of electronics products are fake.
The New York Times set up a dummy Twitter account and paid the service $225 for 25,000 followers. I also admit that once when I was in the agency world years ago, I did buy Twitter followers for a client who wanted to increase the company’s followers. It is very easy – and very cheap. And it was terrible that I did that.
I can tell you that as a professional marketing speaker, I know that event organisers often evaluate speakers by their Twitter followings. And some people do buy followers to get a high number even though the followers are only bots. Even though it might hurt my speaking career, I refuse to do that.
Research from two universities in America found that up to 15% of Twitter accounts are bots. I would argue that the number is probably far higher.
The agency Mediakix once created two two fake Instagram accounts. One used images borrowed from a Los Angeles model. The second used free stock photos. Soon, brands offered both of the accounts free products and more than $500. Both brands and followers were duped.
In another example, Points North Group found that the Ritz-Carlton hotel’s sponsored posts have been published by influencers with followings that were 79% fake.
Captiv8 estimates that brands pay $2,000 for influencers with 100,000 followers and $20,000 for those with a million followers. But The New York Times paid $225 to buy 25,000 followers -- that is 1 cent each. So, if you do the math, being a fake influencer can make you a lot of money. But it makes marketers look like idiots.
Netflix said that 45 million “accounts” “watched” the movie Bird Box. But what do “accounts” and “watched” mean? Many people share the same Netflix account. And does Netflix consider a “watch” to be one second, the whole film, or something in the middle? We have no idea.
Netflix viewership numbers could be fake. We just don’t know.
First, if you are judging the success of your marcomms campaigns by “engagement” or “likes” or “shares,” you have bigger problems than the fake Internet. Still, you should know that there is a vast number of services that sell automated engagement.
The anti-fraud company Sway Ops looked at a single day of Instagram posts that were hashtagged with #sponsored or #ad. What did they find? More than 50% of all engagements were fake and only 18% of comments were made by real human beings.
Engagement is fake and easily manipulated.
Now, remember what I said earlier about bots and web traffic analytics? The same is true for display ad impressions from Google or any other ad network. You might think that the number of “impressions” is the number of humans who saw an ad. But that is wrong. One impression is one web browser – or bot – asking one ad network server to load one ad one time. That is all.
It has nothing to do with human eyeballs. And if half of all web traffic is fake, then at least half of all ad impressions are fake.
The measurement company Adjust looked at a traffic flow sample of 400 million installations over 17 days and estimated that $1.7m worth of installations were being paid to fraudsters faking the activity.
In 2018, researchers found that 11% of all app installations were fake – and that number increased 30% from 2017.
In September 2016, Facebook admitted that it overstated average video view times by 60% to 80% for two years. That same month, an Australian ad magazine obtained Nielsen figures showing a 94% plummet in Facebook video streams in the country due to a correction in how the data was reported.
Also in 2016, Facebook's VP for Europe, Nicola Mendelsohn, said “we’re seeing a year-on-year decline in text". Mark Zuckerberg also said "five years to all video". Now, there are lawsuits that Facebook allegedly inflated video ad views by up to 900%. For publishers, the “pivot to video” seems to have been a big mistake that cost countless people their jobs.
Facebook also reports a “daily active visitor” metric for its Watch product as someone spending at least one minute on the platform per day – but get this! Those 60 seconds do not need to be consecutive. If someone watches one second each of 60 videos, then that is a “daily active visitor”. Seriously.
Remember, TV networks measure a “view” of a show based on whether it was watched for 60 consecutive seconds.
But for Facebook specifically, the issue is more dire. I am sure everyone here knows that Leave won the Brexit vote through an illegal, last-minute ad push on Facebook. And because of the nature of Facebook’s platform, we do not know exactly who bought what ads, what the ads said, where the money came from, and who saw them. But we do know that the last minute ad spend was illegal.
Cambridge Analytica whistleblower Christopher Wylie once said that when Olympic athletes cheat and win, their victories are taken away. We know that Leave cheated. So why is the Brexit referendum victory not taken away? I just do not understand.
I will be more blunt. Foreign states and bad actors pushed enough illegal, last-minute winning votes to Leave with fake news, fake pages, fake groups, fake people, and fake advertising – and all primarily over Facebook. The result? The fracturing of the European Union and the destabilisation of one of the most important democracies in the West. Brexit would not be happening if Facebook did not exist. Donald Trump would not be president of the United States if Facebook did not exist.
One week after parliament’s landmark report on fake news and foreign interference through social media, one week after Damian Collins called Facebook “digital gangsters,” why is that not the top story in every single newspaper and BBC TV and radio broadcast? Where is MI5 and MI6? Why are people not rioting in the streets? Where is the government? Really, I want to know.
The bigger marketing context
To summarise, I will quote Aram Zucker-Scharff, the director of ad tech at the Washington Post:
“The numbers are all fucking fake, the metrics are bullshit, the agencies responsible for enforcing good practices are knowing bullshiters enforcing and profiting off all the fake numbers, and none of the models make sense at the scale of actual human users.”
He continues: “The problem isn't just that the internet is full of fakery and bullshit and bad numbers and malfunctioning metrics and bullshitters and fraudsters. The problem is that all the fake shit is layered on top of other fake shit and it just compounds itself.”
“The internet is full of fake bullshit, run on fake numbers, profitable on fake models, feeding you fake information, supporting fake users. None of this is by accident.”
Still, I do have some suggestions.
First, we need to ask for audited online numbers just as we do for TV, print, and radio. As an industry, we should refuse to accept anything that Facebook, Google, or anyone else tells us unless every single number everywhere is audited by independent third parties. If their platforms are truly legitimate, they should have nothing to hide.
Second, we need to use the same metrics for the online world that we have use for the offline world.
Stop thinking about metrics such as “engagement” and think about the tactical marcomms metrics that have always existed. After all, digital platforms measure only digital touchpoints.
Marketers are simply doing the same marcomms tactics that we have always done – it is just that we are doing them over an increasing number of offline and online channels.
Some examples. Take the tactic of advertising. One metric is brand lift. Target a city with display ads. Measure the brand lift. Target the city with local TV ads. Measure the brand lift. Do the same over radio and outdoor. And then compare the brand lift in the real world to see which channels are the best. (Spoiler: I doubt it will be display ads.)
Take the tactic of PR. One metric is brand sentiment. Do that same process over all online and offline channels and compare the results.
Take the tactic of direct response. The main metric is ROI. Do direct mail, e-mail, Google AdWords, or anything else and then compare the results and returns based on the cost. In this one example, I will say that online channels are good in the single context of direct response. If anything, the Internet is efficient. But efficient is not the same as effective.
Still, can I make a request? Can we stop asking for the “ROI” of every single marketing activity? Most times, marketers say “ROI” when they mean “benefit.” ROI is the proper metric only for the tactic of direct response.
There is no such thing as “traditional” marketing or “digital” marketing – there is only marketing. When you manage your departments, do not segment by online and offline. Segment by tactic. After all, no one ever said that we do “TV marketing” – and no one should say today that we do “social media marketing” or “video marketing”. Those terms are absurd.
Have a team devoted to direct response. They can do that tactic over online and offline channels. Have a team devoted to advertising. They can do that tactic online and offline. Have a team devoted to PR. They can do that tactic online and offline. The tactic, not the channel, should be the focus. And then apply the tactically-relevant metrics to the results. People are generally experts at individual tactics, not individual channels. We should always be channel-neutral rather than believe idiocy such as being “digital only” or even “digital first.”
As BuzzFeed News media editor Craig Silverman put it, “digital has ushered in a golden age of delusion by measurement.” We need to use the right metrics to combat everything that is fake on the internet.
The Promotion Fix is an exclusive biweekly column for The Drum contributed by global keynote marketing speaker Samuel Scott, a former newspaper editor and director of marketing in the high-tech industry. Follow him on Twitter. Scott is based out of Tel Aviv, Israel.