Context, Correlation, Coincidence, Causation

The Golden State Warriors scored 142 points in last night's Western Conference Semifinals. OK. But did they win? In this case, yes, because their opponents, the Memphis Grizzlies, only scored 112. But 142 points would not nearly be enough to beat the Denver Nuggets, who scored 184 points, but still lost to the Detroit Pistons in the highest scoring game in NBA history 184 -186, in 1983. Context not only matters. Context is everything.

Google's 2021 Ad Safety Report, released 2 days ago, was filled with cool stats like "we block[ed] over 3.4 billion ads, restrict[ed] over 5.7 billion ads, and suspend[ed] over 5.6 million advertiser accounts." Um, ok? So what? Is that significant? Does it mean anything? Without context, the numbers are just sports scores and you won't know if the team won or not. The big numbers were intended to sound impressive. But blocking 3.4 billion ads in the course of the entire 2021 is entirely without merit in the context of the 84 trillion bid requests and 8.4 trillion ads that flow through the Google adsystem in a 12 month timeframe. Blocking 3.4 billion ads out of 8.4 trillion ads is exactly 0.0004047619 or 0.04%.

Now I will challenge every reader to do the following Google search "google blocked billion ads." For your convenience, here's the link -- https://www.google.com/search?q=google+blocked+billion+ads Look at the large number of results. Click on a few of them to spot-check the articles that reported on Google's heroic action. I have not found a single article, yet, that has provided context. If you find one, please let me know, and I will amend this article and cite it. If this isn't a statement of the concerning lack of quality in journalism today, I don't know what is. But I am not here to talk about quality of journalism. Citing stats is easy, just like regurgitating press releases. Providing context is hard work; and it requires a few more thoughts and perhaps some more homework.

Context

Those of you who have read my articles before will note that I try to end each article with "So What?" I try to answer the questions, "what does this mean, why is this important, and what should I go do?" Without it, the article would be just like the hundreds of other articles that basically say "here's some numbers." Cool, cool. you do you.

Now let's talk about digital marketing and what marketers can do to uplevel their digital marketing game. Ever look at your the monthly reports you get from your agency? Yeah, those Excel spreadsheets that show you how much you spent? OK, how many ads did you buy? Where did your ads go? What CPM price did you pay? Oh, you get something more detailed like a placement report that shows you a list of sites and apps and how many ad impressions went to each? Cool, cool. But which sites and apps were good ones versus fraudulent ones? How many clicks did you get and were those clicks from humans or bots? Did those reported clicks actually arrive on your landing page? I think, I hope, you get the point. Always look for the context; always ask "so what?" If you don't, then you aren't doing your job as a marketer, you're just watching sports.

To those marketers. bigger numbers always appear better. They're the marketers that have been addicted to larger volumes of ad impressions in digital media and more clicks, which give the appearance of performance. They also think they got a good deal, because they got more impressions, and the average CPM prices were lower, yay! But those are entirely meaningless if you don't have the context of whether those bigger numbers yielded any more sales and business outcomes.

Correlation

Over the years, I have jokingly written quarterly reminders that "correlation is not causation." I cited the charts from Tyler Vigen's site called Spurious Correlations. The shape of the curve for "number of people who drowned by falling into a pool" appears to correlate to the number of films Nicolas Cage appeared in. But correlation is not causation. Nic Cage appearing in movies did not cause more people to drown in pools. Similarly, the rise in per capita cheese consumption did not cause more people to die by bedsheet entanglement, even though the shapes of the curves overlaps very well.

Coincidence

The same goes for digital marketing. There are lots and lots of stats and lots and lots of big numbers. But most of it is correlation, at best, not causation. I would even go as far as to call it coincidence -- i.e. marketers spending money while sales were occurring. Let me re-iterate, not to be cruel, but to be clear. Sales were happening; marketers just happened to be spending money in digital channels, around the same time -- i.e. coincidence. How do we know? There are few cases published publicly. In 2012, eBay famously turned off paid search ad spending in the western third of the country; and they saw no change to traffic and sales from that part of the country, compared to the rest of the country where they did not pause the digital ad spending. The traffic and sales were coincidental to the digital media spending. Airbnb turned off $800 million of "performance marketing" ad spending during the pandemic in 2020. "CEO Brian Chesky told an earnings call, “our traffic levels came back to 95% of the traffic levels of 2019 without any marketing spend … in Q4, more than 90% of our traffic was direct or unpaid." People would have, and were, going to airbnb.com because they already knew Airbnb. The spending on highly targeted ads was coincidental to the traffic.

Causation

Other cases further illustrate that digital media may not be causing any sales or any business activity. When Procter and Gamble turned off $200 million of their digital ad spending, nothing happened; sales of diapers, shampoo and detergent continued. When Chase reduced the number of sites showing their ads from 400,000 to just 5,000, nothing happened; website visits, new card applications, and cards issued continued. When Uber turned off $120 million of their paid app install spending, nothing happened; app installs continued. All that ad spending, and the large numbers it generated in excel spreadsheets, had nothing to do with sales, business activity, or outcomes. Are marketers spending billions of dollars on getting bigger numbers in spreadsheets? Or is their job to invest that money wisely and drive more sales for their companies? Of course, it's the latter.

So let's get on with this. Determining causation is hard. Understanding fancy academic models for causation is hard. So let's skip those. Let me give you a simple experiment which does not require advanced math, but does require courage. Run turn-off experiments for a week, a month. Run those experiments in a region of the country or in just one campaign line. That is the most straightforward way to see if the digital ad spending caused any business activity or sales, at all. It takes courage to suggest this to your boss; it takes courage to follow through and run the experiment. Perhaps you run an even simpler experiment and create a campaign line using an allow-list, separate from your existing campaigns which run with block lists. If the campaign using a short allow-list of mainstream, reputable sites performs similarly to the legacy campaigns, you can start to tell that buying all those ads, showing them on millions of sites and apps, makes no difference.

Let me leave you with a reference to a great case study where the State of Utah Tourism Bureau showed that their CTV ad campaigns caused a lift in tourism, and even computed the incremental spend from these incremental visitors. See the case study in this article (towards the bottom): https://www.dhirubhai.net/pulse/ridiculous-ctv-fraud-ridiculously-easy-solve-ad-fraud-researcher/

So What?

I have to keep my promise to you. I have to end my articles with "so what?" Why is this important? As marketers, we need to move beyond the "sports scores" of big numbers in spreadsheets, from our digital ad spending. We need to understand those numbers in the context of whether they caused more sales and business activity. It's not good enough to just assume the digital ads did something, when the spending was just coincidental to the sales that were already happening anyway. Those sales and business activity were happening at the same time as the digital advertising; they were not caused by the digital ads.

You are ready to uplevel your digital marketing now. What should I go do? For starters, ask for more context around the numbers you get. Next probe more deeply into whether your digital ads caused any sales. If there were sales, check if it was just a correlation or were those sales truly _caused_ by the digital advertising. We've seen a decade of "correlation and coincidence" in digital marketing. It's time to upgrade your digital marketing by moving to "context and causation." That is your call-to-action, marketers, on this Sunday morning. Are you with me? Let's goooooooo!