What is insight when it comes to data analytics? How do you know when you have a key finding that you can use to tell a larger story? Better yet, how do you know your insight is accurate and reliable? How can someone make an accurate claim but the impact is not? Welcome to the world of deep insights and understanding the motivation and nuance driving numbers. Sometimes you have to peak below the surface to see if the data backs up the claim.

Case in point: on February 1st, 2024 Google made a major claim as part of its quarterly earnings statement. Perhaps you saw the headline: 100 Million Premium YouTube and YouTube music subscribers! This was in part to show that Google’s subscription business is performing well with $15B in revenue, which is an easy number to understand when you have 100M people all paying $13.99/month, plus a couple bucks for extra storage on Google 1. Simple math shows 100M paying roughly $15/month (many include upgrades to premium storage levels) equals $1.5B in revenues per month or $18B/year. Not too shabby when Google’s disclosures show they generated $15B in subscription revenue. The math works, yes? Well no. The revenues are a result of how Google tags revenues into certain buckets, but that’s not where the revenue is really generated. So how does Google accurately and honestly make a claim of 100M subscribers, have the revenue numbers to back it up, but also are actively lying to the public?


YouTube's Free "Premium"

Removed from the YouTube announcement, but included with the YouTube Music announcement of the same benchmark (which has since been deleted), is very critical caveat in the fine print: the 100M number includes free trials. But not just YouTube premium free trials, ANY YouTube trial: Music, Premium, YouTubeTV, NFL Sunday Ticket, Google 1 trials, and more.

Let’s assume the number is exactly 100M and work backwards. According to Statista the “paying” subscribers (that includes free trials) jumped from 30M in October, 2020 to 50M in 2021, 80M in 2022, and 100M in 2024.

That sold for $45/month and was discontinued after 22 months before its primary incentive could really kick in: Upgrades to the latest pixel device in 24 months. Set aside that bit of false advertising for a moment and let’s only focus on subscribers and free trials.

After dropping pixel pass Google made YouTube premium a primary promotion for Pixel phone sales, and it’s been working. The Pixel 8 shipped 30 million units at the $899, price point, coinciding nicely with the jump in users observed by Statista.

So the real insight isn’t that YouTube has 100M subscribers. The real insight is Google is cannibalizing hardware revenues to “pay” for subscription revenues through free trials. The real question is how many people maintain their service after that trial ends? We don’t know and Google won’t tell us. Investors may want to know because it means subscription revenues will only grow if they keep selling hardware.

Example: iHeart Radio

As an advertising and media firm, D2 gets approached by media sales people to keep us informed of what their audience is doing, how they’re consuming content and media, and why we should include their platform in our client’s media plans. We have hundreds of these types of meetings every year. Several years ago iHeart caught my attention saying they were the premiere digital streaming platform with 100M app users! (There’s that number 100M again). And while they can point to a 3rd party source to back that number up, it’s 100% absolutely not true. They get the 100M number by counting all app downloads across all devices going back to the launch of the original apps in the app stores. But that doesn’t mean people are using the app, it doesn’t mean they are unique (people download apps when they get new hardware), and it doesn’t mean they’re listening now.

Example: Donald Trump

Ignoring politics and political leanings, listen to how Trump talks. Everything is a superlative: greatest, biggest, fanciest, best, worst, most, perfect, insane, etc. It always lands on the extreme. Notice how he never has a single source, but instead spoken as multiple parties have already concluded what he’s saying is correct. “People are saying”, “everyone knows”, “you hear it all the time”, etc.

How is this insightful? It helps identify if what you hear is a fair statement or amplified by an echo-chamber effect. By claiming a single outcome is the conclusion from multiple sources he’s really saying what he wants you to believe is true because of an echo chamber effect. He’ll say it out loud, people around him nod in agreement, and that becomes “all the people are saying”. Taken at face value it would appear as there are numerous sources for the outcome, but there is in fact only one. Recognizing an echo chamber when you see one is helpful in discounting the value of multiple data points as they are merely a mirror of one data point.

In the first two cases cases of insight I happened to stumble across something that pointed me in the right direction to question the outcome in the first place. For Google it was because they scrubbed the free trial detail from the primary article but left it in the meta detail for open graph protocol markup, which was copied from the YouTube Music announcement that included the detail within the article (since deleted).

In the case of iHeart the number seemed so outlandish given it would’ve made them far larger than Spotify or Pandora in the streaming space at the time that I had to go to the app store pages and saw the added total downloads matched to their claim.

In the case of Trump one must question the definition of "numerous sources" and determine if accurate, or if it should be discounted.