Big data in the media industry: Overcoming a preference for Prophets, Pundits, Polemics, and Pedantics.

If you look closely at the use of data—big or small— in the media industry, you find (most likely) a preference for three make that four sorts of interpreters: Prophets, Pundits, Polemics, and Pedantics. For example, this post is itself an example of Pundit-style analytics.

Frankly speaking, this preference for one of four P’s makes the industry no different from societies-at-large. In just about any society, particularly as the underlying reasons for events become obscure, these sorts of analytical heroes surface.

Prophets—people who can see the future and are here to prepare us.

The insight on offer for many industry luminaries is their professed capacity to see or anticipate that which will happen next. Of particular interest to a prophet is truly the next big thing. These sorts of insights usually begin with phrases like, “The future of your industry looks like this….,” or “The DVD market will….,” or “Three dimensions of consumer demand in 2020….” Many people have built their careers on being prophetic once, or  maybe a few times, and being apparently correct.

The use of data by prophets is, well, very prophetic. A few crumbs of data that can be organized in such a way as to provide anecdotal evidence of a trend are all a prophet requires. Take a chart with a line sloping up or down and just extend the line into the future. Hang out with some “generation [blank]-ers” and describe their behavior as not only eventually widespread but also inevitable.

Let’s be clear and honest about data and analytics—neither can predict the future. The combination of the two can only hope to provide you with some better understanding of the past and present. That understanding might be applied towards the future, but only to the extent that the future is an awful lot like the past.

Pundits—people who can see the truth in the past requiring little more than a bar or pie chart.

Pundits are a powerful class of interpreters as they claim to have the ability to impute general trends from wonky data. Wonky data comes in many flavors, such as: rather bland (like a bar or pie chart) or skimpy (like a sample of ten) or truly sparse (like a single case study).

The favorite sorts of phrases from pundits sound something like: “This industry is failing because of X….,” or “This industry has a preference for three….”

Punditry is powerful because it usually taps into some hunch or belief we have in the underlying cause(s) for experience for which we might lack sufficient evidence (aka, confirmatory bias). The pundit provides a sort of social confirmation upon which we can rely to overcome the uncertainty in the data.  Explanations gets more highly weighted alongside support from a authoritative pundit.

The unfortunate truth is that no analytical method, at least none that is firmly stationed within the class of tools know as scientific or by falsification, will ever prove something beyond doubt. In fact, except in the case of certain methods that treat the counterfactual in a particular way, most tools—like the ever-popular regression—are actually setup to simply question the counterfactual (or null hypothesis). As a result, we simply are reasonably certain that the opposite of that premise may not be the case, or that the data seem to conform to some explanation/expectation within some margin of error.

In simple-but-convoluted english: Data and analytics never prove any premise. Get over it.

Polemics—people who only reveal analyses with the “right” conclusion.

Polemics are a powerful class of interpreters and their impact can be seen in the political debates of governments, industries, and organizations. To the polemic, data and analysis may be tools for understanding but are most importantly tools for influence.

So important is this influential dimension of interpretation that data/analysis that is not sufficient to “move the needle” will be discarded or, in some circumstances, massaged.

Rationally, when there are two sides to an issue and both appear to have support built upon decent data and analysis, its probably time to understand why and how the truth may involve a little bit of both sides from the debate.

Pedantics—people who only believe the sort of analyses that cannot exist

Pedantics operate on what is probably the opposite extreme from prophets and pundits as far as their analytical threshold. For the pedantic, pretty much any and every conclusion is lacking in sufficient rigor for a conclusion to be drawn.

Since no analysis that can be performed will every be a perfect analysis, and those trained in scientific (or highbrow, in the minds of some people) methods are highly sensitive to this inadequacy, the pedantic can shut down any analysis with simple comments like: “But what about….,” or “Now how big is this sample and is it sufficiently diverse to have included….,” or “I think we need a little more….”

And so, if we are going to make decent use of data and analytics, we have to accept that any and all methods come with a grain of error and maybe some salt as well. Even with a little salt, however, and because we can take error into account, these insights have value.

Is Apple misrepresenting iCloud (Photo Stream) in the recent commercial?

In the most recent commercial from Apple, in which the firm touts the simplicity of iCloud, users are seen taking photos that instantly appear on computers linked together via “the cloud.”

The only problem with this presentation of iCloud would be that the scenarios in which the photos are being taken—on the beach, in a a snowy park, etc.— would questionably fit the scenarios in which Photo Stream via iCloud will actually work so automagically.  Why, do you ask?

Because Photo Stream is not setup to work over cellular data networks, the sorts of networks to which you are most likely connected when you are at places like the beach, public parks, etc.  While this sort of instant sync would be excellent, it is not exactly possible (at this point) with an iPhone—even apparently if you have an unlimited data account.

According to Apple’s own description inside the iOS setting, Photo Stream “automatically uploads new photos to iCloud and downloads them to all of your devices, when connected to Wi-Fi.”

Alternatively, from Apple’s own support site: ”

On an iOS device, new photos you take will be automatically uploaded to your Photo Stream when you leave the Camera app and are connected to Wi-Fi. Note: Photo Stream does not push photos over cellular connections.

I added the bold highlights because these highlights matter.

So while iCloud does enable calendar items, contacts, and similar sorts of data do update on iClouded machines almost instantly, even over cellular data connections, Photo Stream is different.

Unless the iOS is doing something other than what it says it does, you will not see the same sort of instant gratification presented in the commercial when taking pictures while connected to cellular data networks.  You have to be hanging out on a beach that offers public WiFi. Oh, and wait until after you close the camera app.

Making sense of the mixed methods for calculating Spotify users and conversion rates

Warning: this is a LONG post.

As Spotify continues to grow, both in terms of the number of countries within which it operates and the number of users signing up for the service—(paid or free), I reckon its time to try and make some sense of the “mixed methods” that are being applied to reckon user base and conversion rate estimates and reports.

My final takeaway away will be that the conversion rates that matter should be between the registered user base and the active free user base, as well as the paid user base.  Tossing out registered users from the calculation—as Spotify is doing by reporting metrics based upon active users— is leading to an over-estimate of the allure of these services in the context of the wide range of service options that might exist.

In other words, by reporting conversion rates based upon Active Free users, the music services market appears to be more attractive as presently priced and structured than that market might actually be.

Now, to be clear, Spotify is/are dealing with not only the expectations of license and investor agreements, but also the cliff of free users being, well, un-freed after their six months of free access to the service.  These dynamics might lead to both the reported ratio of paid to free users and the reported number of registered and active users to move around.  Regardless, the numbers that are hitting the news feeds are moving all over and we might as well be honest about how crazy this experience is.

NOTE: a robots.txt file within the certain folders of the Spotify site prevent a more directed inquiry into the “official” user numbers as reported, but comments to the press make a little backtracking possible.

First: Spotify now reveal(s) it’s free-paid conversion rate based upon a report of active monthly users and not registered users. This method is a bit wonky.

As a result of the method through which Spotify communicates conversion, as free users fall off the active user cliff—by being “un-freed” from the service—Spotify’s conversion rate will increase even though their actual conversion rate for users who have signed up for the service has not changed and may even be falling.


(A) I have 20 users of a service, two of which pay for the service while 18 use a free version.  10% of my total user base is paying for the service.

(B) I deny access to the service to 6 of the prior free users, leaving 12 free users.  One of these free dropouts chooses to pay.  I now have 3 paid users and 12 free users, for a total of 15 users.  20% of my total user base is paying for the service.

The conversion rate in this example increased from scenario (A) to (B) even though no newly registered users emerged and the total user base, in fact, fell in number.

Second, as a result of the switchbacks in metrics reporting, explicit or implied conversion rates for Spotify are all over the place.

Most recently, Spotify execs claims there were 3,000,000 paying users (at either of the two paying tiers) and that this user base was 20% of the total “active monthly” user base.  Or, at least this is how Tim Bradshaw at the Financial Times interpreted the numbers as either presented or hinted to him.

If 3,000,000 were 20% of some larger number, that number would most likely be 15,000,000.  Please feel free to check my math.

And so, we have reason to believe that Spotify now claims approximately 15,000,000 monthly active users.  Where active users implies individual accounts that have used the service during the past 30 days.  And their conversion of active users resides at 10%

Back in November of 2011, Spotify claimed 2,500,000 paid subscribers and an active user base of 10,000,000 users.  This stat is repeated in many places, including an interview for Grammy Week. The quote from Ek the CEO:

You’re talking 10 million active users, 2.5 million subscribers — most of them paying $120 a year, which is double the amount of your average iTunes user.

And so, in November of 2011 it looked like 25% of the active user base was paying for the service, a seemingly higher conversion rate.

While back in March of 2011, the company reported 1 million paid subscribers and a mathematically wonderful figure for users as 6.67 million (or 6,666,666), suggesting a 15% conversion rate (or 14.9925….%).

Finally, the conversion rates that really ought matter—if we are to understand the attractiveness of these sorts of music services— would involve a consideration of the registered user base (i.e., everyone who has signed up and tried the service) as compared to the Free and Paid user bases.

The music industry needs to come to terms with the true appeal of BOTH the Free and the Paid versions of these music services.  Focusing nearly exclusively on the Paid conversion of active Free users leads to a misunderstanding of both the purpose and the draw of these services in the market.

Additionally, when conversion to Paid users is measured as a function of active Free users, the limit applied  to the time period for free usage will result in overstating the appeal (i.e., the increase in the conversion rate) of the Paid services as these Free subscribers drop out.  And as these Free subscribers drop out, we lose them to other non-paying options.

As a result, it appears as if services such as Spotify are increasing in their appeal to the general public when, in fact, over time the appeal may have remained the same.

In September of 2010 the firm reported 10 million  “users,” which back then referred to registered users.  At this point in time, there were a reported 500,000 paid subscribers and the country base for official users was limited to the UK, France, Spain, Finland, Netherlands, Norway, and home country Sweden.  Let’s be honest, these numbers implied that 5% of the registered user based paid for the service.

Since 2010, the firm enjoyed not only expansion to additional countries, but also prime promotion through Facebook thanks to a investor family connection.

The number of paid subscribers has grown by a factor of six since the fall of 2010, from 500,000 to a reported 3,000,000.  If the registered user base grew by a similar ratio, a total of 60 million users would have signed up.  Given the tweaks in the free service limitation since initial launch, combined with certain growth trends and metrics along the way, it seems reasonable that the total registered user base grow by 4x over the period.

As far as I can tell, upwards of 40 million (+/- 5 million) people have signed up for a Spotify account since their launch. If I were to guesstimate their actual conversion rate for registered users to paid users that rate would be 7.5% (+/- 1.5%).

And so, when that conversion rate is measured as a function of registered users who have converted to paid users, this uptick in addressable market and promotion may have resulted in only a slight increase in the true conversion rate for the service. Something about these services are leading them to appeal to only a subsection of a subsection of the market.

Will someone please flip the switch that makes YouTube a formal music service

While browsing around YouTube, or while watching avid YT users browse YouTube, it has become obvious to many people that the basic guts of a music service already exist within the site.  You can search for things, make playlists, subscribe to feeds, etc.

In fact, many people (e.g., generation [fill in the blank] for folks who are trying to classify generations by their assumed differences) already make use of YouTube as a sort of music service.  So many licensed music videos are already on the site, whether as part of the VEVO partnership or otherwise.  And these licenses can cover other files already on the site given the consent of the rightsholder.

And YouTube “Free” is far larger in terms of active userbase that Spotify, MOG, Rdio or Deezer “Free.” If we were to include this set of active media consumers within the market estimates of the size of the “music services” market in the US (or abroad) that market would instantly expand by significant multiples.

In other words, we are probably underestimating the size of the active music service market by either (a) overlooking the way in which people are already experiencing music online, or (b) holding back a platform that could further expand this market with only a few tweaks.

And so, will someone please just take us out of our misery and flip the switch that makes YouTube a more direct music service experience.

I know that in order to make that happen a host of licenses would have to be obtained, licenses that would require negotiations throttled by advances and other assurances.

But millions upon millions of humans are now browsing YouTube listening to music.  It would seem to make some sense to offer both a Free and a Paid version of a streaming (and perhaps also portable) music service integrated with and additional to the video content.

Or maybe I am just nuts.