Current Issue:

Current Issue

On The Cover:
Cumulus Chairman Jeff Marcus

Click here to subscribe to Radio Ink.

Radio Ink Writers

The Fuzzy Math Continues Online


Pandora has just released its August 2012 metrics and, like the previous two months, the numbers are up. That goes against everything we've been told by Triton, which reported all audio listening is down in the summer (through July). Pandora reports "listener hours" for August were 1.16 billion, compared to July which were 1.12 billion.

Pandora does not detail what a "listener hour" is in its report. Triton reported that Pandora had 1.413 million "average active sessions" in July and 1.421 in June. August has not been released as of yet. Pandora again claims it has 6.3 percent of all radio listening, although it provides no detail about how they come to that conclusion.

Pandora also claims active listeners are 56.2 million for August, an increase from 54.9 million in July and 54.5 million in June.

(9/11/2012 2:12:19 PM)
Mary Beth:

Thanks for the nod to my having a grasp on this topic, and for that swing at my head - challenging discourse is always welcome.

My comments were to address the differentials in what was being discussed in the article, that the stats could not be compared.

These comments will only partially address what you reference because, as you know, it takes a whole lot of writing to get a point across in this new convoluted world of radio measurement.

My writings have shown that I believe there is no such things as an "online radio universe," and that it cannot be calculated. Too many stations run by "enthusiastic amateurs" exist, and the landscape changes daily. Too many major players are excluded in the Triton Digital Webcast Metrics.

I've also stated that Pandora has its methodologies, in part coming from some well-known and respected radio research firm (the name, which I don't have time to look up). Please, correct this if in error.

The following two items are pulled from Pandora server logs. The are fact:
1) Listener hours for Pandora during the month of August 2012 were 1.16 billion...
2) Active listeners were 56.2 million at the end of August 2012...

I am in agreement with you (Mary Beth) that calculating the "Share of total U.S. radio listening for Pandora in August 2012 was 6.30%" is an impossible task, and cannot be done without educated guess-work. But educated guesswork is what the entire radio industry is based on; only its numbers come from Arbitron.

Where we disagree is in a need for Pandora to divulge proprietary methods that any radio group would play close to its chest.

If there's such a call to Pandora, we need to have equal call to Arbitron on its breakout to "prove" listener levels are at reported amounts. Not an extrapolation of weighted surveys and PPM stats - all of which fall under the cry of "...considering that Pandora cannot prove there was actually a person listening for every fraction of a minute they count, their numbers remain fuzzy math."

Fuzzy math can be created by anyone attempting to demonize a competitor. Give me a data set and I can prove anything correct or wrong, if I'm talking to people who don't know how to read the data. And that was the essence of what I was saying in my original comment to this article - we cannot continue this apples to oranges comparison without educating readers that claims like this are without basis.

- Ken Dardis
(9/11/2012 10:02:34 AM)
To be fair, Ken Dardis does know what he's talking about. As Triton itself notes, the press release data from Pandora is based on "internal" data, not anything at all from Triton. And since Triton doesn't count anything that occurs for less than a minute, the numbers would never match anyway.
What Ken doesn't address -- that everyone seems to think is the elephant in the room and insists on ignoring -- is that Pandora cannot and will not substantiate its claims of "share of radio listening" that are usually a part of their "based on internal data" press releases. Even when challenged by Radio Ink last month, their steadfast reply was "Based on Arbitron (to which they do not subscribe), Triton (which notes that its ratings have nothing to do with Pandora's internal numbers) and the Census (which doesn't measure time spent listening to radio, satellite radio and internet radio). Since total listening to satellite radio isn't reported anywhere by any nationally syndicated ratings service (and since Sirius XM has no physical way of monitoring and reporting TSL), I wonder how Pandora has generated that data. Since Arbitron doesn't publish total time spent listening to radio with online listening extracted, and since Pandora doesn't subscribe to Arbitron anyway, I have no idea where they are getting that data. There also is nothing that measures total online listening. Triton (whose ratings for Pandora do not match those of Pandora) doesn't measure Spotify, lastFM, Rdio, Songza or myriad other online music streaming sites. The only way to determine total online listening is to get the files from each of those services – something I doubt they would surrender to Pandora. So -- what is the basis of their claim and why won't they share the simple formula they must use to concoct that number?
Ken, considering that Pandora cannot prove there was actually a person listening for every fraction of a minute they count, their numbers remain fuzzy math. And it is fuzzy math if you take it from an analyst's view, wanting to know how they can monetize those numbers. You'd want to know what portion of those hours were used by premium users who do not get advertising, yet pay on $36 a year for the service. Since the paying users who get no advertising are most likely to account for a larger-than-average share of Pandora’s total listening, the only way to monetize that listening is through subscription fees. Somehow I think $36 a year would not begin to cover the royalty fees paid out for these people -- and all the money spent on hiring local ad sellers isn't going to make those listener hours generate revenue.
So, until Pandora releases the actual components (and their sources)of the formula it uses to concoct its claims, and until they differentiate between paid and non-paid usage (something, BTW, even Triton doesn't do -- their monthly numbers include the premium listening), then Radio Ink is absolutely correct -- Pandora is, at best, using fuzzy math.

- Mary Beth Garber
(9/11/2012 8:30:11 AM)
All that's "fuzzy" here is the author's inability to understand they are mixing apples and oranges.

"[A]ll audio listening is down in the summer..." is true, as in "Avg-Active Sessions," "Session Starts," and even (slightly) for "Time Spent Listening," which is all that Triton Digital Webcast Metrics offer.

Stop and analyze the author's statement though, how "Pandora reports "listener hours" for August were 1.16 billion, compared to July which were 1.12 billion."

That which Pandora reports is not connected to that which is reported by Triton - even remotely. They are two different metrics that can exist simultaneously.

One more item, dealing with the claim "Triton reported that Pandora had 1.413 million "average active sessions" in July and 1.421 in June." These are Monday-Friday numbers.

"Monday-Sunday," Pandora had a decrease of only 1,785 persons in "Average Active Session," June-to-July. It had an increase of 48,258,477 "Session Starts" in that same time. 62.29% of Pandora's "Session Starts" occur on the weekend.

This author clearly has a lot to learn about reading online numbers, and on being fair when discussion radio's competition.

- Ken Dardis

Add a Comment | View All Comments


Send This Story To A Friend