How Himalaya Should Spend the $100 Million
They didn’t ask me but here‘s what I think anyway.
I had to reread the headline at least a couple of times:
Podcast Platform Himalaya Raises $100 Million, Launches Apps With Tipping Function
$100 million? What on earth is Himalaya going to do with all that money? Besides, of course, the oddly headline-worthy ‘tipping function’? Then it occurred to me: The Oprah and LeBron Show. The two stars would richly deserve that money just so long as their deal includes three important words: Only on Himalaya. At that point, Himalaya is only two tweets away from over 80 million high engagement Twitter impressions. And that’s the whole game, of course. Content really is king. If there is any doubt about that just ask Netflix.
On the other hand if all that’s delivered for the $100 million is exclusive distribution deals with those who need no further introduction other than their first name, that will be a crying shame indeed.
The good folks at Himalaya didn’t ask me but here are my thoughts as to what they should bring to the market with all that money. My qualifications for this, you ask? Virtually none other than I produce two podcasts. That, and I have noticed something really important missing from current podcast services which I simply cannot believe wasn’t delivered ages ago by at least one of the major podcast platforms.
Nobody can tell me how many people are actually listening.
The nearest analogue is the widely referenced, industry standard ‘download count’. It’s virtually a meaningless number beyond the broad, qualitative characterization of the audience size. Technically a download is nothing more than moving a digital file from one place to another — such as from the podcast hosting service to your phone. That’s it. That’s all you know for sure. Beyond that is anybody’s guess.
Even when a blockbuster like NPR’s Serial podcast has been downloaded over 300 million times, there is still only one rather useless fact known with 100% accuracy: one file or another was moved from one location to another that many times. Whether that file was ever been decoded by a piece of software to generate audible sounds to which one or more pairs of ears listens — that’s an entirely different matter. With Serial, if just one percent of the downloads never gets played for whatever reason, that means the audience has been over estimated by three million. Think that’s bad? If one listener downloads all 32 episodes of Serial, that counts as 32 downloads, not the one listener it really represents.
Ugh.
Serial has a huge audience without a doubt. But exactly how huge? Nobody knows.
Of course, the same goes for the microscopic-by-comparison audience for my podcasts. I think I have a certain number of listeners based on the download counts for each episode. But how many pairs of ears have actually listened to an episode and for how long? I don’t have a clue. Nobody does.
For now, the plates all stay in the air so long as advertisers are happy with the gross analogue of download count to audience size. Thank goodness not too many of them have yet to start asking that nasty question “how many pairs of ears and for how long?” Not to mention that other, even nastier question “and did they just skip over my ads?”
Suddenly images of dominos falling, houses of cards collapsing and steam locomotives colliding head on spontaneously pop to mind.
This is an easily solved problem. Before I get to that, however, a little background. Electronic payment companies have, for years, used something called device fingerprinting as a means of uniquely identifying the remote device with which they are communicating. It’s just like it sounds: examine various characteristics of the remote device — a phone for example — apply an algorithm to these characteristics and out pops a unique identifier.
For the payment companies alarm bells go off, strobes flash, sirens wail and the CFO’s phone rings in the middle of the night when somebody claims to be you and that pesky device fingerprint doesn’t match. While it’s possible you upgraded your phone since the last transaction, which means a new device fingerprint of course, it is much more likely it’s some miscreant just pretending to be you. That miscreant will go to great pains to look and behave remarkably like you in every way until, in the end, the device fingerprint doesn’t match. At that point, the jig’s up. Payment declined.
The term device fingerprint is somewhat of a misnomer. It’s nowhere near as deterministic as a human fingerprint. However, if it’s sufficient to deter fraud in the electronic payments industry with its low tolerance for errors, it really should be more than sufficient to uniquely identify a device within the Himalaya ecosystem.
So let’s assume Himalaya implements some version of device fingerprinting in order to establish how many unique devices have contacted them one way or another. Let’s also further assume just one pair of ears is associated with each of those unique devices. That’s very accurate for any sort of mobile device — unless you are in the habit of sharing your phone with someone? — and somewhat less accurate for devices primarily intended to be played through-the-air like the Apple HomePod or the Amazon Echo. But let’s go with the one pair of ears per device assumption for the time being.
But, wait a sec, what about the rest of the person attached to those ears? Let’s just say, for the time being, the less we know about them the better. I’ll explain more in a moment.
The only other thing Himalaya has to do is keep track of how a given podcast episode is consumed within their ecosystem. The base unit of such a system is vector-like: at what hour, minute and second within a given episode did a unique pair of ears start listening, and for how many seconds did they continue listening. The vast majority, of course, will start at the beginning and then play the episode until such time that they don’t want to listen any more. That activity can be captured in exactly one ‘listening vector’, as I’ll call them. If the listener starts and stops an episode three times, that’s three listening vectors Himalaya needs to capture. So long as exactly one listening vector is captured for each start-and-listen event, Himalaya will have all the information it requires about listening behaviour.
Before I go any further, it’s worth mentioning that capturing these listening vectors is a cinch if the podcast is being streamed directly from Himalaya — everything is in one place. But what about all those mobile devices which have downloaded episodes from Himalaya for offline listening? This is only slightly more complicated. Simply keep the episode listening vectors on the mobile device until such time that device makes contact with Himalaya again. Upload the listening vectors at that time. With that, Himalaya has a complete listening history for that unique device and the ears associated with it. In the event the device never makes contact with Himalaya again then, yes, you lose the listening vectors stored on that never-heard-from-again phone or tablet. Whatever small error this introduces, it tends to underestimate the audience as opposed to the other way around. That’s better from an audience measurement perspective. It’s also easy to quantify the error related to the missing information. It’s plus or minus one pair of ears in each such case.
So with the combination of device fingerprinting and a complete database of listening vectors what does Himalaya have? After some trivial processing, nothing less than the ability to know, within fractions of a percentage point, exactly how many pairs of ears have listened to each individual second of any podcast episode.
I thought that might get your attention.
From this treasure trove of data everything else is built.
Most important to me, the podcaster, Himalaya can provide data which shows me how many unique pairs of ears have listened to each individual second for each of my episodes distributed through their service. Of course, what they should also do is plot that out for me in a variety of ways so I’ll be able to easily visualize the parts of my episodes which work and the ones which don’t. Ideally, Himalaya will also give me the ability to slice, dice, aggregate or drill down into the data at will. At the very least, it would be great if they let me download the raw data for these kinds of machinations.
However, in order to find the party with the most to gain or lose, follow the money which inevitably leads you to the podcast advertiser. They would have access to the same kind of analysis but in their case they will care mostly about the portions of the podcast episodes where their ad spots are located. They’ll be able to determine second-by-second how their ads performed.
This level of detailed analysis has knock on effects to other capabilities. What the advertiser pays, for example, should be related to the precisely measurable audience size for their spots. Put an ad on The Tim Ferriss Show and you pay a lot per ad. Put an ad on one of my podcasts — please! — and, well, let’s just say you pay a whole lot less. But the price per unique set of ears will actually be the same for either Tim’s podcast or mine. The advertiser simply has to make a decision as to the number of ears they can afford to attract for a given ad and the most effective content in which to embed their campaigns.
This would also enable additional features for which the advertiser might be willing to pay a premium. How about the ability to disable the fast forward button during their ad spots? If that’s a little too on-the-nose for either the advertiser or the audience or both, how about the ability to move the listener right to the offer code part of the ad when they hit the fast forward button and then continue with the podcast content with another click of the fast forward button?
This approach also enables Himalaya to sculpt the listening experience based on how the pair of ears wants to pay for their episodes. If they want to pay for a Himalaya subscription, the ads are automatically skipped when the subscriber plays that episode. As if they were never there. If they are listening ‘for free’, then they get the whole episode with the ads included. Of course, this capability would be tailorable by the individual podcast. If Malcolm Gladwell wants his fabulous, and fabulously popular, podcasts available to subscribers only, that would be an option for him. He could even further tailor it so the first episode is free, no ads, and for the rest the audience would pay per episode but only if they listen to more than, say, the first three minutes. Or perhaps just have the listener pay for the total number of seconds to which they have actually listened to any one of Gladwell’s episodes.
Don’t worry, Himalaya, the cost of implementing all of this will be little more than a rounding error for LeBron and Oprah. In fact, I’m almost sure they won’t even notice.
So what about the rest of the person attached to those wonderfully monetizable set of ears? Don’t we care about that person? Of course we do. But here’s an unassailable fact standing in the way: every piece of additional information we collect about those ears beyond what is absolutely necessary to identify them as unique, is one piece of information which, sooner-or-later, somebody is going to object to Himalaya collecting. In other words, it becomes a privacy problem and the world is becoming justifiably more concerned about that all the time.
So, Himalaya, do the device fingerprint thing to support your audience measurement objectives but absolutely no more than that. Accept the fact there are some tantalizing demographics which will simply be out of reach in this proposed paradigm. The Facebook of podcasting is exactly the opposite of what you want to be. Moreover, you will do well to be totally transparent about what information is used in your fingerprinting algorithm stressing the fact it contains no personally identifiable information. You may even want to go one step further and publish it as an open standard adoptable by any and all stakeholders including your competition. All the users of that algorithm would know is that some pair of ears out there likes to binge listen to episodes of The Oprah and LeBron Show.
Remember, even if you do all that, you will still be able to tell your advertisers, with great certainty, how many pairs of ears have been listening to each individual second of each episode and that’s an undeniably valuable proposition in its own right.
Also remember, if and when those anonymous ears decide to create a Himalaya account, they will first be given the opportunity to read your very simple and easily understood Terms of Service. These will clearly state what additional information Himalaya intends to collect and what it deems acceptable uses of that information. If that is indeed acceptable, the anonymous pair of ears can come in from the cold and identify the person attached to them. On the other hand, if a given pair of ears does not agree to those Terms of Service, no account would be setup and end of story. That is, other than the ears will remain anonymous which is probably the way they want it.
John Wanamaker, the American retail and marketing pioneer, was famously credited with the saying “half the money I spend on advertising is wasted; the trouble is I don’t know which half.” If Wanamaker were alive today, I wonder if he would update that for advertising in podcasts. Something along the lines of “some of the money I spend on [podcast] advertising is wasted; the trouble is I don’t have a flying clue how much. Maybe all of it.”
However, this has nothing to do with the suitability of podcasts as a medium for advertising. In fact, I would anecdotally argue the exact opposite. It’s a fantastic medium for advertising. It’s way more engaging than music and is available in all sorts of places — like moving cars, for example — where video is simply not an option.
All podcasting lacks is a tool for the accurate measurement of the audience.
But finally back to The LeBron and Oprah Show for a moment. Before Himalaya were to sign these superstar, surefire audience magnets to a multi-megabuck, Only on Himalaya deal, doesn’t everybody want to know precisely what is being paid for the attention of that massive audience? Not the least of whom Oprah and LeBron?
I’m quite sure they do. That’s why, with the features described here, they would want to be Only on Himalaya in the first place.
©2019 Terence C. Gannon
Thank you so much for reading. You can also listen to this essay an episode on the Not There Yet podcast, read by the author. Sincere thanks to my friend and colleague Tim Beck for vetting the technical assertions of this essay and helping to rein in the loopier ones.
