Interview Transcript

Disclaimer: This interview is for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise.

Given your experience, I thought we could use Megaphone as an example to discuss how a top podcast would be monetized.  

Yes, there are several opportunities, but programmatic might be the biggest one, in my opinion. I don't see how the podcast world will evolve differently from the digital media world. That's why I believe programmatic will be so crucial, especially for brand advertisers that need to generate impressions and awareness. Buying and monitoring 100 shows is challenging unless you're buying programmatically or through the Spotify Audience Network. The Spotify Audience Network is a line item on a media plan. It's easy to buy, measure, and determine if everything was delivered. That's where I see a lot of the opportunity. Host-read advertising will continue to grow because it's effective and measurable with ROI. Unlike most direct response advertising, it is smaller scale but very effective and has a high CPM. Most direct response advertising is massive scale with a low CPM. Podcasting continues to work because it's such a personal experience. However, brands will only buy part of their campaign this way, in my opinion. They'll purchase a lot through audience targeting. But we can go through these questions.

Could you start by describing a typical Megaphone podcast? What does it look like in terms of monthly listens, ad spend, etc?

It's hard to define a typical podcast on Megaphone. Megaphone targeted enterprise players on a large scale, like Fox, which has multiple podcasts. Some of them are big, and some are small. So, they're all included.

Was there a minimum number of listens that you required?

There isn't a minimum listening requirement to participate with Megaphone. However, to secure a special deal, perhaps a lower CPM against download, you need to have scale. If you're a smaller participant, you might pay $100 a month, and it may not necessarily be beneficial for you to participate, as your program might not generate substantial revenue depending on its size. Our focus, especially during my tenure, was on the major players in the space. It's the 80-20 rule. That's where the revenue is. We aimed for scale, and often, scale is the aggregate of a network.

On SPAN today, you have Megaphone, Spotify Originals, Exclusives, and some of the top Anchor programs. How does one get into that network?

For Anchor programs, scale and brand safety would be considered. At one point, it was the top 200 shows on anchor, but I'm not sure if it has expanded beyond that. As for Megaphone, it would include all the enterprise players out there, from Fox to CNN and Warner, to ad networks like AdLarge and Westwood One. iHeart was on there for several years. They purchased their own platform and eventually migrated over, but they forfeited a lot of revenue in doing so. Then there are their owned and operated programs. If you think of Spotify, they have two sides. There's Megaphone, which primarily seeks advertising inventory and works with the big players, and then there's their content. Joe Rogan brings more than just ad impressions. It brings monthly average active users to Spotify, among other things. They place those on Megaphone and can fill the unsold inventory. Spotify does not sell the shows on Megaphone directly. They only sell their owned and operated programs directly. Programs like Joe Rogan and Call Her Daddy have the scale to do that.

What's the average number of ad slots on Megaphone inventory?

I can't provide specific information about Megaphone, but I can tell you how we used to view it. The publishers of the shows decide the number of ad slots. They can include as many as they want. The early recommended theory was two pre-rolls and a mid-roll for every 15 minutes of programming.

There are platforms like iHeart that insert many more ads in their mid-rolls than the formula I just shared with you. Spotify, for instance, doesn't sell post-roll for its Spotify Audience Network. It's quite challenging to sell post-roll as direct response advertisers aren't fond of it. In most cases, post-roll is sold programmatically because it can go for a low CPM and provides additional revenue to the show. I believe there will be pressure to insert more advertising into programming due to the expected downward pressure on CPM and increased pressure on volume. Spotify has already moved quickly in this direction. However, the challenge is that ads for direct response advertisers don't perform as well when they're cluttered with other ads. These direct response advertisers still account for a large portion of the purchased inventory.

What's the difference in the CPMs between pre and mid-roll, typically?  

At Spotify Audience Network, at least when I was there, it was public knowledge that we sold the pre-roll at the same CPM as we sold the mid-roll. We didn't differentiate. The DTCs are focused on the mid-roll. They don't like to buy a lot of the pre-roll brand advertisements because they get 60 seconds to communicate their message if they do mid-roll. Pre-roll tends to be 15 to 30 seconds in length.

Why is that?

You don't want to deter the listener from walking away from the show. If there are two to three minutes of advertising at the beginning of the show, people are likely to skip it or the show altogether. That's the theory, and there's probably some data to support that.

How do you segment the ad formats today?  

If you consider the shows on Megaphone, the partners are selling direct. Take, for example, the William Barnes podcast, you would be selling directly, probably to the DTC shops. The CPM would be based on performance. There's a lot of pressure for a low CPM because if you have a lower CPM, it would theoretically perform better in terms of response. But those CPMs overall can range anywhere from $20 to $35 if you're selling direct. If you're a special show, you can command more than that.

And that's host-read?

When I sold Malcolm Gladwell's show directly many years ago, we sold out all the inventory at a $75 CPM, and it was all direct sold advertising.

So you're going to an advertiser and saying, we've got these slots on Malcolm Gladwell's show. This is the price. Do you want to buy it? And that's guaranteed inventory?

Yes, that would be guaranteed if you're selling direct. Malcolm Gladwell is a bit of an exception because he does the host read. You get Malcolm Gladwell, who some consider one of the most important thought leaders of today, voicing your brand or product. There's a great deal of value in that.

Is that permanently on the file? For instance, is Malcolm Gladwell’s voice pre-recorded and then inserted into the WAV or the MP3, and is that permanently fixed?

A show could include what we call "baked in" advertising. It's not the best term, but with Malcolm Gladwell, we never used baked in advertising. We guarantee you one million impressions at a $75 CPM, or $75,000. We'll deliver that and then remove the ad. Some sellers and publishers still use baked in advertising. Direct-to-consumer advertisers prefer this because if there's a recurring event, like the William Barnes Christmas special, it might resurface next Christmas. Even though the offer might be different, you're still visiting that site and making a sale.

What's the current ratio between baked in and dynamically inserted host-read ads?

Direct-to-consumer advertisers continue to push for baked in ads. I'm not sure of the exact percentage. Most partners are selling dynamically inserted ads because it's more practical. If you're on your favorite website, you're not seeing baked in ads. If you're reading the Financial Times, those ads are constantly rotating. I would say many of the smaller players trying to generate some revenue still use baked in ads. There are platforms like AudioBoom that primarily focus on baked in ads, but they're trying to transition more towards dynamic ads. Direct-to-consumer advertisers prefer baked in ads, but they're still buying based on impressions from dynamically inserted ads. This could affect the CPM. If the baked in ads are performing better, the CPM could be higher. The issue is that you don't have all that back inventory to sell.

If Malcolm Gladwell voices the ad, it's dynamically inserted. Let's say it's season six of Revisionist History. This will boost seasons one through five because it draws more attention. You get new listeners who will go back and listen to previous episodes. So you have that inventory for sale. There's also the Spotify Audience Network. If you're on Megaphone, which Malcolm Gladwell is no longer on because he works with iHeart, but hypothetically, the Spotify Audience Network would fill that back inventory. If Malcolm Gladwell had a relationship with programmatic players, that would be the third tier. So it goes from direct sold, to the Megaphone setup, to the Spotify Audience Network, and then there's still available inventory. You could use programmatic hashtags and try to sell some of that remaining inventory.

Typically, how many of the top shows on Megaphone are sold directly by direct sales?

The most popular shows with the best hosts that are brand safe are likely to sell the highest, probably more than 50%. I would wager that a show like Joe Rogan is 75% sold out directly, and Malcolm Gladwell's show is also likely 75% sold out. The question is, why isn't it 100%? Selling 100% is challenging due to the timing of the advertising. Campaigns don't always align. You have a fixed number of ads in the show, and the timing of ad campaigns, like Malcolm Gladwell's, may not coincide. For instance, one campaign might want to start on November 15, but there's still some inventory left from a campaign that ends on November 1.

Aligning the inventory with the advertising campaign is a challenge in achieving 100% sales.

Yes, it's been quite tricky in the podcast space.

Let's take Malcolm Gladwell as an example. Suppose he's launching a new season of Revisionist History. How do you estimate and sell the number of impressions to advertisers?

You can project based on past seasons and the available inventory. You also have the opportunity of the back catalog. However, you don't always get it right. A significant challenge in advertising arises when a new program is launched, especially a limited series. It's tough to sell, particularly the first season, because you have no idea how it will perform. You can model it after a similar show, but it could either take off or flop. If it flops, you have a lot of work to do with the advertiser to rectify the situation.

But Malcolm Gladwell would have a stable average monthly listenership, right? So, you would have an idea of the impressions.

Yes, with dynamic ads, you can project within a range. The same applies to Joe Rogan. If you're getting more downloads or impressions than anticipated, Spotify Audience Network or programmatic could fill that. I've seen shows that didn't perform well initially, but something happened in the news, and they exploded. There's no way you could have sold that. So, it's all programmatic, Megaphone Targeted Marketplace, or something similar, filling it.

If he's generating, let's say, two million impressions a month, how much would you try to sell?

You would try to sell as much as you possibly can. Ideally, you'd want to hit the nail right on the head. But if you're going to make a mistake, it's better to oversell than undersell. And then deal with it by extending the campaign another week to deliver against it, or repurposing the leftover budget. For instance, if you've got $10,000 left over, could we repurpose that somewhere else in a show?

What does a host like Malcolm care about? 

Most hosts, especially those of Malcolm's caliber, have the ability to reject advertisers. For instance, Malcolm might reject oil and gas companies. If you lean more towards the liberal side, you're likely to reject oil and gas. However, if you're hosting a show like Rush Limbaugh's, you might accept oil and gas.

Typically, there are guidelines stipulating that you can't reject more than 20% of the advertisers we bring to you, or 20% of the categories if you're making any sort of guarantee in the space. This isn't always the case, but it's something we try to manage.

What's Megaphone's take rate for direct sold ads?

Megaphone doesn't directly sell anything. It's all handled by Spotify Audience Network. If I'm hosting the William Barnes show, I'm selling directly, and I use Megaphone's platform because I consider it the best. It also brings me a significant amount of additional revenue.

Megaphone doesn't monetize any of the direct sales that Malcolm Gladwell or I would do for our podcasts?

No, they do not. They don't take a share. They just want inventory to sell against audience at scale. Now, for something like the Joe Rogan show, which is under Spotify, they have a few of their shows that they sell directly because they're really big. They can get them in front of the big brands.

But Joe Rogan probably gets a cut of that. I'd imagine he has some deal with them where he's likely to get a cut of what gets sold there.

Yes, if I had a podcast on Megaphone and launched it tomorrow, they wouldn't sell me direct. I would just get a share of the Spotify Audience Network. I apologize for the confusion.

Let's say for my podcast, I use Megaphone. I would then say, okay, I've sold 50% of my ad slots directly to X DTC brand. I pre-record my host-reads. I would then have to inform Megaphone or SPAN in Megaphone about the available slots and the CPM I want.

Actually, you don't share the CPM or the available slots. Megaphone can see the slots available and will deliver in open holes. Now, if you sold to Coca Cola, you may block Pepsi or the entire drink category because you don't want a competitor in there with the show that you sold direct, or you'll have the advertiser complaining. Megaphone will fill as much of that inventory as they can, and you'll receive a check at the end of the month.

But if I'm a big host, like you said, following the 80-20 rule, most of the revenue comes from the big hosts. Ideally, I would want to sell everything directly.

Absolutely, that's the goal. However, it's quite challenging to achieve. I've rarely seen anything sold out. There's usually inventory available for any host. Yes, you want to sell direct because you're not doing any sort of revenue share. If you sold direct at $25, it's better than sharing a $40 CPM on the Spotify Audience Network.

How would you compare a big creator on YouTube to this? YouTube takes a cut of everything. The rate might be different, but how do you compare the two mediums?

It's really not comparable. You can't sell directly on YouTube, although you can sell baked-in ads on YouTube. The host could read something into the show and get paid for it.

But YouTube is still going to sell ads on top of that, correct?

Yes, and then you share the revenue with YouTube. I believe it's a 50-50 split. I think most of the big players that are audio-based are most concerned with what YouTube is doing in the space. However, if you're an audio creator or publisher, you might think, "If I put my show on YouTube, I can make some additional revenue."

What are they concerned about on YouTube?

The scale that YouTube brings is significant. Video is becoming more important to many podcasters, especially as celebrities and thought leaders move into the space. I predict Spotify will start to play more in the video game, as will other players like Art19, Wondery, and Amazon. It's a bit more challenging for iHeart and Sirius, who don't yet have a proprietary video outlet. But that's going to become more important. I think user behavior will expand the number of people listening or viewing podcasts because, although podcasts are successful because they're audio and intimate, video excites people because they get to see themselves on screen.

Going back to the waterfall, how do you see the percentage of inventory sold directly by the top shows changing in the next five to 10 years?

First, consider Spotify as an example. There's seasonality to the advertising because they're working with many digital advertisers. December is a strong month for advertising, while January tends to be soft as the new year kicks off. When I was selling the Megaphone targeted marketplace, I would have to warn my big partners not to compare January to December, but rather compare January to the previous January. You're up 100% year over year, but you're down from December. That doesn't mean there's a downward trend in advertising. It's just seasonality.

Let's say you're Malcolm Gladwell or within the top 20% of global podcasts, in 10 years, will you be selling more or less direct ads off audience networks?  

I think something like the Spotify audience network fills between 25% and 50% of the inventory they can get their hands on, depending on the time of year, what's blocked, and competitive separation. There's a lot that goes into it, from the outside looking in. I believe it will be direct sold. But if you look at the digital marketplace for digital advertising as a whole, the last I checked, 87% of digital advertising was sold against audience or programmatically. I think podcasting is going to edge its way to those numbers at some point.

However, they lack hosts. They don't have the host-listener dynamic, right? This is somewhat present on platforms like YouTube. But my question is, how much power does the host have to prevent the program from moving to programmatic?

I don't know if you've read Bryan Barletta and Sounds Profitable, it's worthwhile free content. He makes a point that I've always agreed with. If we're selling programmatic, we should strive to keep those CPMs high. We should not follow the path of digital advertising, which saw a significant drop in CPMs, putting a lot of people out of business.

You have the DTC shops and brand advertising. There's $140 billion being spent in the US on digital advertising, compared to less than $2 billion in podcasting. The goal is to tap into that $140 billion. To do that, we need to understand how brand advertisers buy, what they want, and what's going to be efficient and effective for them. Programmatic audience at scale, along with host reads and unique programs, can be a solution.

If you're a brand advertiser, you need to buy unique programs that will stand out. Maybe working with a host that aligns with your brand could be beneficial.

But why would Joe Rogan or Malcolm Gladwell allow Megaphone or Spotify to monetize their inventory when they have the power? It's not like the Financial Times linking up their website to Google.

I think the answer to that is, although I don't know what Joe Rogan's deal looks like, they've guaranteed Joe Rogan a certain amount of revenue.

Yes, I mean, he's a different case. Let's consider a top 20% podcast, globally.

There are still a lot of deals. The minimum guarantees that Spotify, Wondery, Sirius, iHeart, and to a certain extent, Acast, are offering are substantial. If you're giving a minimum guarantee, in many cases enriching the host, you're saying, "Here's how I've got to monetize this if you want this big paycheck." So there's a lot of that out there. But yes, hosts, and I don't think a host like Joe Rogan is going to be displaced by programmatic. He's got a big show. He's powerful.

That's my point.

But there are, what, four million podcasts? I hear different numbers. Let's use a third party to guarantee it's brand safe and let's go for volume against audience.

If I understand correctly, the top 20% of podcasts, dominated by hosts, will always want to capture a significant part of the inventory host-read direct because they capture all of the dollars, right?

Yes.

Arguably, the real opportunity then is programmatic, but mainly for the rest of the inventory excluding the top podcasts. It's the rest of the inventory after direct sales, which could be, let's say, 50%, I don't know what it is. And then there's the long tail of podcasts, where the opportunity lies. The four million, or rather the 3.2 million, the 80% of the long tail that's not being monetized yet.  

As I mentioned earlier, the super-premium shows likely won't participate in the programmatic space if they're associated with Megaphone. However, they might participate in the Spotify Audience Network because it allows for control over the CPMs, which tend to be relatively high. This doesn't negatively affect the high CPMs obtained through direct selling. There's a significant amount of inventory out there, including this elite group of programs.

On average, what percentage of inventory on Megaphone is sold directly by the host?

It's approximately 30%.

30% of the ad inventory?

Yes, 30% of the ad inventory. I say that because Acast recently released some numbers indicating a 30% sell-through. I'm basing my estimate on those figures. I don't believe it's much different anywhere else.

But that sell-through is through their audience network, on Acast?

Acast's figure includes everything. I estimate about 30% because when I consider Megaphone, they work with many enterprise players who have large sales teams or network sales.

I thought that figure was quite low, not high. Do you consider that high or low?

It's probably higher than most. A network like Fox might be higher, there could be another network that's lower. So, I'm comfortable with that number.

Let's assume 30% is sold directly by the host. They go out, sell the impressions, and so on. So, does that mean 70% of that inventory then goes to Megaphone and SPAN and can be sold via SPAN?

Yes.

How does that work?

SPAN targets specific audiences. If it identifies the right listener, it'll deliver an ad to them. As I mentioned earlier, they fill between maybe 25% and 50% of that remaining inventory.

Roughly 35% of the total ad sales, at most, can be sold programmatically at the bottom of the waterfall?

Yes. And Spotify doesn't do that. The publishers themselves would have to set that up. Or if they're a network, for instance AdLarge, they have a relationship with several programmatic players and they fill the inventory after SPAN.

Why doesn't Spotify do that?

It's only recently that we've been able to utilize VAST technology to access that inventory. Many of the agreements made previously wouldn't have included it. I wouldn't be surprised if Spotify fills that inventory in the future. If I were Spotify, I would aim to tap into that inventory in future deals. However, the market is still evolving, and programmatic hasn't been a major player until recently. It's growing rapidly, but from a small starting point. I would suggest keeping an eye on this in the future. I believe these partners will want to fill that inventory now. The Spotify Audience Network fills at a higher CPM and at a very good rate. So, if you're hooked on it, you'd want to maintain that.

What's the average CPM on SPAN that you mentioned?

It probably ranges from $25 to $35.

How is SPAN targeted?

SPAN targets against the audience. It uses Nielsen data, ingests it, and works to pinpoint the audience. It's not perfect, nothing in targeting, especially in podcasting, is precise, but I believe it's the most precise targeting in podcasting. It uses the IP address, which typically has an average of seven users, but with recency and frequency, you're able to narrow that down. Most other targeting is done by modeling, using similar data, but it's not as precise.

Why can't Acast use IP addresses?

They can, but it's expensive and requires a lot of work. Acast's technology is probably one of the ones I would consider outside the top four. They use AdsWizz, so they would use AdsWizz targeting, which I believe is model-based or contextual-based, so it's not as precise. They could build it out, but it would take a lot of work.

For Malcolm Gladwell's podcast, for example, the inventory moves from Megaphone to SPAN. Then every impression served detects the IP address and the audience, and checks if it can match an advertising campaign. What type of advertisers are looking to buy via SPAN?  

Spotify has been able to use its large, global sales team to sell this. If you think about Spotify, they're probably selling an audio buy, which is programming first. For a $5 million investment advertiser, the podcast portion might be $500,000. Media buyers are under so much pressure that they probably don't study the targeting methods of any of these partners, in most cases. They're likely thinking, if I negotiated hard against that overall audio buy and if the SPAN is at a $30 CPM, I'll let that one go. So, a powerful team like that can generate strong revenue for the Spotify Audience Network and their publishing partners.

Are these ads served by dynamic insertion, or is it streaming insertion?

As far as I know, Spotify is the only one with access to their own inventory for streaming. That might have changed, but they can see if an ad is streamed through. With downloads, it's a bit more complicated. According to IABv2, an ad is considered listened to if it's played for 60 seconds within a 24-hour period. But it's hard to confirm if the ad was actually served.

The impressions you sell directly and via SPAN are based on downloads. With dynamic insertion, you can't really tell.

That's the methodology everyone uses. It's not perfect, but it's what's available. So, yes, you could be buying an impression that never even had the chance of being listened to. If you stream, you at least know that the ad is streamed through. So there's a much greater likelihood that it's been listened to.

You can capture a higher CPM, but inventory falls off. The longer a show goes on, the more inventory falls off. For example, the fourth mid-roll may not have been streamed or listened to. It's complex because most people on the buying end don't understand this.

Streaming insertion is only on Spotify inventory. So, Original, Exclusives, and all of the hosted stuff on Spotify.

They're almost the only ones that can do it because they've got the app. Now, Apple Podcast could tell you if it was streamed through and you can go to their metrics and data and look at it, but you can't tell which ad you sold to whom. Spotify can see that, which is an advantage.

How good is the targeting on SPAN today?

I believe it's the most precise in podcasting.

But it's just using Nielsen, though?

Yes, it is. It's not perfect, but I believe it's the best out there, no questions asked. We're working in a cookie less environment, which makes it difficult. The game changer for Megaphone was when my CTO suggested bringing in a third-party to target against audience data. I said it would not only change our business, but it would change the podcast business.

But I don't understand how that's differentiated if it's just using a third party.

Because for the first time, Nielsen allowed someone to ingest its data instead of modeling off it. So we were able to know if you visited Kelly Blue Book in the past three minutes and you popped up, we would categorize you as auto.

But iHeart or Sirius can't buy that?

They do it differently. I can't really tell you how they target, but I think they model. They'll say, well, this show looks like this show, or this show has 60% auto intent.

But you must pay Nielsen a lot of money for that.

Our work with Graham Holdings was quite interesting. We collaborated with TV stations to make a connection with Nielsen, which we couldn't do on our own. This connection allowed us to differentiate ourselves. Nielsen's credibility in the business is what attracts advertisers, even though I'm not sure how precise their targeting is.

There's no reason why iHeart or Sirius can't do the same with someone else?

Their targeting might not be as precise, but it's still very appealing to advertisers. If I approached an advertiser who's been buying podcasting on the branding side, I doubt they could tell you how everyone is targeted. They're typically told, here’s our targeting against men aged 25 to 54, who are in the market for a new automobile.

Of that inventory, let's say, 30% gets sold directly, 70% goes to SPAN. If you fill half of that, you're left with 35% for programmatic. Could you explain what programmatic actually is in podcasting today, between the private marketplace, direct, and the guaranteed buys.  

If you're operating in the programmatic space, the private marketplace is where you'd want to be. If it was large enough and you could manage it, that would involve me, as a media buyer, promising to sell you one million impressions against a specific audience at a $15 CPM. You'd agree to the deal if you liked the programming. The deal would then be executed programmatically, eliminating the need for paperwork and manual exchanges.

However, programmatic is never guaranteed. You don't know what you're going to get, and even in a private marketplace, you're not guaranteed to get the inventory if you're bidding on it. Programmatic is often associated with bottom feeders. I don't mean to disparage brand advertisers, but they're not as concerned about the environment they're in. They just want the audience at the best possible CPM, so they'll bid low in hopes of capturing that. If they're not delivering, they might increase their bid. This attracts a variety of advertisers, not just the big brands, but also those promising quick weight loss. The CPMs can be very low in this scenario. Despite these challenges, I believe this business is growing rapidly.

When I use Spotify for Podcasters and it tells me I can build my audience targeting finance people at a $15 CPM, what am I buying?

A lot of that is audio streaming and Spotify music inventory. They were starting to incorporate some of the podcast inventory into that methodology, and they've probably done more of it since. I haven't really followed that aspect of it, but that would be SPAN advertising back to the partners.

How does the private marketplace differ from a guaranteed buy, which is the other format?

The private marketplace operates on a guarantee. It functions similarly to a deal between you and me, which we would then execute dynamically.

Why isn't it an auction?

It's because you're guaranteed. For instance, we will reserve the inventory for these 10 shows at a $15 CPM, and you can purchase a million impressions, which we will deliver. In an auction, you would place a bid without knowing if it will be accepted or how many impressions will run in specific shows. If you're targeting against an audience, you can't see the inventory, hence you can't predict what the inventory will look like. Unless you're personally overseeing it in a dynamic system, there's no way to determine the inventory's appearance or if someone else will claim it before you do.

How does this compare to Google's auction on websites?

Google's system is cookie-based, so they have a better understanding of the available inventory. In this case, you won't know what it looks like until you cast it. You might have a slight idea if you target differently. AdsWizz is probably the largest player in programmatic, as a platform. There's likely enough inventory that you'll get it if you try to purchase it that way. However, I haven't seen what it looks like from the buyer's side, only from the seller's side.

From the seller's perspective, how do you establish the infrastructure to offer an auction?  

A company like AdLarge, for instance, has inventory and AdsWizz will provide access to any inventory they have. So, AdsWizz is creating the demand and connecting it with the supply, taking a 15% cut.

It's like an Ad Exchange and a DSP, or just a DSP? 

They're connected to the DSPs.

So, they're an Ad Exchange then?

Yes.

Essentially, the buyers are on Trade Desk, they'll be bidding on audio, and it would go programmatically via AdsWizz to Spotify.

There are many different access points for selling locally and creating demand. You can collaborate with AdsWizz to establish that connection. Triton, another third party, is moving demand out of space and into podcasting. They have demand and can meet the supply.

From a technological standpoint, how do you envision the evolution of the bidding mechanism from the private marketplace today to, potentially, an auction in the future?

Both are going to grow. I believe the bidding process will expand. When I consider digital advertising, one of the biggest aspects is that these large brand agencies prefer to purchase programmatically. It's efficient. They avoid dealing with numerous individuals, and they profit from their technology to the brand. They charge for media, but also for the use of their technology, which adds another percent or several percent to the cost. Consequently, there's pressure from these large agencies and branding houses to buy in this manner. However, since it's a cookie-less environment and less precise, the process has been somewhat sloppy. It's improving, but the progress has been slow.

You can't target the CPM on the private marketplace, typically, if it's around 30 to 35 if you go direct.

You can set floors too. But with the private marketplace, you can keep the CPMs similar to selling direct. It's just a more efficient method. They'll suggest a $14 minimum. You probably should go below a $14 minimum if you want to maximize revenue. My guess is the average in a private marketplace is between $10 and $20. Again, it's just a guess as I don't spend much time there.

But what's the difference in the take rate between SPAN and programmatic for Spotify, for Megaphone?

Megaphone isn't using programmatic, they're just selling. The programmatic would be sold by the third party, by the publishers themselves, and that would fill in after SPAN. You might fill another 20% of your inventory at a low CPM.

Spotify doesn't actually take any rate for anything apart from SPAN?

Except for the owned and operated shows that they sell directly.

And if you sell via SPAN on Megaphone at a $35 CPM, what does Megaphone take? What does the publisher get?

The standard split for the Spotify Audience Network is 50-50. They take 50%, which is higher than what most people in the industry take. Despite being the highest, they can sell it at a premium because it's Spotify, a cool platform with great programming. They sell it as part of an overall media buy, capturing a higher share than others. I’ve heard that others, even their direct sales are below the Spotify Audience Network share. In my opinion, they're doing a very good job.

However, I heard that in the third quarter, iHeart sold out 92% of the inventory they had access to on the show. Through direct sales, selling against their network, and then programmatic, they were able to fill 92%, which is probably higher than it would be if they used direct sell and added Megaphone and programmatic. But those Megaphone CPMs are high. I don't know what percent of that 92% was programmatic. They sell a lot of programmatic. That's what I heard in the market the other day.

Back to the programmatic piece. Right now, on Spotify or Megaphone, they don't sell any programmatics to the publisher. Let's say I'm Malcolm Gladwell. I pre-record a host read, I sell a chunk direct. The rest then comes down to SPAN. They sell everything they can. Let's say there's about 30% or 35% of the inventory left. That then goes out to programmatic. But if I'm not mistaken, I, as a publisher, have to link up to AdsWizz with myself or some other programmatic platform.

That's correct.

That's a hassle.

It is indeed a hassle. But depending on your scale, it can create some good revenue. Some people argue whether any impression should go unfilled. My argument is that there's a benefit to that. The downside is that we're essentially pressuring the market to pay lower and lower CPM.

But moving forward, how do you see Spotify trying to play into programming?  

They could possibly buy, or they could link to it themselves with VAST tech. But they would have to put their inventory in a DSP. There could be a possibility that they create or build the technology to do so. But yes, they could use Triton or AdsWizz to make that link.

They need to link up to the DSP, right? The buyers are on Trade Desk.

My thought is that the fastest way from point A to point B is to just buy a small player that has the technology and create that link into the DSP network. That, in a sense, is programmatic but you just don't see it if you're a buyer, you don't see it on the same screen as you see all your other busy. It's somewhat of a hassle for buyers.

How would you compare the position of Trade Desk to Spotify for audio?

The Trade Desk fills some of Spotify's inventory outside of podcasting. Everyone is tapping into podcasting. There are five programmatic players that have a lot of the same inventory. What's happening with the Trade Desk is that they have people coming in and buying Spotify streaming. I don't believe any of Spotify's podcast inventory is in the Trade Desk, but it wouldn't be Spotify's inventory. It would be Malcolm Gladwell's inventory.

How is Spotify going to benefit from any programmatic advertising in the future?

I'm not sure how those contracts read, but it would have to be contractually allowing them to continue to fill that inventory. They could possibly put programmatic in and say that it's Spotify Audience Network, just filling inventory. They could buy or invest in a demand platform or work with a demand side platform.

They would have to create or build the technology to create that link. I'm not sure if they want programmatic to compete with Spotify Audience Network, right now. The CPMs are so much higher in Spotify Audience Network than they are in programmatic, as a whole.

So, it's less scalable, like you said.

Right.

It's less scalable. You can only do the top brands, the top shows, that way.

For direct sales. But Spotify Audience Network fills everything they have.

SPAN is basically, like you said, it comes down from after Malcolm Gladwell sells his post-read stuff directly. You get a chunk of the inventory, but then you at Megaphone. 

There are billions of impressions in Spotify Audience Network, right now. They just want to sell massive scale against the audience. I heard they did over 200 billion.

If I'm an advertiser, where am I buying that on SPAN? Am I on a Spotify platform?

You're buying blind. I would come to you and say, we've got all this programming in Spotify Audience Network. We're Spotify. It's brand safe because we work on GARM guidelines and we can use a third party. We own Chartable, we own Podsights, so we can prove that your advertising works. But these advertisers don't see where their ads run beforehand.

It's a manual sales process, at scale, because you need a person, a sales representative, to go out and tell the advertisers the impressions and then he buys at scale via SPAN.

Yes. He suggests targeting specific age ranges and gender.

That vision is a great opportunity. However, the idea of monetizing the long tail via programmatic seems distant now, as there is no DSP or Ad Exchange.

I'm sure they're considering it. There are intelligent people over there. As I mentioned, some of these other players are capturing a lower CPM than Spotify is sharing with some of its partners. The revenue from Megaphone's targeted marketplace is crucial to publishers. If you're on Megaphone, it's part of your business plan. You monitor that number monthly and want it to be as high as possible. It can make or break your organization. Spotify wants as much inventory as possible to target as precisely as possible. They have billions of impressions to serve. Spotify is a seal of approval on the inventory. They've done an excellent job with it. I believe that as programmatic grows, Spotify's Audience Network will continue to grow as well. They'll figure it out. If they need to shift and start selling some programmatic, I believe they'll manage it, but I can't assist them.

In your opinion, what is the biggest risk to Spotify?

They need to keep a close eye on YouTube and what they're doing in this space. Amazon's Art19 and Wondery are also noteworthy. Amazon has deep pockets. They don't target audiences at Amazon, but they have built some models on Art19. If Amazon can figure out how to buy against purchase habits in this space, which is challenging without cookies, they could do something extraordinary. They could invest a lot of money in the podcast space, which would be insignificant to them. Their shareholders wouldn't mind if they incurred a loss for a couple of years. Those are the two key things I would be aware of if I were Spotify.