Four Myths of Bundling

Explore

2️⃣

Myth 2. Revenue from bundles should be allocated based on usage

Our Myth-maker lives in the Bay Area tech scene where Usage is seen as the Great Equalizer, so when asked the question “how should revenue from a bundle be distributed among providers”, our Myth-maker quickly and energetically exclaims: “By Usage of course!!”

A number of years back, I was looking through cable viewership data, and found an interesting data point. At the time, ESPN and History Channel were getting almost the same amount of usage - e.g. if you looked at their rating points, viewership share, etc, they were quite comparable. And yet if you were to look at their carriage fees (the amount of money the cable company pays in fees to each provider out of your monthly cable bill), there is a ~20x difference. History Channel can get from $0.20-$0.40 per month per subscriber, while ESPN collects $4-6 per month per subscriber.

So is this unfair? Shouldn't they be paid the same, based on Usage?

Thesis 2: The most fair way to distribute revenue to providers in a bundle is by Marginal Churn Contribution, not Usage.

I propose that there’s an alternate valuation scheme here. Colloquially, it is referred to as “Anchor Value”. I would describe it more precisely as “Marginal Churn Contribution” (aka MCC) . In other words, I think that the best way to distribute money to provider X within a bundle is to ask “if I were to remove X from the bundle, how many people would churn?”. Similarly, you could ask the same question from a subscriber acquisition perspective (“If I were to add X to the bundle, how many new subscribers would I earn?”) ー also known as Marginal Subscriber Acquisition Cost Contribution (aka MSacC). But for simplicity, I’ll pretend they are interchangeable for now and focus on the churn-based model.

If you believe this framework, then the reason ESPN would receive ~20x the carriage fee of the History Channel is because removing ESPN from the bundle would theoretically cause roughly 20x the number of subscribers to churn as would removing History Channel from the bundle.

⁠

For example, let’s imagine that if ESPN were removed from the cable bundle then 10% of subscribers would churn. If the average monthly cable bill is $50/month, and roughly 100M households subscribe to cable, this would mean 10M households canceling, or a loss of 10M * $600/yr = $6B in annual revenue at risk. So the cable companies agree to pay the $6B, but distribute it across all 100M households, resulting in a carriage fee of $60/yr or $5/month.

One way to calculate WholesalePrices

So not only do I believe that the correct way to distribute money in a bundle is not by Usage, I also believe that the value of a product in a bundle can be determined mathematically in a way that approximates the Marginal Churn Contribution. The general statement of this principle would be as follows:

Corollary 2.1

WholesalePrice (ProductX in BundleY)

= RetailPrice (BundleY) * MarginalChurnContribution (ProductX in Population of BundleY)

Here’s the definition of some of those terms

WholesalePrice (ProductX in BundleY) is how much revenue should be distributed from BundleY to the provider of ProductX per subscriber of BundleY

RetailPrice (BundleY) is the price for the whole Bundle

MarginalChurnContribution (ProductX in Population of BundleY) is the percentage of people from BundleY who would churn if ProductX were removed

WholesalePrice is an interesting concept. The idea is that every provider in the bundle gets paid a portion of the customer’s bundle payment for every customer. In some sense, we’re distributing the SuperFan value of the bundle over the broader bundle population which includes both CasualFans and NonFans.

Putting this corollary in layman’s terms, it says that if I’m going to distribute money from my bundle between products, I should run the exercise of removing that product from the bundle and seeing how many customers I lose, and establish a total amount of money I would lose. Then I should turn around and distribute that sum across the population of my bundle and pay the provider that amount per subscriber.

Another way to calculate WholesalePrices

So that’s interesting and useful, but is less practical because it requires removing products in order to establish fair WholesalePrices. I think we can come up with an even more interesting and powerful corollary:

Corollary 2.2:

The best bundling outcome will have:

WholesalePrice (ProductX in BundleY) = RetailPrice (ProductX)

* SuperFan% (ProductX in Population of BundleY)

Let’s define the terms again

WholesalePrice (ProductX in BundleY) is the same as above: how much revenue should be distributed from BundleY to the provider of ProductX per subscriber of BundleY

RetailPrice (ProductX) is the a-la-carte price for ProductX. In this case, we’ll further define this to mean the price at which SuperFans of ProductX will pay for ProductX if it were offered a-la-carte.

SuperFan% (ProductX in Population of BundleY) is what % of the customers of BundleY are SuperFans of ProductX, when BundleY includes ProductX.

Now let’s try an example. For example, imagine that I am trying to include a music service (ProductX) into my bundle (BundleY).

Given that most music subscription services today are priced at ~$10/month, let’s assume that the RetailPrice (MusicService) is $10.

In round numbers, let’s presume that ~10% of the population of BundleY subscribes to one of these music services.

For now, we’ll estimate that the SuperFan% of the MusicService within that population is 10% as well (this is important, we’ll come back to it).

Therefore, the formula says that the appropriate WholesalePrice (MusicService in BundleY) <= $10 * 10%. In other words, the WholesalePrice is $1.

This may seem counter-intuitive at first. How can I take a service that (successfully) charges $10/month and only give it $1/month per subscriber? Let’s run through a hypothetical negotiation:

First, the BundleY company will point out that although this is 1/10th the price that the MusicService is used to, the overall revenue collected will be the same since the new customer base is 10x as large (implied because the SuperFan % is 10%).

Then, the MusicService provider will retort that this is unfair because the BundleY company is providing the service to some CasualFans, who (by definition), ascribe some non-zero value to receiving the service, and they will argue that they should be compensated for that as well.

And finally, In return, the BundleY company will argue that the MusicService no longer has to do any subscriber acquisition marketing, churn management, etc and therefore the price should actually be less than $1.

And in some of these cases, they will agree and the negotiations will generally stabilize to the equation above: RetailPrice * SuperFan%.

In fact, I think this pattern of negotiation is roughly what each cable carriage dispute oscillates between and where it eventually settles.

Ok but I said that this holds true in the “best bundling outcome” - what does that mean?

Let’s take two extremes:

Fully Overlapped SuperFan base: In this extreme, the existing population of BundleY has 100% overlap with the current ProductX SuperFan base

Fully Distinct SuperFan base: In the opposite extreme, there is zero overlap. In other words, there is no customer who is a SuperFan of ProductX who is an existing BundleY customer, and vice versa, there is no BundleY customer who is a SuperFan of ProductX

And just to make the math specific, let’s assume that BundleY has 40M subscribers paying $10/month and ProductX has 10M subscribers paying $10/month.

Let’s start with Case #1: Fully overlapped SuperFan base. So in this case, of the 40M BundleY customers, 10M of them are already SuperFans of ProductX and are paying $10/month for the service. There are two interesting perspectives:

From the perspective of the provider of ProductX, this is a rough situation because if they bundle into BundleY they will likely lose 100% of their a-la-carte customers. So they expect to be fully compensated across the 40M customer base. So they will argue that they currently get $100M/month ($10/month * 10M subs) and so to retain that, they will want the same amount of money across the 40M BundleY subs, so they will want $2.50 / month. This is the same equation as Corollary 2.2 - the SuperFan% of ProductX in the BundleY population is going to be 25%, so they will want 25% * $10/month = $2.50/month. So far so good.

But then we get to the perspective of the bundler. The bundler’s thought process will be that after they add ProductX to the bundle, they are still going to have 40M subscribers, and so they’ll be collecting the same amount of revenue. Where is this $100M/month payment come from? They could potentially raise prices - after all 25% of their customer base would still see a savings since they are currently paying $20/month for the two services. But what about the other 75%. This is tricky, and they will likely balk at this proposition.

Ok let’s look at Case #2: Fully Distinct SuperFan base. Now in this case, none of the current 10M ProductX subs are part of the 40M BundleY subs. Let’s look at the perspectives again, perhaps in reverse:

The bundler will be very excited about this. They will reason that they can add ProductX to the BundleY and get 10M new customers. The existing BundleY customers will get ProductX “for free” and BundleY will now have 10M new customers. So BundleY will now net $100M more in revenue each month. Dividing that across the 50M subs, BundleY would be willing to pay $2/month in fees to ProductX. Another way to look at this is to look at Corollary 2.1 - the MarginalChurnContribution (or MarginalSAC contribution in this case) for ProductX in BundleY is presumably 20% - i.e. 20% of BundleY’s overall customers (after ProductX is added) would churn if ProductX were removed. So the WholeSalePrice = the RetailPrice of the Bundle ($10) * the MarginalChurnContribution (20%) = $2/month.

The provider’s perspective is very similar to Case #1. They will likely lose all of their a-la-carte customers so they will expect to be compensated $100M/month. Given the new customer base will be 50M subscribers, they will be ok with $2/month in fees which happens to be the same as the RetailPrice of ProductX ($10) * the SuperFan% of ProductX in the new population (20%).

Another way to state this is that the bundle is “optimal” when Corollary 2.1 and Corollary 2.2 are equal: i.e.

The bundle is "optimal" when:

RetailPrice (BundleY) * MarginalChurnContribution (ProductX in Population of BundleY)

= RetailPrice (ProductX) * SuperFan% (ProductX in Population of BundleY)

This is interesting because establishing SuperFan% is much easier than understanding MarginalChurnContribution. For the former, you can look at the a-la-carte subscribers as a proxy, whereas for the latter, you eventually have to remove the product from the bundle to get a proper estimate.

But perhaps even more interestingly, this leads naturally to a very counter-intuitive conclusion in Myth 4: that the best bundles are ones that minimize superfan overlap. But let’s get through Myth 3 before going there.

Some things I’m leaving out for a separate discussion

For brevity, I’m leaving a few things out so I’ll just list them briefly:

How can one determine the correct RetailPrice and SuperFan% for a particular ProductX? The theoretical best way is to launch the a-la-carte service, wait for it to hit a terminal # of subscribers, and then do the math. But that’s often impractical. So thus far, I think this has mostly been done by negotiation, but I do think there are better ways to approximate these values today. However, I’ll leave that to a separate discussion (see

Extras⁠

)

What about unstable SuperFan%? The example above all assumes that the SuperFan% has stabilized. If the ProductX provider believes that they have 10M subs today but will have 20M next year, then the negotiation becomes trickier. For products like well-established cable networks, this seems reasonable, but for new products like a new music service, this is tougher. In the end this reduces to a similar question to #1 - how can one estimate SuperFan% with imperfect data. So separate discussion :)

What if BundleY and ProductX have substantially different prices? This is trickier as well since it’s not as obvious that a-la-carte customers of ProductX will immediately purchase BundleY as a substitute. We’ll explore this a bit in Myth 3, but there’s some more framing to be done beyond that as well.

A closing example

Just to close this section, here’s another real-world example: In January 2014, the WeatherChannel and DirecTV went through a contract dispute haggling over their carriage fee (AKA WholesalePrice in our constructs). The WeatherChannel put out a

press release⁠

as a part of their dispute arguing that DirecTV would lose 1.6M subscribers without the WeatherChannel. The

LA Times reported⁠

that the typical WeatherChannel carriage fee is $0.13 per subscriber per month, implying that the MarginalChurnContribution is ~0.1-0.2% (on a DirecTV ARPU of $50-100/mo). In other words, of the 20M DirecTV subs, all that is necessary to justify that carriage fee is for 20k-40k subscribers being WeatherChannel SuperFans and at risk of churn… much much less than the 1.6M than the WeatherChannel argued. This is an interesting case because the MCC is so small that it is very difficult to assess - this is no surprise since as bundles grow, the goal of the bundler is to add enough goods that any individual anchor would have very little impact on the bundle if it were to be removed. In doing so, it makes MCC very hard to measure, which leads to brinksmanship negotiations - i.e. sometimes ProductX actually has to be removed from BundleY to see the true MCC. One guess (just a guess) is that DirecTV took WeatherChannel dark, measured customer complaint volume, and then came back to the negotiating table with better data.

⁠

Next Section:
`Myth 3`⁠
, shift from the provider perspective to the consumer perspective

⁠

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.

2️⃣