A few extra ideas that didn’t make the cut:
Practical mechanisms for establishing MarginalChurnContribution? Other mechanisms for establishing SuperFan%?
This got separated into it’s own writeup:
Can Providers make independent decisions?
Another common (but more minor) Myth is that Providers can make independent decisions on being behind a paywall vs not. We experimented with this on YouTube - offering the ability for channels to offer a-la-carte subscriptions, and other platforms like Google Play and the Apple App Store have similar constructs. It initially seemed surprising that most creators were unwilling to put content behind an individual paywall, and yet were very excited to get deals that allowed them to distribute through a cable network like Nickelodeon or Disney. But using the theories above, this makes sense. The way they would phrase it is “if I put my channel behind a $5 paywall on YT, I’ll at best keep 1% of my fans and it’s hard for them to make up the value of the other 99%. But if I distribute on Disney, I don’t give up any fans because everyone already subscribes to cable and has access to Disney.” This leads to an important conclusion:
A Provider’s decision to go “behind a paywall” is a function of (a) the composition of the distribution of that bundle and (b) the other participants in that bundle
Does ordering matter? Would you see different results if providers join/leave the bundle in different orders?
It turns out that the answer is yes! This insightful argument was pointed out to me on Twitter by
The net impact of this is mostly on
ー i.e. you can’t simply calculate each providers SuperFan% and determine a WholesalePrice, you need to run the same calculation for all permutations in which the bundle could be created, and then average the resulting WholesalePrices for each good. It’s worth reading more on
. One of the most useful examples is done with a gloves simulation: i.e. you have 3 people, 2 with right handed gloves and 1 with left handed gloves. Your job is to create a matching pair. How much should you pay each person for their marginal impact on the bundle? Shapley proved that the left handed glove owner should be paid 2/3rds of the value while the right handed owners get paid 1/6th each.
What about bundles where the cost of delivery is not zero - e.g. delivery services, etc?
Short version is that you have to add a “Minimum” to the WholesalePrice calculation which is the DeliveryCost of the goods. But it’s also important to note that there is often economies of scale in these delivery costs - e.g. shipping costs, warehousing costs, driver costs, etc often scale such that they become cheaper both as there are more subscribers (volume based) or when there are overlapping product needs (e.g. use just-in-time delivery services for urgent goods, and fill in empty spaces in delivery vehicles for goods that don’t need to be delivered immediately). In any case the formula would be adjusted to something like:
Minimum [ RetailPrice (
) * SuperFan% (
in Population of
What about bundles that depend on “Breakage” (customers not using the good) vs ones that don’t?
The easiest way to interpret this is as a variation on the previous question (i.e. it’s a non-zero delivery cost). If there is truly no economy of scale in delivery (i.e. the marginal cost of delivery doesn’t decrease with increased volume), then I’ve generally encouraged companies to find ways to add things to their bundle that have either zero or declining DeliveryCosts.
- Great stats on TV (HH share, etc)
- Jim Barksdale and Mark Andreessen on “there’s only two ways to make money in business: One is to bundle; the other is unbundle”
- Wharton Bundling Lecture - uses slightly different terminology (positive vs negative correlation) but fundamentally similar ideas
as a way to think about applying MCC in subset permutations (i.e. products that together form different outcomes) [pointed out by
, for a look at the original version of this write-up