Ads in streaming, differential pricing, and the pursuit of ARPU

Price discrimination, sometimes called differential pricing, is the practice of setting the price of substantially similar goods or services at different levels for different customer segments, based on the demand levels of those segments. The canonical example of price discrimination is that of airline dynamic pricing: passengers traveling for business purposes tend to be less sensitive to transportation costs than passengers traveling for leisure, so flights that are characteristic of business travel — booked very close to the date of travel, shorter than a week — are priced higher than those of leisure. The purpose of price discrimination is to reduce or eliminate consumer surplus, or the aggregate value of savings enjoyed by consumers who would have been willing to pay a higher price for the product than its equilibrium price. Consumer surplus is depicted in the diagram below, from my book, Freemium Economics:

The practice of price discrimination is predicated on the notion that different groups of consumers are willing to pay different prices for substantially the same good. The practice of price discrimination is ubiquitous, and it is often accomplished through bundling or quality tiers such that the different groups of people being exposed to different prices are not being offered identical products at the same time.

Price discrimination can take three forms:

  1. First-degree price discrimination exists when the seller is able to charge each individual consumer their reserve price, thus reducing consumer surplus to $0;
  2. Second-degree price discrimination exists when a supplier prices goods on the basis of quantity consumed, such as through bulk discounting or buy-one-get-one-free promotions. This form of price discrimination takes place when the supplier knows that different groups feature different demand curves and price elasticity sensitivities but it cannot identify those groups a priori. Discounting based on quantity incentivizes customers to spend more money than they perhaps otherwise would have absent the offer;
  3. Third-degree price discrimination exists when a seller can identify different groups based on their price elasticity sensitivities a priori and sets prices for those groups accordingly. An example of this is student pricing for cinema tickets.

These three forms of price discrimination were defined by Arthur Pigou in his book, Economics of Welfare, which was published in 1920. At the time, the notion of first-degree price discrimination must have been somewhat theoretical, or at least, scoped narrowly to very specific situations. First-degree price discrimination requires data aggregation and analysis capabilities that likely didn’t exist such that customer-level pricing could be calculated in real time with any level of precision or rigor in the 1920s.

But this capability is taken for granted for modern digital products, as I describe in Netflix already operates an ad network. Next stop: Content Fortress. Customers mostly expect that digital product experiences be personalized for them on the basis of their past engagement, and systems like recommenders and algorithmically-sorted content feeds are pervasive. Given that the full digital product experience can be personalized for users through behavioral data, surely so too can pricing?

In a paper titled First-Degree Price Discrimination Using Big Data, the economist Benjamin Shiller investigates the incremental value of web browsing data in implementing first-degree price discrimination for subscription packages for Netflix. He finds that, relative to using mere demographic data, the utilization of a user’s web browsing history in pricing a Netflix subscription could increase profits from a first-degree price discrimination model relative to a second-degree model by 12.2% vs. 0.8% for demographic data alone. From the paper:

I find that web browsing behavior substantially raises the amount by which person-specific pricing increases variable profits relative to second-degree PD – 2.14% if using all data to tailor prices, but only 0.14% using demographics alone. Expressed in total profits, this difference appears more striking – 12.2% vs. 0.8%. Web browsing data hence make first-degree PD more appealing to firms and likely to be implemented, thus impacting consumers. Substantial equity concerns may arise – I find some consumers may be charged about double the price some others are charged for the same product.

Ethical questions arise from this thought experiment, and some of these practices have been disrupted by platforms since the paper was written. But in a digital context, price discrimination can apply to more than a product-level price point, especially for freemium products. The freemium model is really an exercise in first-degree price discrimination, achieved through personalized experiences delivered at price points that are tailored to the individual user.

Two generally-accepted preconditions for the utilization of price discrimination are that a firm can differentiate the levels of demand and price sensitivity across groups (or, in the case of first-degree price discrimination, across individuals) and that a firm can enforce the differentiated price levels, for instance, by preventing a secondary market from arising. For digital products, a lack of transparency around the differentiated price levels is probably encapsulated in that second requirement: any given user would likely be incensed to learn that they paid more than other users for substantially the same product. Returning to the Netflix example from the paper: putting aside ethical concerns, while it might be technically feasible to build a price personalization engine at the level of an individual user, it’d be wholly impractical commercially, as those individualized prices would almost certainly become public.

The concept of price discrimination is an interesting backdrop against which to consider the current upheaval in the streaming ecosystem. As the streaming wars rage, the primary competitive front has shifted from subscriber counts to Average Revenue per User (ARPU): when Disney reported Q2 earnings last week, it boasted 221.1MM active subscriptions across all of its streaming services versus 220.7MM subscribers for Netflix, but with an ARPU less than half that of Netflix’s. (Note that the subscriber/subscription comparison between the services is problematic because Disney counts per-product usage as independent subscriptions, regardless of whether it happens as part of a bundle.)

Disney announced in its earnings release that it will institute a 38% price increase for its Disney+ streaming service, to $10.99 from $7.99, on December 8th. Disney also revealed that its ad-supported Disney+ tier, Disney+ Basic, will launch that same day — at a subscription price point of $7.99.

As Ben Thompson points out in his analysis, this strategy is clearly in service of improving ARPU and not subscription growth:

  • Non-subscribers that previously thought $7.99 was too expensive have no reason to revisit that decision and will remain unsubscribed;
  • Some users will find ads completely repellant but will be unwilling to pay more for the ad-free service and will churn;
  • Some users will continue paying $7.99 for the ad-supported tier despite the degraded experience;
  • Some users will upgrade to the higher-priced tier, unwilling to tolerate ads and whatever other feature differences exist in the ad-supported tier.

Contrast this to Netflix’s strategy, which will see its ad-supported product tier priced at a lower level than its ad-free tier. Of the ad-supported tier, Netflix’s co-CEO, Ted Sarandos, said:

We’ve left a big customer segment off the table, which is people who say: ‘Hey, Netflix is too expensive for me and I don’t mind advertising’….We’re adding an ad tier for folks who say, ‘Hey, I want a lower price and I’ll watch ads.’

This approach, by definition, can’t improve ARPU: while it will almost certainly boost overall subscription numbers, it will add net new subscribers at a lower price point (and potentially cannibalize some existing premium subscribers), reducing overall ARPU. This isn’t a bad strategy on its own, but it has to be considered in the broader scope of the streaming market. Disney is pursuing increased revenue with a deliberate attempt to grow ARPU that very likely won’t result in net new subscribers; Netflix is pursuing increased revenue with a deliberate attempt to grow its subscriber base that will decrease ARPU. Even assuming no cannibalization of premium subscribers by Netflix’s new ad-supported tier or loss of current Disney+ subcribers who would rather churn than pay more money or be exposed to ads, these two strategies represent different sides of a bet on the total addressable market of a subscription service. Disney, with its pricing strategy to chase ARPU, believes that it has more existing, price-inelastic users that will tolerate a higher subscription cost than there are new users to be acquired for the service. And Netflix believes the opposite.

I’ve spoken about the primacy of product personalization in light of the degradation of advertising efficiency engendered by platform privacy changes: if users are more expensive to acquire, advertisers must yield more value from them to finance user acquisition, and product personalization is one means of doing that. While I don’t think Disney or Netflix have been meaningfully impacted by recent privacy developments, the predicament they face is similar: as the streaming market has grown more congested, competition for user attention and wallet share has intensified, and it logically follows that acquiring customers would grow more challenging. Netflix is attempting to adapt to this by filling out the lower end of its supply curve, and Disney is moving some portion of existing users up the demand curve.

Photo by Artem Beliaikin on Unsplash