Everything you want to know about subscriptions
A subscriptions taxonomy; why annual payments are a bad idea (for companies); subscriptions as price discrimination
Paid subscriptions and membership schemes are in vogue: there’s one for pretty much everything under the sun — newspapers, gyms, music and film streaming, bank accounts, restaurants, delivery services, education apps, dating apps, &c, &c.
As we’ll see in this post, there are good reasons for companies to introduce paid tiers. At the same time, there is a lot of confusion about the purpose and mechanics of such schemes, even at the companies that run them — anecdotally, many companies introduce them to ‘be like Amazon Prime’ or ‘because VCs love them’, without much thought on what they’re actually trying to achieve.
In this post, I hope to shed some light on the properties of subscriptions, and why/when they work.
Types of subscriptions
A PM I worked with at Monzo classified subscription products using this taxonomy:
Pay to access: services where you need to pay a fee to consume the product — e.g. Netflix.
Pay to upgrade aka freemium: services where most features can be used for free, but that offer extra benefits for a monthly fee — e.g. Spotify, Tinder.
Pay to save / earn: services where a monthly fee results in lower costs or other kinds of rewards (that are worth more than the monthly fee) — e.g. Deliveroo Plus, Santander 123.
These are not mutually exclusive. For example, Amazon prime is both pay-to-access (in that it unlocks next day deliveries) and pay-to-save, in that deliveries are free. Still, this taxonomy is useful, because the strategies companies need to execute differ depending on the primary type of the subscription service.
Before we discuss that though, we need to consider one more dimension: what is the objective of the company when designing a subscription service.
Subscription objectives
A subscription service can really only serve two purposes (besides optics a la the aforementioned ‘VCs love them’):
Deliver profit directly: in this case, the subscription itself is the focus of the business. For example, the point of the Netflix fee is to make money in itself.
Drive consumption: under this model, the benefits the subscription unlocks cause users to spend more on the company’s other products — for instance, the idea behind Amazon Prime or Deliveroo Plus is that users will place more orders with these companies. Here, companies often ‘lose’ money on the subscription itself (i.e. the fees they collect are more than offset by the cost of the benefits they offer), but make it up through the users’ higher consumption — though there is a nuanced case that we’ll discuss further on.
The second model can be further broken down into two (non-mutually exclusive) variants: using the subscription to drive acquisition or higher engagement from existing users. For instance, given Amazon’s high penetration rate*, I imagine Prime was conceived to get existing Amazon users to spend more on Amazon; in contrast, Santander’s 123 account was designed primarily to attract new customers.
*As an aside, it’s useful to distinguish between penetration rate and market share: the former refers to the % of the population who have transacted with a company; the latter refers to the % of total spend captured by the company (it’s a function of penetration and % of each user’s spend the company captures).
Putting the two together
In general, pay-to-access and pay-to-upgrade models aim to generate profit from the subscription, whereas pay to earn aim to drive engagement and/or acquisition.
There is a case to be made that pay-to-upgrade can have a secondary aim of driving profitable engagement — though I can’t think of any good examples of that (it’s usually the case that users who subscribe to paid tier will become more engaged; but that engagement is unlikely to be profitable in itself, except insofar as it ensures the user doesn’t cancel their subscription.)
Finally, there are three ways for a pay-to-earn subscription to aim for direct profit - but not in a straightforward manner. This is because if the main benefit of a subscription is that users earn or save money, users will only sign up if they save or earn more than the subscription costs them. That said,
some companies rely on ‘breakage’: that is, their subscription does offer benefits that more than offset the subscription’s cost, but most users don’t make full use of the benefits (think of how many people sign up for gym membership but never go). In my view, this isn’t a moral business model — it’s basically duping users.
some companies offer subscriptions that result in cost savings compared to users’ buying the subscriptions’ benefits separately. In such cases, the subscription includes benefits whose value to users is higher but whose cost to the company is lower than the subscription fee. For example, you can imagine a subscription service that offers a bundle of streaming options (Netflix + Hulu + Paramount + Disney+) for less than the cost of paying for each of those separately. This can work if the company offering the bundle can negotiate a special deal from each of these streamers to buy accounts in bulk for a price lower than retail. In this case, the company offering the subscription is essentially a retailer or distributor for other services.)
finally, there can be a combination of a subscription fee + cost per use that is profitable thanks to the subscription fee itself - we’ll come back to this further down.
Strategies for different subscription types
The focus of a company offering subscriptions ought to differ depending on the subscription type:
Pay-to-access firms need to focus on optimising their marketing & distribution channel mix (this happens at the point where the return on marginal investment from each channel is the same); the can also try tiered strategies (as Netflix is doing, with their ad-supported cheaper tier; this is bringing them to pay-to-upgrade territory, the difference being that there is no free tier).
Pay-to-upgrade firms can use their free tier to entice users to try their product; they must then focus on selling their subscription product to these users (an interesting argument in the HBR is that a very high conversion rate to the premium tiers isn’t always a good thing: it suggests the company isn’t attracting enough users to its free tier. A 10% conversion rate from 1,000 users is better than a 50% rate from 100 users). These companies also have more leeway to experiment with pricing — they have the flexibility to show a different price point to each user, and therefore run experiments to gauge demand elasticity.
Finally, pay-to-save companies should be putting all their effort behind figuring out how to minimise cannibalisation: they must balance the level of the benefits they offer and the fees they charge so that only the users who wouldn’t otherwise act in the way the company wants them to act take up the offer — see the next section for more details on this.
Further considerations
In this section, I jot down some thoughts on interesting aspects of various subscription models, namely: the concept of upfront annual payments, pay-to-earn as a cap to cannibalisation, subscriptions as a means of price discrimination, and the factors that determine whether a subscription aiming to deepen engagement will in fact do so.
Upfront annual payment
I find pay-to-access subscription types the least interesting — they are pretty much your standard transaction: a customer wants a good or service, they pay for it, the merchant provides it. One aspect worth commenting on is the ubiquitous practice of offering users the choice of paying upfront instead of paying monthly, usually at a discount — see, e.g. Disney+:
I think this is usually a bad pricing tactic.
Let’s reason through this: suppose a company like Disney+ offers their service for $10 a month. Suppose also that 50% of users cancel their subscription sometime within their first year. This is what Disney’s financials will look like if 100 users sign up:
Churn at the end of the year is 50%. However, the average churn throughout the year is 23%. If no users churned, a user would be worth $120/year to Disney (revenue-wise, at least); but in reality, users are worth ~$92 each.
So, Disney reasons, if we offer users the option to pay upfront for the whole year in return for two months fee, they will be worth $100 (= $120 * 10/12). Since $100 is higher than $92, this is a good deal! And so they offer the discount.
But what this overlooks is the fact that the people who are most likely to choose to pay upfront are the ones who are least likely to churn in the first place!
Under reasonable assumptions, offering an annual payment results in roughly similar revenue to not doing so — you can check this by playing with this basic model I’ve put together (make a copy to edit). The idea behind the model is to segment users into low-intent and high-intent — the former being users who will definitely churn at some point within the year. Using this segmentation, and playing around with scenarios on how much of total churn will come from low-intent users, and what % of low- and high-intent users will opt to pay upfront, we can see that the reduction in churn will usually be offset by the discount offered to people who wouldn’t have churned anyway.
All this said, there are three cases when the practice makes sense:
The first is when most users are low-intent and therefore likely to churn. Of course, in this case, it might be better for the company to improve its service and retention rate, rather than trying to lock users in for a year!
The second is when it takes time for users to familiarise themselves with the service and to make the most of it, in which case nudging users to commit for longer makes sense. A company could tell whether this is happening if churn spikes in the first few months but stabilises after that.
The final case is when interest rates and inflation are high. In such an economy, a company might be better off accepting lower revenue for collecting cash upfront.
Note that it’d be absurd if users paid more for the one-off upfront fee than for monthly instalments (unless the economy is in severe deflation mode): the company would be worse off, because at best it’d generate the same amount of money, but taking a hit on cash flow; at worst, users will churn/default on their payments. This should be obvious, but I came across a hardware manufacturing company that planned to introduce a subscription service, under which customers would pay less if they opted for monthly instalments. Their reasoning was that a) their prospective investors liked subscription models, and b) a monthly plan would make it easier to sell more products to the users in the future. Even if this were true though, it’d not offset the cash flow impact (particularly acute in a hardware company, which incurs high costs for manufacturing), nor the cost of users defaulting on their monthly fee.
Pay-to-earn as a means of capping cannibalisation
Here’s a question that’s often overlooked: if Amazon Prime does indeed cause users to buy more from Amazon because of the free shipping or next-day delivery (and the profit from the increase in purchases more than offsets the cost of these benefits), then why doesn’t Amazon offer free shipping/next day delivery to all users?
There’s an argument here that it’s not Prime’s benefits that drive higher consumption, but users’ opting to pay for a fee in itself that does it — that is, there’s a psychological element here: “I’ve paid for this service, so I might as well make the most of it”. I think this is true to an extent: there’s the sunk cost fallacy that affects most people’s decision-making, but also the fact that users have subscribed means Amazon is always top-of-mind for them — so next time they’re shopping online, they’ll definitely check Amazon.
But the real answer, I believe, is that pay-to-earn/save subscriptions exist to cap cannibalisation. Cannibalisation is a catch-all term referring to an action that erodes a company’s own revenue in some way. For instance: if P&G launches a new shampoo, not all sales from that shampoo will be incremental — some will come from consumers who’d otherwise buy the company’s existing products.
In the case of things like free shipping, cannibalisation means that some (or even most) users will spend more when shipping is free — but, crucially, some won’t. So, if Amazon were to offer free shipping to everyone, it would end up giving value away for free to users whose behaviour wouldn’t change.
So, by requiring payment in exchange for free shipping (even when that payment is actually lower than the shipping fees, so all users who sign up for Prime are financially better off), Amazon avoids giving this value for free to those users who don’t seem to care about shipping costs (and whose behaviour is therefore unlikely to change when shipping is free); this may include very rich people, those who don’t care for speed & next-day deliveries or light users who only purchase things online infrequently.
In other words, such subscription schemes are basically a form of price discrimination: companies let users self-select into price-sensitive and price-insensitive groups and only offer better value to the price-sensitive group. (Of course, in the case of Amazon Prime, this doesn’t work perfectly — Amazon ends up giving free value to users whose behaviour won’t change: power users. Super-loyal customers who’d go to Amazon for all their shopping anyway will sign up for Prime regardless of whether or not they’re price-sensitive.)
An even better example than Amazon Prime is Santander’s 123 account. This account costs £4/month to maintain, but users can earn up to £50/month in cashback and interest. Santander could offer these benefits to all customers with a current account — but why would it? By offering it as a separate account, with a tiny bit of friction, it ensures that customers who aren’t looking around for good deals won’t receive money that won’t change their behaviour; only those bargain-hunting customers, who’d never open a current account paying nothing, will choose the 123 (this has been a particularly good tactic over the past decade: with zero/negative interest rates and low inflation, most customers haven’t been hunting around for good deals on their bank accounts, and deposits have been ‘sticky’. So there was no point in offering interest and cashback to all customers).
Pay-to-earn as price discrimination, pt II
This section is rather heavy on microeconomic theory; this is where I’d normally advise you to skip it if this isn’t your cup of tea, but I won’t — this stuff is just too interesting!
I can’t claim the insights here are my own — everything that follows is a summary of a post by Byrne Hobart, which is itself a summary of a chapter in David Friedman’s Hidden Order. Both of these sources are complex — I read both multiple times and exchanged several emails with Byrne (who was very kind with his time) before the concepts finally clicked; this is my attempt to make them as simple as possible.
To understand how subscriptions can be used as a precise price discrimination tool, we first need to discuss utility and demand curves. A demand curve shows how many units of goods customers are willing to buy for a range of prices. Typically, the higher the price, the fewer the units people are willing to purchase. To see why, imagine a very simple market with a single consumer. Let’s say we’re talking about video game cartridges for the Nintendo Switch. The first question is, how much is a video game worth to our consumers?
The answer is that it varies! If our consumer has a Switch but hasn’t bought any games yet, the first game they buy is worth a lot to them — say, $30. Once they have a game though, subsequent games aren’t worth as much. Suppose our consumer has bought Mario Kart. They’d like some variety, so they’re willing to spend some more money, but not as much; let’s say they’re willing to spend $15 to buy their second game. The third game will be worth less still — and the 100th game will probably not be worth much at all to them. We can plot this as follows (we call the value of a game to the consumer ‘utility’; marginal utility is the value of the latest game they buy):
Suppose now that there’s a single company making all Switch games, and that it costs this company $4 to make and sell a single cartridge. What’s the optimal price to charge to maximise profit? As I’ve explained in a previous post, the answer is to charge a price so that marginal revenue is equal to marginal cost. In our simplified model, the company should charge $15: at that price, our user is willing to buy 2 units (since 1 unit is worth >$15 to them, and the second is worth exactly $15); at lower prices, the users will buy more products, but the profit for the company will be lower:
What the company would really like to do is sell each cartridge at price exactly equal to its worth to the user; it would charge the user $30 for the first cartridge, $15 for the second, and so on. And guess what? The company can do this through a subscription fee!
To calculate the right fee to make this work, the company needs to know the consumer’s utility function — i.e. how much each unit is worth to them. In the age of big data, companies have a pretty good idea of that. So, here’s what the company can do:
It knows its cost of selling 1 cartridge is $4. So, it can sell cartridges at $4 and not lose any money per cartridge.
It also knows that at that price, the consumer is willing to buy 7 cartridges (since the 8th cartridge is worth less than $4 to them).
The company therefore knows that the user is willing to pay up to $78 for these 7 cartridges — that’s their total utility from the table above.
So what the company can do is tell the user ‘I will sell you cartridges for $4 if you pay a fee of $50’; this $50 is equal to the user’s spend limit ($78) minus the cost of the cartridges at that lower price (=$4*7 = $28).
Boom! The user is happy to do this, and the company will now make a profit of $50, more than double the $22 it’d make if it charged a flat price.
This is pure alchemy! You can follow the workings by making a copy of this file.
Will a subscription lead to higher engagement?
As we’ve discussed, pay-to-earn aims to increase engagement. Since pay-to-earn is essentially a form of discounting the price of a service, the factors that determine its likelihood of success are basically the same that affect price elasticity of demand. It’s worth being a little more specific on the availability of substitute goods factor, which we can refine/expand as follows:
User ‘stickiness’: in industries where users stick to one provider (for instance because of habit, friction, or monopoly), companies will find it hard to increase engagement. In such industries, pay-to-earn subscriptions will more likely than not end up subsidising users without changing their behaviour. In contrast, in industries where users find it easy to switch among many competitors, subscription plans can drive loyalty and market share gains (which is why they’re so prevalent in the airline or hotel industry, for instance; of course, if an airline were flying a single route on which it had exclusive rights, it’d be crazy to offer a subscription plan).
Market share: companies with high market share are less likely to benefit from subscription services. Since there is little user spend left to steal from competitors, the possible incremental gain is small, and the risk of subsidising existing behaviour is high. The only caveat here is that some companies can use subscriptions to expand their category (although arguably this is a matter of defining the ‘market’ over which ‘share’ is calculated correctly in the first place; for example, one might argue that even if Deliveroo had a 100% market share in the food delivery space, it could still benefit from a subscription service because people would order more in total. But if the original market were defined as ‘food’ as opposed to ‘delivery’, Deliveroo’s market share would not be 100%.)
Conclusions
In summary, if you’re considering a subscription service for your company, consider the following:
Be clear on your objective — is it to make revenue, or deepen engagement? Plan your activities accordingly.
Don’t spend too much time building different billing systems — annual payments are unlikely to work; if you really want to try them, run an experiment before rolling out to your entire user base.
If you’re aiming for deepening engagement, consider the price elasticity of your product, and explain why you think a subscription strategy would succeed in increasing engagement.
Run the maths as described above to set the optimal price for your subscription service — you’ll need a good understanding of the demand curve for your products before you do that, so it might be useful to run a few pricing experiments beforehand.
And that’s all! If you liked this post, please subscribe so you’re notified whenever I publish, or follow me on Twitter. Feel free to comment or email me questions/criticism (acatsambas@gmail.com).