Firehose #182: 💰 Subscription addiction. 💰

A framework for subscription success. Plus: Airbnb and Roblox go public, how to start a newsletter, and could you survive "Home Alone"?

One Big Thought

Subscription has become the default way that businesses purchase software. The SaaS ecosystem was worth $640B in 2019, up from next to nothing just 15 years ago, and is likely worth over $1T today.

Subscription has become an increasingly prevalent consumer business model as well. I don’t have a neat graph of market cap over time for consumer subscriptions, but the market cap of these businesses in aggregate is likely larger than for B2B. Legacy consumer utilities like Verizon ($250B market cap) and Comcast ($236B) have been subscription oriented for years, but internet native companies have built even more value with subscriptions in the last decade. Amazon* alone is worth $1.6T today, and its high margin Subscription Services segment is currently on a $24B revenue run-rate. 126 million Americans subscribed to Amazon Prime as of last month. Apple*, at nearly $2T in market cap, has grown its Services segment to $54B revenue in 2020 — larger than any other segment except iPhone, and at more than twice the gross margin of Apple’s legacy hardware business.

The categories of media, food, gaming, and fitness have been more recently transformed by subscriptions, thanks to companies like Netflix* (73 million US subs), Doordash (5 million Dashpass members), Microsoft* (15 million XBox Game Pass subs), and Peloton (3 million members), respectively. Even old school categories like insurance (Lemonade with 739K members), education (Chegg with over 4 million subscribers to its “Services”), and healthcare (One Medical with over 500K members) are susceptible to a subscription disrupter. In aggregate, these players and their competitors likely add up to or exceed the value of Amazon’s Prime member base.

Subscriptions are attractive because they are predictable, but not every product or service is a natural fit for subscription. First principles would suggest a few things must be true for a subscription to stick and grow rapidly.

Spinning the data flywheel

First, the subscription service must get better the more your customers use it. The data exhaust from transactions should lead to a more appealing service in the future. If this condition is true, a virtuous cycle driving more transactions is really beneficial to the customer, not just the company selling the service. This quote from the S-1 filing of Stitch Fix* in 2017 lays out the argument well:

We believe we are the only company that has successfully combined rich client data with detailed merchandise data to provide a personalized shopping experience for consumers. Clients directly provide us with meaningful data about themselves, such as style, size, fit and price preferences, when they complete their initial style profile and provide additional rich data about themselves and the merchandise they receive through the feedback they provide after receiving a Fix. Our clients are motivated to provide us with this detailed information because they recognize that doing so will result in a more personalized and successful experience. This perpetual feedback loop drives important network effects, as our client-provided data informs not only our personalization capabilities for the specific client, but also helps us better serve other clients. For example, a client may tell us in her feedback from a Fix that a size small blouse was the right size, but fit too tight in the shoulders. Based on the detailed information we know about that client’s body proportions and fit preferences and the item specific merchandise data we gather about the blouse, our stylists, informed by our algorithms’ recommendations, will not recommend that blouse to clients with similar body proportions and fit preferences. We believe our client feedback loop and its corresponding network effects ultimately contribute to better personalization for all of our clients.

Stitch Fix was inspired by Netflix*. In the Netflix model, success is measured by low churn. Netflix has industry leading churn, as shown below.

What’s most impressive is that Netflix has increased prices many times in the last decade, on the order of ~10% each time, with little evidence of increased churn. The most recent price increase in October 2020 raised standard plan to $14 (+8%) and premium to $18 (+13%).

If you can keep churn low while raising prices over and over, then you have solid evidence that your service is better than it used to be. For Netflix* and Stitch Fix*, that’s all about aggregate customer data.

Tapping into recurring behavior

Second, the subscription must tie to a recurring behavior. The causality must flow from the recurring behavior to the subscription, not the other way around. A simple change to subscription pricing, without addressing the underlying behavior, will not be successful.

Consumables like food are great opportunities for recurring subscriptions. Nearly 30% of Doordash customers opt-in to its Dashpass service because doing so unlocks better economics for those customers, who have a recurring use case for takeout food. Peloton’s subscription works because the benefits of exercise only accrue if activity is repeated multiple times per week. Social ties can even increase workout frequency. Pandemic aside, Peloton has meaningfully grown monthly workouts per connected fitness subscription. Calm* has done the same in the mental health space.

To the contrary, many of Netflix’s competitors claim to have a similar recurring behavior of media consumption, but in reality consumers sign up for a single show and shut off their subscription the next month. These services are not really subscriptions; instead, they are employing a transactional model in a subscription pricing scheme. Similarly, some of Peloton’s online fitness competitors do the equivalent of “channel stuffing.” In January, when New Year’s resolutions are top of mind, they deeply discount their monthly plans in an “annual” offering and pull forward the year’s revenue up-front. Most of these services churn out users at high rates once users hit a subscription cliff.

Path to upgrade

Third, a subscription offering should self-select your best customers and give others an aspirational path. Amazon Prime* is the canonical example. As of 2018, Prime members spend $1,400 annually vs. $600 for non-Prime users, shopping 26 times vs. 14 times per year. Amazon is happy to subsidize shipping because it makes more contribution dollars on these customers in aggregate. Faire* has a similar program called Insider, in which the company subsidizes shipping for its highest volume customers.

Upgrade paths can even exist without a de facto subscription. Both Chewy and Stitch Fix* offer an auto-ship feature, which gives meaningful discounts to the most active customers in exchange for predictable revenue. In 2019, nearly 70% of Chewy’s revenue came through auto-ship. At the time of its IPO, Chewy attributed its stellar net dollar retention to the high percentage of auto-ship customers, who spend 6% more than the average non-auto-ship customer in each purchase. That adds up over time!

In short, subscriptions are not a panacea. Price changes do not solve product/market fit on their own. But, if your product gets better with more usage, addresses a naturally recurring behavior, and has a potential “upgrade tier” of users, a subscription model could be very successful. I’m excited to see more companies add subscription offerings to their consumer services in creative ways.


Tweet of the Week

Links I Enjoy



So many IPOs; so little time! Between Wish, Doordash, Roblox, and others it’s been quite a couple weeks for consumer tech.

Airbnb filed its S-1 the week after I published my last Firehose on Doordash. I decided rather than do another S-1 post to share some work from others. This balanced post by Mauricio Prieto compares Airbnb’s journey to that of Bookings Group. As shown above, gross bookings is nearly identical on a time lag. The companies respective bottom lines look much different, however. Bookings Group achieved similar growth with very healthy EBITDA margins, while Airbnb burned several hundred million in 2018 and 2019. That’s despite a marked advantage for Airbnb in organic traffic (~20% of revs vs. 30%+ for Bookings Group).

Airbnb is no doubt a special company. It has invented a category of travel and now has the challenge of driving that model into the mainstream. Time will tell if its inventory advantage, which Prieto observes is receding, and its organic brand halo will ultimately win out against a more horizontal, commoditized competitor at the top of the funnel. Some of my favorite people in tech work there or have done so in the past. I am rooting for them all.


Newsletter-ception. →

I joined a newsletter co-op called “The Type House” a while back. It features some friends and colleagues from other popular publications, many of them on Substack.

We collaborated recently on a “how to” post for newsletter writers. If you’re interested in getting into the game, I’d encourage you to check out tips from yours truly and others!



Roblox filed its S-1 and it’s a doozy. The company has built a dual flywheel business that’s both social network and content development ecosystem. On one side sits 36 million DAUs. On the other sits 7 million active developers with 18 million unique experiences, of which 12 million were active in the last 12 months. I calculated that a DAU spends an average of 2.6 hours in Roblox, which is so large it feels incorrect, but not crazy given how immersive Roblox is. Its monetization is still lower than I would have expected at $39 ARPDAU, but with 23% FCF margins the company is going to invest heavily in increased monetization. While pre-COVID growth was slowish, the pandemic really bolstered this online community.

By the way, of the $1.2B in bookings Roblox saw in the last 9 months, more than half was deferred revenue! Consumers essentially float half a billion to the company in pre-paid in-game currency and then keep a meaningful chunk of that currency in game. It’s a massive float that makes me wonder if we should think of Roblox as a neobank! Along those lines, the company’s biggest cost bucket is payment processing (~25% of bookings).


The long haul.

A new study (disclaimer: not yet peer reviewed) shows that those infected with COVID-19 have enough immune cells to fight off the virus 8 months later. Some patients could potentially have immunity measured in years, but the response was not super consistent across patients.


“Home Alone” for the holidays.

Paint can to the face? Crowbar to the chest? VSauce broke down the traps from iconic Xmas movie Home Alone and demonstrated that those burglars definitely had their work cut out for them.

Enjoyed this newsletter?

Getting Drinking from the Firehose in your inbox via Substack is easy. Click below to subscribe:

Have some thoughts? Leave me a comment:

Leave a comment

Or share this post on social media to get the word out:


Disclaimer: * indicates a Lightspeed portfolio company, or other company in which I have economic interest. I also have economic interest in AAPL, ADBE, AMT, AMZN, BABA, BRK, BLK, CCI, CRM, GOOG/GOOGL, FB, HD, LMT, MA, MCD, MSFT, NFLX, NSRGY, NEE, PYPL, SHOP, SNAP, SPOT, SQ, TMO, TWLO, VEEV, and V.