Ximalaya and the Economy of Ears

Five takeaways from the largest online audio platform in China's IPO filing

Hi friends!

Between the rise of Clubhouse (and the subsequent copying all across the board) and the recent developments in podcasting by Spotify and Apple, audio seems to be all the rage.

Ximalaya, which is the largest online audio platform in China, just filed to go public.

Ximalaya is a platform that connects users to billions of minutes of (non-music) audio content across a wide range of genres produced by professional to amateur creators. It is considered a pioneer in the online audio space, sometimes called the “economy of ears”.

Here are some interesting takeaways from their filing which shed light on the company and might suggest where things are headed in the audio world in the West.

I. Intentional focus on third-party devices and platforms to expand reach

Ximalaya ended the quarter with 250M monthly active users, growing ~24% y/y.

It reaches these users in three ways:

  • Mobile app: 42% of users

  • IoT and in-car devices: 20% of users

  • Third-party apps which embed Ximalaya: 38% of users

Over half of the users come from third-party platforms and devices, thanks to Ximalaya’s strategic focus on them

Ximalaya has done three things to expand its reach beyond just mobile:

  • They work with over 60 automobile makers in China, including Tesla China, Mercedes-Benz, BMW, and Audi to provide in-car audio content through pre-installed devices

  • They work with third-party mobile app developers and mobile device manufacturers to embed their features and access to their content library into their mobile applications

  • They also have their own IoT product portfolio of 6 products including smart speakers

It’s interesting to compare this with some of Spotify’s recent moves, including a mini-player on Facebook and the “Car Thing” smart player as well of course as Spotify’s integration with smart devices across the board.

II. A focus on “audio content” broadly

Ximalaya has over 280 million tracks of audio content across 100 genres, totaling approximately 2.1 billion minutes on their platform. Music aside, they have focused on audio content pretty broadly and have content including:

  • Podcasts: Ximalaya has a large variety of podcasts including personal development, business, leisure, sports, parenting, and technology on the platform

  • Audio entertainment including comedies: Ximalaya is the go-to destination for audio entertainment, in particular, traditional Chinese comedies such as Chinese Xiangsheng (or “crosstalk”) and Pingshu (a form of storytelling) and has ~1M albums1

  • Audiobooks: Ximalaya has ~4.5M titles sourced primarily from third-party IP partners.

  • Audio courses / educational content: Ximalaya has 9.5M albums of premium knowledge sharing which include courses for children and adults. Some are professionally produced, while others are user-generated.

  • Audio live streaming: Content creators on the platform tend to use audio live streaming to connect with fans and also create new kinds of content which integrate recorded content, such as audiobooks or radio dramas through dubbing and creative role plays

In the west, Spotify has music and podcasts (and some of these other use cases might sometimes be uploaded as podcasts), but it seems like Ximalaya has really leaned into audio as a format.

I wonder if over time Spotify will expand into more non-music and podcast content including specially produced comedies, meditation, fitness (guided audio workouts), and even audiobooks.

III. A vibrant content and creator ecosystem

Ximalaya has over 5M creators on the platform and talks about its “PGC+PUGC+UGC” strategy. What is that?

Ximalaya thinks of its content (and creators) across three segments:

  • Professionally generated content (PGC): This includes partnerships and licensing deals with top-tier publishers, online literature platforms, content creators, and key opinion leaders, including getting the rights to audiobooks and other content, as well as popular instructors who create courses. Ximalaya has exclusive licenses to ~70% of its top 100 albums. Ximalaya also is a leading player in acquiring rights to adapt popular content into audio series.

  • Professional user-generated content (PUGC): Ximalaya has a marketplace that matches content creators with high-quality copyrighted content that they can produce. It also helps train these content creators and helps them produce the content. This allows these creators to have a higher reach and better monetization than they might have otherwise. Ximalaya typically grants these PUGC creators rights to use the content they have licensed and shares a percentage of revenue generated from the applicable PUGC with them.

  • User-generated content (UGC): Ximalaya makes it easy for anybody to create and upload and get distribution for a wide variety of audio content. The user-generated content creators help maintain a breadth of content on the platform and also serve as a funnel into PUGC creators (many of the PUGC creators started as UGC)

In terms of volume of creators, about 2.1K produce PGC, 4.6K product PUGC and the other 5.15M produce UGC.

In terms of listening time, 15% comes from PGC content, 33% from PUGC content, and 52% from UGC content, illustrating the importance of the 6K PGC and PUGC creators and the power law in effect in that ~50% of the listening time comes from these ~0.1% of content creators.

Spotify’s leaning into exclusive content with Joe Rogan and Barack Obama is interesting and parallels the PGC focus of Ximalaya. In addition, however, I was fascinated by the PUGC approach, of essentially elevating talented people and giving them content to produce that Ximalaya separately acquired the rights to. I wonder if Spotify ever starts doing something similar.

Similarly, arguably none of the big platforms in the US (Apple, Spotify) has made it extremely simple to create a podcast/audio show. That’s one thing Clubhouse got right, and even Zoom and Descript on the pure “podcast” side. Spotify’s acquisition of Anchor was also a move in this direction.

IV. A multi-modal monetization model

Like many other Chinese internet companies, Ximalaya has a diversified monetization model comprising of many streams, best illustrated by the graphic below.

  • Advertising: Ximalaya makes advertising revenue from display ads and audio ads and has introduced programmatic advertising which now accounts for ~50% of advertising revenue.

  • Subscriptions: Ximalaya’s subscription revenue, which is the single largest source of revenue includes two things i) a monthly unlimited subscription which provides access to unlimited streaming of a large inventory of albums, and ii) one-time payment options for specific albums or pieces of content which can be as low as 50 cents. Currently, 13.3M users are active mobile paying users.

  • Live-streaming: Ximalaya has ~3.5M live-streaming MAUs. These users generate revenue for Ximalaya by purchasing virtual gifts and items for creators, which Ximalaya takes a cut from.

  • Educational Services: A newer segment, Ximalaya has essentially packaged its courses into subject-specific boot camps or vocational training programs or education for children. The foundational learning programs for children <12 years old in particular have been popular.

  • Others: A smaller segment comprising of revenue from the sale of IoT devices or fees they receive for the rights to convert their audio content to text.

The chart below shows the split of revenue over time, with education a recent and growing stream. The others are relatively constant in their splits (i.e., growing at similar rates).

A similar diversification is pretty constant across most Chinese companies, though no Western companies have cracked it. Spotify is arguably one of the most diverse with both subscriptions and advertising and is improving on it by adding one-off subscriptions to certain podcasts and building a dynamic advertising network. As the live-streaming platforms like Clubhouse think about monetization, the numerous examples of virtual gifting success in China might be a good source of inspiration.

V. Financials suggest “podcasting” has better economics than music for platforms

Ximalaya did $620M in revenue in 2020, growing 51.3%. In Q1 of ‘21, it grew revenue 65% compared to the same quarter the previous year.

A part of that revenue growth is driven by user growth, which is growing ~24% y/y, and the rest by higher revenue per user through higher engagement (active users spend 141 minutes per day on average in the app and this has been growing 5% y/y) and better monetization.

Given the nature of the business, content costs are a big expense but perhaps smaller than one might think. In aggregate, the cost of revenue for Ximalaya was ~51% (and has been declining as a percentage year over year). This is primarily driven by revenue sharing agreements (32%) which is the money paid out to the content creators and hosts and content costs (6-7%) which is the cost of licensed copyrights which are amortized over time.

Spotify has a lot lower gross margins (~25%), driven by the royalties it pays on the music side, which many believe is part of the reason for its push in podcasting. Ximalaya’s numbers could be a sign that even with exclusive content and first-party production, one can sustain higher gross margins in this type of content.

The other big cost driver for Ximalaya is sales and marketing, which is ~50% of revenue. It has also been reducing over time but is still high, as Ximalaya spends on user acquisition. It will be interesting to see if they can sustain some level of user growth as they reduce this.

Additional Reading

A few additional reads for those interested in learning more.

  1. Nieman Lab Interview about podcasting in China

  2. China Tech Buzz podcast episode on podcasting in China

  3. A profile of Ximalaya in Inkstone from 2019

P.S. If you see this then that means I didn’t have a chance to edit this piece before it was sent out probably because of my second vaccine dose reaction. Apologies for that!

Thanks for reading! If you liked this post, give it a heart up above to help others find it or share it with your friends.


If you have any comments or thoughts, feel free to tweet at me.

If you’re not a subscriber, you can subscribe below. I write about things related to technology and business once a week on Mondays.


An album here refers to a series of episodic content from the same creator(s)

Employee compensation and one-year equity grants

Hi friends!

Recently, some companies such as Lyft and Stripe moved to one-year equity payouts rather than the typical four-year equity grants that are common in the technology industry.

The headline of the article is misleading (or a genius PR move on Stripe / Lyft’s part) and paints this as a win for employees (“speed up”), which it isn’t.

I’ll go deeper into this today and cover:

  • Background on equity compensation

  • Contributing events in the past year

  • Motivations for companies

  • Employee perspective

  • Why early stage-startups shouldn’t adopt it

Background on equity compensation

Today, typically, employees at most tech companies are compensated in equity. For most public (and later-stage private companies), this takes the form of restricted stock units or RSUs (essentially shares), and for most startups, this takes the form of options that entitle the employee to buy shares at some price.

In both cases, the typical structure is as follows:

  • 4 year total vesting period on the initial grant

  • 1-year cliff, meaning nothing vests until that first year, after which the rest vests linearly either monthly or quarterly

This initial grant tends to be followed by “refresher grants”, either upon promotion events or on some fixed period (every year after X years), which have a similar vesting schedule.

At most later-stage companies, one way to think of this grant is as follows: the company makes an offer to the employee which includes $X in equity vesting over 4 years, where $X translated into a certain number of shares based on the price at the time they start.

In some sense, employees divide the $X by 4 to think about their effective compensation per year.

But the reality is that most companies, even public ones, tend to grow their valuation (even net of dilution) each year, so when the further out stocks vest, they are worth more than the initial grant price.

Looking at big tech data

Here’s a simplified view. If a big tech company gave an employee a $200K grant ($50K per year on the surface), vesting over 4 years, and they sold the grant each year as it vested, how much would it be worth?

I ran some of the rough numbers for Google, Amazon, Apple, and Microsoft with data from 2005 onwards.

Employees benefit from the appreciation, and so a ~$200K grant actually is worth a lot more, and especially the Year 3 and 4 value of the grants, which tend to be worth 75-100% more than the $50K “sticker value”.

The obvious takeaway is this: replacing the four-year grants with one-year grants granted every year is bad for employees unless adjustments are made where the average sizes each year are larger.

So why might companies be wanting to switch their equity compensation payout packages?

Contributing events in the last year

We’ve seen two things happen over the last year, both of which I suspect might have played a part in being the impetus for this change from Stripe and Lyft.

1) A huge step up in the valuation of private companies going public.

In part because of heated public markets with low-interest rates and record multiples, companies such as Snowflake and Coinbase went public at valuations ~10X more than their most recent financing just a year or 18 months prior.

While great for those companies, one obvious effect of it is that certain employees, who essentially joined between the two rounds, now have equity packages far beyond what the company might have expected to pay (and potentially the value these employees have created or may create).

This tweet captures that sentiment well.

To put things concretely, someone for whom HR targeted compensation of ~$100K/yr in equity for 4 years might be making $1M/yr for 4 years (likely far more than their market rate especially given they took no risk) given the ~10X increase in valuation just 3-6 months after joining, while someone a lot more senior who joins post the IPO (~3-6 months later) makes less.

One could argue that this is just a part and parcel of the Silicon Valley game, but companies might counter saying there is the opportunity to use the equity more effectively.

2) High volatility in stock prices in the last 12 months which leads to effective compensation for new hires being heavily dependent on timing

We saw a “crash” in markets in February and March in 2020, and an even more incredible recovery and boom that followed, which meant that people who joined in a 3 month period might have seen their entire stock grant appreciate 5-8X, and be making a lot more than those who joined a few months on either side for essentially no reason.

Examples of such stocks include Snap, Pinterest, and Tesla, as well as a whole host of SaaS companies, which are up 6-7X from March.

By offering grants in one year increments, timing becomes less of a factor because only the first year or equity rather than all four years of equity get artificially inflated (or deflated) because of large changes in stock prices

Motivations for companies to move to one-year payouts

The events above illustrate why some companies might contribute to wanting a different payout structure.

So what are the advantages of a one-year pay-out structure?

  • Reduce dilution: Companies that have seen their valuations rise over the years may view this as a way to reduce dilution if they’ve found a four-year vest plus their valuation growth often makes their compensation go above market range. The flip side to this is that since other companies will offer a four-year vest, it may discourage joining the company in the first place.

  • Stronger “Pay for performance” culture: In many ways, the timing one joined and that initial 4-year grant today dictates compensation at many places, especially later-stage startups. That initial grant has an outsized effect on compensation, even though the employer doesn’t know much about the employee. By switching from one large grant to a yearly grant, companies can pay more to those who are performing at a high level in years 2-4 than give everyone a grant size assuming some level of average performance at the time they’re hired. Obviously, refreshers are one way to do this in the typical structure, but this model makes it easier to compensate more closely tied to performance.

  • Reduce “rest and vest” incentives / staying just to vest initial grant/timing impact: At certain startups and later-stage companies which take off, employees in years 3 and 4 might be sticking around just to vest their initial grant and be resting and vesting or “earning more” than higher-performing similarly senior employees who joined later. This equity model can reduce some of that by tying the one-year payouts to current performance. Some might say you can always fire underperforming employees, but there is a middle ground where they aren’t underperforming but just not performing at the level of say the grant value they are vesting each year (and firing employees has other knock-on morale effects)

The flip side to this is that it could make companies who adopt this approach less competitive in the talent market. In addition, it could decrease loyalty in that there might be less of a reason to stay beyond year 1 into year 2-4 if the new grant sizes aren’t appropriate and market competitive.

Impact to Employees

As I mentioned earlier, absent no adjustments to the size of grants, a 4 year to 1-year move hurts employees. But how much?

That depends on how fast the company is growing and whether the employee sells immediately or holds for 4 years (often naturally the case in private companies).

To illustrate, I modeled a simplified example where an employee sells upon vesting (note that in private companies where the employee can’t sell, the effect of the switch is even more compounded).

So what happens if an equal-sized grant is used with one year vesting as in the 4-year vesting?

The employee is worse off by about 31% (or 55% of the initial value of the grant) because they don’t get to enjoy the benefits of compounding growth on the value of their later grants.

One natural question to consider is how much bigger the annual grant size has to be in a one-year payout model compared to the annual size in a four-year grant. The answer is below.

It will depend on how fast the company stock/valuation is increasing, but it’s striking that even at 15-20% growth rates, the size of the grant has to be 25-35% higher to keep the employee as well off.

In other words, if you were getting $400K vesting over four years, you need ~$135K per year in one a year payout model to be indifferent, assuming the company is growing 20% per year.

Note that while companies might argue that they will pay more in years 2-4, in the 4-year schedule, companies would have had some form of refresher anyway, so I’ve excluded that consideration in the calculation.

Early-stage companies

While it may make sense for some public or quasi-public (1-2 years from IPO) companies to adopt this above approach, where the equity is more like cash but with a slightly wider range of outcomes than big-tech equity, early-stage startups should not adopt this approach.

Already, hiring at the early stage is quite difficult. With the cost to start a company being lower than before and the market being hotter than ever, many employees wonder what the point of joining early on for 0.25-2% of a company is when they could found a company and get 20X that.

In addition, startups are still relatively cash-poor (or at least want to optimize for conserving cash) so tend to have lower base salaries than alternative offers, made up for by options that have the potential to appreciate a lot more.

Over 50% of startups fail, and employees who join these startups are aware of this. But they know that if that isn’t the case, their options can appreciate a lot more and be worth a lot more than had they taken an equivalent job at a later-stage / public company.

By moving to a one-year equity grant model, early-stage companies would be curtailing the upside benefit, without really protecting the downside in any way. If the startup isn’t working your equity still ends up being 0. But if it does work your equity ends up being worth a lot less.

As you can see in the simplified example below, the delta can end up being very meaningful - in this case, 3X times the initial value of the grant! Obviously, companies might issue larger grant sizes if doing one-year payouts, but the size of those would have to be huge to still provide the same upside.

Thanks for reading! If you liked this post, give it a heart up above to help others find it or share it with your friends.


If you have any comments or thoughts, feel free to tweet at me.

If you’re not a subscriber, you can subscribe below. I write about things related to technology and business once a week on Mondays.

Programmatic SEO, the long tail and customer acquisition

The tactic Zapier, Kapwing, Airtable and others use to acquire thousands of users for free

We’ve all heard the adage “distribution is king” or “CAC (customer acquisition cost) is the new rent”. While starting a business has never been easier, reaching customers has never been harder.

So the key question becomes: how does one acquire customers cheaply?

The other day, Patrick O’Shaughnessy asked about clever customer acquisition tactics people had seen to which I responded with the tactic of programmatic SEO, particularly with a focus on the very long tail and niche keywords.

I thought this was a reasonably well-known tactic, but received several DMs asking for more clarification and for advice on how this might apply to them so decided to make that the topic of this week’s newsletter.

I cover:

  • SEO and Programmatic SEO

  • How startups can implement programmatic SEO

  • Rounding up some examples: Zapier, Kapwing, and EquityZen

SEO and Programmatic SEO

SEO, or search engine optimization, refers to optimizing a product or service’s website to rank highly in search engines such as Google for the keywords one cares about to generate free traffic from them.

The algorithms of search engines are pretty complex and search engine optimization has many intricacies to it, but at a very high level, the two things which determine what web pages rank for a particular keyword are:

  • On-site factors: how well is that webpage optimized for that keyword (does it mention the full phrase directly in the title? or some of the phrase? in the body of the text?)

  • Off-site factors: how many “backlinks” (i.e., webpages from other websites) does the webpage have? Are they from authoritative web pages? Are those web pages related to the topic?

Programmatic SEO refers to creating landing pages on your website dynamically on a large scale to target the very niche and specific search terms people may use where your product or service may be helpful.

Yelp, Zillow, and Tripadvisor are some examples of the first wave of companies that leveraged programmatic SEO from the very start.

Search for any restaurant, apartment/house address, or landmark respectively and you will likely come across Yelp, Zillow, or Tripadvisor on the first page of Google. Here, the use case is pretty obvious - they are review/info websites and so generate a page for each place, and each page has some information/reviews about the place.

By generating a webpage for every individual place, they make it more likely that they can rank for that keyword (because of at least a well-optimized webpage for that keyword), and also can capture the traffic searching for specific restaurants rather than just general keywords like (“food review app” or “best french restaurants near me”).

What is the benefit of targeting more specific or niche keywords?

  1. Broad and generic keywords are very difficult to rank for, whereas more specific keywords can be easier to rank for because there is less competition for them. While any of these specific keywords individually is low volume, in aggregate, they can be fairly meaningful.

  2. These specific keywords tend to be higher intent and more likely to “convert” than broader or generic keywords.

Programmatic SEO can be used to generate hundreds or thousands of web pages, each meant to target an individual one of the keywords, such that very quickly a website can also start bringing in free traffic from these keywords without putting in a lot of effort for each specific keyword.

How startups can leverage programmatic SEO

Some of the examples above such as Yelp, Zillow, and TripAdvisor might seem a bit obvious: of course, they should have one webpage per restaurant/address/place.

But this approach can be leveraged by many more startups, including SaaS startups and others which even need users to be logged in to use their product.

The high-level approach is:

  1. Identify the keywords or terms users may use if they’re looking for something like your product or service

  2. Generate landing pages, one for each keyword (typically with some dynamic content)

1. Identifying keywords

This is in many ways the hardest part and is highly dependent on the product. The general goal is to think about all the niche and long-tail types of keywords that people might be using to look for something where your product could help.

Here are some specific ways to identify these keywords:

I. Think about specific vertical use cases for horizontal use-case products

For horizontal use-case products, one useful exercise to think through all the different vertical use cases that customers may be searching for where your product can help?

If you are a note-taking or spreadsheet software, perhaps “note-taking” is hard to rank for but “meal planner template” is easier. For example, Airtable has a template gallery with 200+ templates, each of which has its own page optimized to rank for that type of template.

Many users find Airtable through searching for a specific one of these templates, as opposed to a general keyword.

II. Think about modifiers to your primary value proposition which make it more specific

For products with a clear and specific value proposition, recognizing that people search for that value proposition in very specific ways depending on their own case use, including with company names or geographies. In general, consider appending some of the following to the primary keywords.

  • company names (e.g., “verify employment <company>” vs just verify employment)

  • specific examples of the thing in question (e.g., “<meme such as Willy Wonka tell me again> meme maker” vs just “meme maker”)

  • categories and subcategories (e.g. “hire <category such as web development> freelancer” vs hire freelancer”)

  • geographic locations (e.g., “hire web development freelancer <city> rather than just “hire web development freelancer”)

For example, Zapier lets you link 100s of web apps with each other. But most people looking for integration don’t search for “integrate any two apps”, they search for “<app name> < another app name > integration” or “<app name> integrations”.

Recognizing this, Zapier has basically built a webpage for each integration pair (think of it as having n*(n-1)/2 pages for this purpose where n is the number of apps it supports) to rank highly for many of these keyword sets.

Also, Zapier has built a webpage for every single app to target the “<app name> integrations” keyword, and so sometimes for certain keywords, two of their web pages may be relevant and show up as below.

Zapier gets millions of visitors for free from search engines!

As another example, Truework is a startup that provides employment verification. While it may be difficult to rank for the general term “employment verification”, many people who are looking to verify someone might search for the specific company that they want to verify employment for, so for example “employment verification <company>”. Truework has recognized this and built one webpage each per company, to capture this long-tail search across 100s and 1000s of company keywords.

III. Use Keyword Research Tools

Free tools such as Google’s keyword tool or Similarweb can help you understand:

  1. What modifiers and related keywords people are using related to your core keywords.

  2. What keywords competitors are using to drive traffic

Don’t focus too much on the volume of searches the keywords get, but search for patterns in terms of what categories of keywords people are using. Do a lot of people search for these by geography, with company names, with other types of categories (e.g., type of food), with modifiers such as “best or top X” etc.

2. Generate Landing Pages

Once you have your keywords, the next step is to generate landing pages.

Since typically these landing pages are generated dynamically, the key is to ensure that you have the right template for the landing page and the content readily available (in your database or through scraping) such that you can generate a landing page for each keyword programmatically.

A few things to keep in mind:

I. Optimize that webpage for search engines

Make sure your title and the text in the body mention the keyword you’re targeting with that webpage. Ideally, also include it in the URL, and have a clearly definite site hierarchy.

Zapier is a good example, with a well-organized site directory.

  • They have one webpage per company, which shows the list of integrations available for that company to target the "<company name> integration” type terms

  • They have one webpage per integration pair, which has a call to action to integrate the two services, to target the “<company name> <company name> integration” type terms.

  • Each of these pages includes that term in the title, description, and URL

II. Reinforce that you can help users with their *specific* keyword rather than a general one

A good example here is Kapwing, which is a general online photo and video editor. It also has a template gallery with a page each for specific memes. When searching for a specific meme followed by “template” or “maker”, Kapwing often shows up in search results. Over half of its traffic comes organically from search engines, in part because of this technique.

The landing page for each template clearly reinforces that they can help you with that specific ask, rather than the general terms. They show the meme image along with a description. They also have a clear call to action, which brings me to the next point.

III. Have a clear call to action, which if needed redirects the users to a specific state, directly tied to their request.

The landing page should have a clear call to action to take the user to your product/service (if the content itself wasn’t your product/service). What’s important is that, especially if you redirect users to a log-in page, make it clear how they can fulfill their request with your product, by having a clear call to action or even previewing the potential options for a user.

The Kapwing landing page image above is a good example, with a specific “Make it” call to action, which takes the user directly to the photo editor (without requiring a log-in).

Zapier also does a good job specifically showing what the different things you could do are, with a “Try it” button associated with each. This makes the user more comfortable that their specific use case can be addressed before they get redirected to the login page.

Rounding up examples

I’ve mentioned a few examples above of how this has been implemented across different use cases, but just wanted to summarize and call out a few others1. I’ve excluded some of the obvious “consumer” ones here such as Zillow, Tripadvisor, and Yelp, but those are also good examples.

  • Zapier is a master at this and recognized that people search for very specific companies when wanting integration, and gets ~7M visits a month, over half from search engines, thanks in large part to this technique.

  • Kapwing recognized that people typically are searching for a specific meme they want to make rather than a general meme maker, and gets ~6M visits per month, over half from search engines, somewhat thanks to this technique.

  • Truework recognized that people looking to verify someone’s employment will search for the specific companies they worked at. Of the ~150K visits month, 75+% comes from search engines. In fact, almost a whole percent of visits come from one such keyword: “Amazon employment verification”.

  • Airtable recognized that they can onramp people into their multipurpose horizontal use-case product through very specific templates that they might be looking for. This is a smaller channel for Airtable, with only 5% of their 16M visits through search, but is likely a more meaningful channel for new organic user growth.

  • EquityZen, which allows investors to buy stock in pre-IPO companies, recognized that people might search for specific companies (e.g. “<xyz company> stock” rather than just “invest in private companies”. It has a page optimized for each company and typically ranks in the top 5-10 for each of them. Over 75% of its ~800K visitors came from search engines, with 4 of their 5 popular search terms being of this exact form: “discord stock”, “stripe stock” etc.

Additional Reading

A few other resources I recommend for those interested in learning more:

  1. Essay on Zapier’s rise by Sacra, which touches on this growth tactic

  2. Essay on Zapier’s use of SEO for growth

  3. Moz on SEO strategies for programmatic linking

  4. Examples of these landing pages: Airtable, Kapwing, Truework, EquityZen, Zapier

  5. A thread by Trung Phan on Masterclass’s use of SEO leveraging a similar approach (niche) but with content and more targeted at second-order questions.

Thanks for reading! If you liked this post, give it a heart up above to help others find it or share it with your friends.


If you have any comments or thoughts, feel free to tweet at me.

If you’re not a subscriber, you can subscribe below. I write about things related to technology and business once a week on Mondays.


All data here is from SimilarWeb

The Democratization of IPOs

Why it matters and how it is happening

Hi friends!

There has also been some activity recently around the potential for retail investors to participate in IPOs at the IPO price - the “democratization of IPOs” so to speak, which is what I’ll go into in this piece.

I’ll cover:

  • The need for democratizing the IPO

  • The brokerage-centric approach

  • The company-centric approach

The need for democratizing the IPO

In Jeff Bezos’s recent shareholder letter, he included a note he received from a couple that bought Amazon shares when it IPO’ed. Given the appreciation, each share is up ~2000X. And so unsurprisingly, their investment in Amazon, which they’ve held until now, can cover a house purchase.

Recently, companies have started to go public much later on in their journeys and be valued at a lot higher valuations. When they do go public, retail investors that have access to just public markets have seen opportunities for such returns as in the case of Amazon diminish, as illustrated in this chart.

As I’ve written about previously, SPACs are one way to address this by encouraging more companies to go public sooner in their life cycle.

But not all companies will choose to go public via a SPAC, some will still want to do an old-fashioned IPO.

Even with companies going public later, there is still perhaps room for a 5X or 10X or more from the IPO for high-quality companies with long runways.

For the retail investor looking to invest in such companies at IPO, one of the frustrating elements is that they typically aren’t able to buy in at the IPO price. By the time they are able to buy in, the shares have already popped, sometimes more than 100%.1

As I’ve written about before, these pops can be quite meaningful, with over a quarter of them popping more than 50%, which essentially means a potential say 5X return from IPO, becomes a potential 3.3X return.

One way around this for companies is to go public via a direct listing, but if they need to raise primary capital then they have to raise a private round and then go public slightly later at a higher valuation anyway (this was somewhat the case in Roblox).2

But for the companies that choose to still IPO, the best case for retail investors is being able to get an allocation in the IPO itself and invest at the IPO price - “a democratization of the IPO”, so to speak.

Let’s dive into the two ways to democratize the IPO: the brokerage-centric approach, and the company-centric approach.

The brokerage-centric approach

In some sense, there has always been an opportunity for *some* retail investors to participate in IPOs. IPOs often have an institutional/retail split in terms of who the shares are allocated to, with ~10% typically earmarked for retail investors.

Brokerages such as Fidelity and TDAmeritrade for example, allow their customers to participate in IPOs (i.e., apply to receive an allocation), but require a minimum account value of $100-500K (depending on the IPO) and $250K respectively.

Besides, given that these IPOs are heavily oversubscribed, and the brokerages prioritize allocations based on the assets in their account or the revenue they bring in for their brokerage, often to actually receive an IPO allocation, the assets one needs can be a lot higher. It's safe to say that so far, participating in IPOs has been restricted to the 1% of retail investors.

That might be changing though.

Recently, brokerages SoFi and Public have announced the ability for customers to invest in IPOs. There have also been reports about Robinhood exploring the same feature.

The way this works for SoFi (and presumably Public) is similar to how TDAmeritrade and Fidelity have done it. 

The basic process is that those who are eligible can submit an indication of interest, and then the brokerage will allocate IPO shares based on how much allocation they get from the company going public and the criteria they choose. Customers can then do whatever they want with the shares, pending some restrictions.

I’ve detailed the key aspects below:

  1. Eligibility: SoFi will require a minimum of only $3K in your account to be eligible compared with the typical $100-500K

  2. Allocations: Those in good standing (who haven’t flipped IPOs in the past) will be prioritized by the brokerage based on several factors such as shares requested, account balance, and history.

  3. Restrictions: While one can sell shares immediately, brokerages discourage this “flipping” by limiting the ability of people who do so to participate in future IPOs. For Sofi, that flipping window is 30 days.

While the process is largely the same, the key is that the minimum account balances needed to participate are a lot lower, effectively democratizing the process. 

What will be interesting to see is whether these brokerages can secure enough allocations for the IPOs in the hot companies to meet the demand. After all, it won’t be the same level of democratization if only the top customers of these brokerages can get any allocation if they receive too much interest. Or imagine requesting 1000 shares and only receiving one.

The company-centric approach

Another way to democratize the IPO is a more company-centric approach, where companies exercise some judgment in choosing which retail investors participate in these IPOs and reserve some IPO allocation for them through a “directed share program”.

There’s a famous Buffett quote that goes:

In large part, companies obtain the shareholder constituency that they seek and deserve.

Companies know this well, and take great pains to “craft” the shareholder base they want, especially on the institutional side. But there’s an opportunity for more companies to do so on the retail investor side.

This is especially true for many marketplaces and consumer companies where retail investors are not only customers but often also “suppliers” or producers or developers that keep the platforms going.

Companies can use their IPO allocation as a way to allow their most loyal customers or “suppliers” to get in on “Day one”3 and participate in the upside.

We’ve seen a few examples on the “supplier” side, such as:

  • Uber set aside 3% of shares of its IPO offering for eligible drivers. To be eligible, drivers must have completed >2500 trips and have been active in 2019. 1.1M drivers (~28% of the total drivers) were eligible to participate and could buy up to $10K of shares. They also provided a one-time cash bonus to these drivers.

  • Airbnb set aside 7% of shares of its IPO offering for eligible hosts. Eligible hosts could pre-register to buy up to 275 shares (ultimately receiving a max of 200 because of excess demand). Given Airbnb’s 115% pop on day one, those hosts that participated more than doubled their investment in one day (though I imagine most continued to hold).

I was surprised Doordash didn’t do something similar (though they did give Dashers cash bonuses) when it went public, and I imagine going forward we’ll see companies like Instacart perhaps go this route when they go public (if they choose to IPO). 

On the customer side, Deliveroo which went public recently set aside shares in its IPO for its most loyal customers, who were able to buy up to £1000 worth (via a platform PrimaryBid). 70,000 customers took them up on the offer, buying a total of £50M. While the outcome wasn’t great in this case given that Deliveroo stock plunged ~30%, I think the idea of allowing your most loyal customers to get in slightly earlier is a noble one that will stick around. 

Overall, this approach allows companies to “reward” their loyal customers and suppliers, and handpick their customer base to an extent. Also, properly capturing retail demand can also result in potentially higher pricing and so a more efficiently executed IPO (in terms of lower pops). However, execute it “too efficiently” and companies may need to contend with IPO “drops” and their loyal customers feeling hard done by, although the ones that stick it out will presumably be okay in the long run.

Closing Thoughts

With retail participation in the markets increasing writ large, and more scrutiny on the one-day gains available to institutional investors, there have been changes to both the process of going public and also innovation and democratization of retail investors ability to participate in IPOs, from both brokerages and companies themselves. This does reduce some of the “imbalance” in the markets and is a welcome change.

One IPO I’ll be watching closely in this regard is Robinhood’s - after all, they have the ability to both democratize IPOs as a broker (starting with their own), and also allow their most loyal customers to be shareholders in Robinhood itself.

Thanks for reading! If you liked this post, give it a heart up above to help others find it or share it with your friends.


If you have any comments or thoughts, feel free to tweet at me.

If you’re not a subscriber, you can subscribe below. I write about things related to technology and business once a week on Mondays.


Yes sometimes these shares go downwards as the lockup expiration approaches and the float increases, but they often are still higher than the IPO price.


This is changing, with the SEC approving the ability for companies to direct list and raise primary capital


Obviously, IPOs are happening later in a companies life cycle and so “Day 1” is becoming later and later. Perhaps crypto and initiatives like Republic will actually enable getting in on “Day 0”.

Coinbase and the cryptoeconomy

Hi friends!

Coinbase is a leading cryptocurrency exchange and will be going public via a direct listing on April 14th. Its shares last traded hands at a ~$100B valuation in private markets.

With most startups going public these days the key question is “can the economics improve long-term?”. What I find most interesting with Coinbase is that the key question is “can the economics stay this good long-term?”

In this piece I’ll cover:

  • Coinbase overview

  • Coinbase business and financials

  • The bull and bear case for Coinbase

  • S-1 tidbits

Coinbase overview

Coinbase began in 2012 with the mission to “create an open financial system for the world”.

What began as a way for retail investors to buy, send and receive bitcoin via a simple and intuitive interface has over time expanded to a broader portfolio of products and services for retail and institutional investors, as illustrated below.

Coinbase serves three main sets of customers:

  • Retail users: Coinbase offers its 56M retail users a safe, trusted, and easy-to-use platform to invest, store, spend, earn, and use crypto assets.

  • Institutions: Coinbase provides 7000 institutions such as hedge funds, money managers, and corporations a one-stop-shop for accessing crypto markets.

  • Ecosystem partners: Coinbase provides over 115,000 ecosystem partners such as developers and asset issuers a platform with technology and services that enables them to build applications that leverage crypto.

Coinbase business and financials

Coinbase’s financials and metrics are truly mind-boggling.

In Q1 2021, they made ~$1.8 billion in revenue, more than all of 2020 and 2019 combined. And they had ~40% net income margins and over 60% Adjusted EBITDA margins.

They essentially grew ~850% y/y while generating 50%+ EBITDA margins.

The Rule of 40 is a rough rule of thumb that captures the growth and profitability trade-off which says that for a good (typically SaaS) company, their growth rate and profit margin should add up to 40% or more. Coinbase is over 900%.

So how exactly is Coinbase making so much highly profitable revenue? The reason is two-fold:

  1. High take rates on retail transactions

  2. The current state of the Crypto Market Cycle

High Take Rates on Retail Transactions

Coinbase makes money in two ways:

  • Transaction revenue: This revenue is derived from commissions on the trading volume generated by customers on our platform and tends to align with crypto-asset price cycles. In 2020, Coinbase made~ $1.1B of transaction revenue.

  • Subscription and services revenue: This revenue is more tied to the assets on the Platform and their use in products such as Store, Save, Borrow/Lend and Stake. In 2020, Coinbase made ~$45M of subscription and services revenue.

Over 95% of the revenue that Coinbase generates comes from transaction revenue, i.e., commissions on trades from retail and institutional clients, so let’s focus on that.

Digging one level further, Coinbase’s institutional trading volume makes up about 60% of total volume. However, Coinbase makes 95% of its transaction revenue (and 90% of its total revenue) from retail trading volume.

So how is Coinbase able to generate so much retail trading revenue?

Coinbase was obviously early in the space and has been able to become the trusted brand for consumers wanting to participate in the cryptoeconomy.

They have 56M retail users, over ~6M of whom transact every month. Their easy-to-use UIs and interfaces as well as their trusted brand make it the go-to onramp platform for consumers for crypto. This is evident in the fact that over 90% of users were acquired organically, and Sales and Marketing spend as a cost of revenue is under 10%.

This has enabled them to command premium pricing, particularly in retail transactions.

The take rates on retail transactions is a lot higher than institutional ones, meaning that retail revenues make up the bulk of revenues, even though they make up only 40% of transactions by dollar volume:

  • the take rate on retail transactions is ~1.25-1.5%

  • the take rate on institutional transactions is ~0.05%.

I’ve noted before that Coinbase itself has different take rates for different users based on the “UI” you use (3-4% in the regular flow vs 0.5% for Coinbase pro). Users who use the basic flow pay more in fees than those who use the more sophisticated UI of Coinbase pro.

This “tax” on unsophistication is a big driver of Coinbase’s revenue. If everyone used Coinbase pro (or paid the equivalent of Coinbase Pro rate), Coinbase would make ~60%-67 less money.

The Crypto Market Cycle

Transaction revenue is dependent on trading volumes, and in particular retail trading volumes, which are correlated with crypto-asset price cycles and bitcoin prices.

Coinbase notes that: “While we have grown rapidly, our growth has not been linear. Instead, it has come in waves aligned with crypto-asset price cycles which tend to be volatile and draw new customers, investors, and developers into the ecosystem.”

And at the time of Coinbase’s going public, we are in the midst of the fourth crypto-asset price cycle.

As the price of Bitcoin has increased from $7K to ~$60K, trading volumes have increased fourfold (16X when annualized) from $80B (2019) to $335B (Q1 2021).

It took almost 2 years for example for Monthly Transacting Users, which is a key metric that determines retail transaction volumes, to reach q1 2018 levels, but now they’re at 6.3M, >100% higher than the previous peak.

They mentioned a few times in the S-1 that rather than quarterly or yearly performance, they prefer to be measured on very long-term horizons and across crypto-asset price cycles. It will be interesting to see as a public company what kind of guidance they provide in quarterly earnings calls.

So far, most of their guidance has been scenarios around the number of monthly transacting users (a range of 4-7M).

The bull and bear case for Coinbase

The bull case for Coinbase is quite clear: they are the trusted brand, with phenomenal top-line and bottom-line numbers, and the cryptoeconomy is still early in its lifecycle.

Coinbase is at a run-rate of $3.2B in net income (based on Q1) and growing top-line at triple digits year over year, which event at a $100B valuation would be a < 35 P/E, which is arguably a lot more reasonable than SaaS companies at >25X revenue.

Also, there is a case that crypto is still relatively nascent if you consider that the crypto addressable market could be ~3B+ people, and Coinbase is currently at ~50M.

And Coinbase is capturing an increasing share of the cryptoeconomy, as in the chart below.

But I think the bear case is more interesting and worth touching on.

Coinbase’s current earnings are hardly “normal” - revenue and transaction volumes are at an extreme given the current state in the cycle and their profitability is unsustainably high.

One bear case is obvious: that crypto dies down and Coinbase consequently does not live up to its lofty valuations.

But the other scenario is worth exploring, what if crypto continues to do well but Coinbase doesn’t do well.

How might that happen?

There’s a classic Jeff Bezos quote that goes:

"Your margin is my opportunity."

And Coinbase’s margins present quite an opportunity. Yes, they are the trusted brand in the wild west of Crypto, and so customers were willing to pay 2% to buy these assets, given their simple and intuitive UI.

But now Crypto has a capitalization of over $1T. As the market has matured, competitors have come in, and I expect more to continue to do so.

  1. On one side, you have the current crop of consumer fintech companies in Robinhood and Square and Paypal, which can at least match the ease of use if not the trusted brand. Yes, their offerings are not as fully-fledged as Coinbase and don’t really give you access to the actual coins. But given that most people want access from an investment perspective they might not care. I’m also certain these companies will invest more heavily in crypto given the current trajectory of the market, and they’re already doing quite well. In fact, Robinhood had 9M people who traded crypto last quarter. That’s more than the 6M who traded on Coinbase each month!

  2. Then, you have the traditional brokerages - the Schwab’s and Fidelity’s who have to be looking at Coinbase’s numbers and crypto prices and thinking that for the benefit of our customers (and our shareholders), we have to be able to give them access to these investments.

  3. Lastly, you have the other crypto native brokerages such as Gemini and Binance, which tend to have lower transaction fees but perhaps not the same level of trust and brand.

To the extent that 90% of Coinbase’s revenue comes from retail trading fees, I do expect that with the maturation of the market and increasing competition (and increasing emphasis from competitors on Crypto), the fees come down.

One of the other benefits for the first two sets of companies above is that they don’t need to monetize through crypto trading, which should further result in more pressure on pricing.

  • Robinhood for example makes money from the payment for order flow on stocks and options, and as long as free crypto trading is a way to bring customers into their ecosystem cheaply, they can get away with close to zero-fee crypto trading.

  • Similarly, only 10% of Cash App customers have used it to buy Bitcoin before. But these customers have a gross profit per active customer and engagement in our broader ecosystem as bitcoin actives use of other products, such as Cash Card and direct deposit, more frequently compared to the average Cash App customer.

Coinbase alludes to the fee pressure that may arise:

Similar to other financial products, as the industry matures we anticipate fee pressure to emerge over time. Our strategy is to maintain our position as a trusted brand in the crypto space and develop new products to enhance our customer value proposition and offset the effects of any future fee pressure. If we are unable to capture value through the development of new and existing products and services or if fee pressure emerges more rapidly than we anticipate, our operating results may be adversely affected.

How might they maintain pricing power in retail?

  1. Their trusted brand will definitely help, which is evidenced by them almost becoming synonymous with crypto purchases for the average consumer, and their 90% organic acquisition number.

  2. Growing non-investing use cases: Coinbase states their flywheel is: retail users and institutions store assets and drive liquidity, which they use to expand the depth and breadth of products which then attracts new customers. I’m more skeptical.

    While I do believe Coinbase has a big advantage for those users who want to participate in the cryptoeconomy more heavily, for those who just want to buy coins from an investment perspective, I don’t know if the breadth of products necessarily matters.

    21% of retail users who invested also engaged with at least one non-investing product per quarter, and if I were Coinbase, I would be trying to drive that number higher, since those users are much more likely to remain on the platform regardless of competition.

The same way brokerages were able to charge fees and commissions which eventually eroded in part because of Robinhood, Coinbase could see a similar thing happen with its retail commissions. Overall, Coinbase is in part a magical business because Crypto was “the wild west”. As that is increasingly no longer the case, it will be interesting to see if Coinbase can remain a magical business.

Other S-1 tidbits

Before the upcoming direct listing, the news around Coinbase centered around Coinbase’s “mission-focused” culture, and the fall-out from that.

While I won’t go into that, their S-1 does provide some tidbits that suggest they are at least opinionated if not unique in some of their approaches.

Additional Reading

A few other resources if you’re interested in digging further.

  1. Coinbase S-1

  2. Robinhood note on how Crypto fits in with their broader mission

  3. Square shareholder letter which touches on how Crypto helps Cash App

  4. Meritech Capital’s Coinbase S-1 teardown

Thanks for reading! If you liked this post, give it a heart up above to help others find it or share it with your friends.


If you have any comments or thoughts, feel free to tweet at me.

If you’re not a subscriber, you can subscribe below. I write about things related to technology and business once a week on Mondays.

Share Tanay’s Newsletter

Loading more posts…