There are two main avenues that consumer tech companies and social platforms can take to monetize their products.
The first one: monetize on owned property. At a high level, this means tying revenue directly to users’ usage of your service. This looks like anything from selling ad space, to proposing paid subscriptions, taking transaction fees, offering in-app premium feature upgrades, etc. Most often, products monetize through a mix of these different methods.
This is the model of most social networks, apps, consumer web services, etc.
The second: monetize the underlying infrastructure. Infrastructure, here, broadly meaning the components that form the required backbone of the consumer-facing product. This means turning parts of your business that were historically cost centers (custom-made software tools, data center builds, etc.) into revenue drivers. Although the infrastructure was originally made to serve the product in some form, many companies have been able to spin out new B2B offerings based on processes, tools, and IP that was originally created for internal use only.
Some examples of this are well known: what became AWS started as an internal project around 2000 (!) as a way to solidify Amazon’s web infrastructure, which was struggling to handle the company’s hypergrowth at the time.
The first version of the Unreal Engine, the game physics engine which today underpins countless AAA video game titles was originally developed to be used as the engine for the 1998 game Unreal.
More recently, Snap Inc. has been following a similar strategy with their suite of camera tools and augmented reality technology originally built for Snapchat’s Lens features. Launched in 2018, Snap Kit offers a set of developer tools that let you integrate various parts of the Snapchat toolset into your own products. The Camera and Story kits in particular are prime examples of this infrastructure strategy. Rather than spending months trying to build an AR lens feature into your own app from scratch, save time and a pile of money by just integrating Snap’s Camera directly, and building on the shoulders of their existing ecosystem.
From planning to execution
Turning these cost centers into revenue drivers obviously sounds like a sound plan on the surface. But in practice, executing such a strategy essentially means launching an entirely new business unit. Building a product for internal use is one thing, but selling it and making a sustainable revenue stream from it represents a mammoth task that can pose a real threat of distracting management from the core business.
With that in mind — knowing that it can be resource-intensive and a heavy strategic move — it’s worth exploring some of the key factors that tend to push companies in the direction of growing their B2B offering alongside their core B2C business.
The first and most obvious reason is diversification and handling of a transition to maturity. Particularly for consumer social companies whose user revenue over a given period can be subject to significant variance and are tricky to forecast, locking in fixed-term large ticket contracts in an adjacent field offers partial insurance against a downturn in the core business’ metrics. Similarly, large social network companies often face the challenge of managing the transition/expansion to a new stage in the company’s lifecycle as user growth begins to slow down.
Opening new streams of revenue focused on high-value contracts with comparatively little extra R&D required represents an attractive line item to investors and can be pointed to by management as a new source of growth opportunity when the core B2C business starts showing signs of reaching the upper end of its S-curve.
As it turns out, Snapchat serves as the perfect illustration of a situation where a slowdown on the horizon prompted revenue stream diversification.
In a turn of perfect (?) timing, Snap announced the launch of Snap Kit in mid-June 2018, the first quarter in Snap’s reported history in which daily active user numbers dropped quarter-on-quarter. The launch came as part of a two-pronged strategy to expand the Snap ecosystem outside of the app — it was around this time that Snap fleshed out their content-first strategy in a heavy push to grow advertising revenue.
Bringing Snap Kit to developers to build on the Snap ‘infrastructure’ for their own projects is a perfect example of the second avenue of monetization.
Allowing 3rd party clients to build on top of your existing technology also offers a cheap and attractive source of data to fuel your own R&D efforts (licensing agreements allowing). Through this lens, expanding into a B2B offer, beyond serving as a diversified revenue stream, acts as a long-term investment to bolster your own tech innovation — and you can actually charge for it.
On an even longer time frame, this facilitates expansion into new verticals or into capturing new target markets and use cases in your existing verticals.
Here we come back around to turning cost centers into revenue drivers. This newly generated revenue effectively compounds, as each new client’s unique application of your owned IP brings with it new information and insights augmenting your own understanding of the range of possibilities it enables.
Licensing access to the fruits of your proprietary knowledge serves as a short-term, low-added-cost revenue driver, and an opening for long-term compounding growth opportunities.
And that brings us to our subject of the day.
Introducing BytePlus
Last week, ByteDance, the tech company behind TikTok, launched a new B2B service offering in the West under a new business unit known as BytePlus.
BytePlus’ core offering covers 4 main pillars:
- BytePlus Effects: the technological tools used in TikTok’s (and, originally Douyin’s) live camera effects and AR engine
- BytePlus Translate: similar to, well… Google Translate — offering text, voice, and image translation, and the AI/ML backbone that powers it
- DataRangers: a suite of different tools to gather, manage, and manipulate data at scale
- BytePlus Recommend: the “best-in-class recommendation algorithm” most famously known for driving TikTok’s hypergrowth
I’ll add: BytePlus’ concept isn’t an entirely brand new thing — ByteDance recently first launched these offers in the Chinese B2B scene under a new brand known as Volcengine.
Very similar to Snap, BytePlus brings gesture detection SDKs, body motion SDKs, beauty kit tool suites, and more to 3rd party developers to develop their own external IP. But where Snap has focused more heavily on its advertising and content strategy (which, it’s still worth noting, ByteDance has been working on too), BytePlus represents ByteDance’s decision to prioritize building a larger tech infrastructure ecosystem and extracting the most value possible from their crown jewel — their recommendations engine.
The famed recommendation engine, initially born as a news article recommendation tool for Taotiao, looked like it would be kept under the exclusive usage of ByteDance through the last couple of years — even as countries tried to force ByteDance’s hand into technology transfer agreements through 2020.
What factors, then, drove the recent decision to license access to this famed, historically secretive and exclusive IP? Why now?
The answer, I believe, brings up some of the points we explored above.
Revenue from TikTok (and Douyin — the Chinese version of TikTok) is believed to represent only a small part of ByteDance’s overall (rapidly growing) income, despite being the company’s most successful product by user numbers. Though having grown its advertising tools over the last year or so, alongside new features such as livestream donations, games, and e-commerce integrations, monetization avenues within TikTok have been relatively limited.
BytePlus, then, serves as a way by which to monetize the core value driver underlying TikTok, rather than risk alienating users by attempting to maximize revenue growth within the app itself.
Just as we explored above, layering a B2B offering on top of a B2C data collector acts as an attractive proposition to pitch long-term growth opportunities.
Furthermore, digging into the terms of service associated with BytePlus’ various infrastructure tools, it’s quickly apparent that they’ve been careful to maximize the compounding returns of licensing their sandboxed technology to outside clients.
BytePlus has been careful to ensure that all IP rights to software, data models, algorithms — basically all technical developments — built by clients on or around BytePlus’ underlying technology belong to BytePlus.
In fact, they went as far as to cast a broad net of ownership over “any intangible ideas, residual knowledge, concepts, know-how and techniques related to or learned from BytePlus’s provision of the Platform or Services”. This terminology is purposely vague, but associated with the company’s claim over clients’ “operational and technical data relating to the Platform and Services” paints a picture in which BytePlus is perfectly positioned to benefit from the compounding technical knowledge generated by having clients using its technology in new verticals and use cases.
The technical data derived from new implementations of BytePlus’ software models acts as fuel for BytePlus’ continued innovations and offers feelers into potential new markets to whom the company will later be able to address new products and services more directly.
Meanwhile, the (aggregated, anonymized) user data generated by client applications serves to further improve the recommendation algorithms and AI models that underpin the company’s infrastructure offers. As Byrne Hobart points out, ByteDance’s customization and recommendation algorithms can be applied in a content-agnostic way: “picking up a critical mass of useful signals is more important than knowing exactly what those signals are.”
Through sufficiently structured and defined taxonomies associated with the underlying content, in a broad sense, ByteDance doesn’t need to know exactly what type of content interests what type of person. This is the exact theory that allows ByteDance to offer its algorithms to other companies, ones that don’t need a short video recommendation tool.
A Theory of Attention
At a certain level of abstraction from the content itself, an AI/algorithm effective enough at pattern-matching user behavior can be used across a wide variety of use-cases almost unrecognizable from the original format for which the algorithms were developed.
In essence, the launch of BytePlus serves as a major step towards building ByteDance’s generalized “Theory of Attention”.
Most recommendation engines (take social network timeline algorithms, for example) essentially have to progressively feed an AI model for understanding what type of user is interested in what flavor of content. In most cases, this model is limited to one format of content. Twitter is pretty good at knowing what type of tweet I will want to see, and YouTube is great at knowing what type of video I want to see. Neither would (at least initially) be very good at doing the other one’s job.
And that’s exactly where ByteDance is looking to come in.
Rather than build new individual models for understanding what catches my attention on one site, what catches it on another, and so on, ByteDance wants to build an abstraction layer to package together all different signal sources to understand what underlying factors tend to lead to catching my attention at all.
By extending feelers into new verticals and use cases, ByteDance is aggregating a critical mass of signals to understand the factors that grab attention in all different types of contexts and situations.
In this sense, BytePlus’ recommendation algorithm service starts to bear resemblance to content discovery and native advertising companies like Taboola or Outbrain. Their models aren’t dependent on having access to a large abundance of a specific type of content, as long as a sufficiently wide variety of content exists for their algorithms to parse and display to the users.
For ByteDance, augmenting this Theory of Attention opens new opportunities for further optimizing their existing product line to increase revenue generated per user, and acts as R&D data to generate new products and services, which themselves feed back into this aggregated data.
To use tech Twitter lingo: they’re building a flywheel!
A last key point of interest in this move by BytePlus is what it symbolizes for ByteDance’s global expansion efforts.
As of mid-2021, TikTok remains banned in a handful of countries, most notably India. Under the pretense of national security, and as a response to the China-India border clash in 2020, TikTok has been entirely shut out of the Indian market for over a year. India, at the time, accounted for 1/3rd of all TikTok installs.
BytePlus faces no such restrictions. India’s ban targets TikTok (and several other Chinese apps), but no other ByteDance properties. As such, BytePlus serves as a new point of entry into the Indian market, opening it to a population of 1.3B Indians able to generate the valuable behavioral signals which fuel ByteDance’s technology. In fact, the company has already been forging some early partnerships with Indian firms, touting Indian clients such as Gamesapp on its site.
The initial criticisms various governments had around TikTok largely focused on its recommendation engine, and the extensive personal information it collects from users in order to power it. It’ll be interesting to watch how governments react in the long run to this almost identical technology being rolled out into their countries, perfectly legally circumventing bans and restrictions.
My (likely not that bold) prediction: nothing will happen.
TikTok’s recommendation algorithms and the data they collect served as a very public target throughout countries’ geopolitical spats with China. By virtue of its hypergrowth and ubiquity among younger generations, particularly in India, TikTok served as an easy boogeyman to attack.
Despite being the same technology as the one they previously criticized, governments won’t react to this expansion. The technology is less visibly prominent than it is with TikTok, and its extracted value much more nebulous. It’s much easier to rally political outcry against a prominent consumer-facing product than it is to criticize a single layer in a tech stack.
I’ll be closely following BytePlus’ next moves and new clients. Their clients so far are spread across several verticals — many, brand new sectors for ByteDance — including US fashion & e-commerce (Goat) or travel booking (Wego).
ByteDance was the first Chinese company to seriously challenge the existing tech giants in the social space. Now they’re setting their sights on their B2B offers. BytePlus marks the first large-scale data company out of China to cross borders and challenge (much larger) entrenched players like Azure, Google Cloud, and AWS.
And thanks to their famed crown jewel, it looks like they might just stand a chance.