Observations & Notes - February 2020
This is a set of reflections and things I’m paying attention to these days. I discuss bottoms-up adoption curves in life sciences, localization for regulated industries, the bundling of design tools, and much more.
Bottoms-Up in Bio: The “as-a-service” model is entering the world of life sciences. More startups popping up help outsource experiments, innovating on either workflows or procedures, the chemistry involved, or through the use of new machinery. Examples of outsourced activities range from sample prep to synthesis to mass spectrometry to bioassays and beyond. Companies building in this space include Emerald Cloud Lab and Strateos (formerly Transcriptic and 3Scan).
While the “AWS for life sciences” sounds nice in theory, it does come up with its set of challenges. For one, large pharmaceutical companies will outsource some of this experimental work already with contract research organizations (CROs), which are services companies, not product companies. The large enterprise buyers in this space don’t trust fledgling startups. Thus, startups providing lab work or experiments as-a-service tend to serve smaller biotechs or other startups. Some startups are great customers because you can grow with them, but it can be unstable ground to build a lasting business on. There’s opportunity not only for specific companies to emerge, but also to develop the strategies and knowledge base for bottoms-up adoption in life sciences, similar to how its emerged in Silicon Valley the past few years for SaaS and dev tools.
Translation and Localization in Regulated Industries: The translation industry is one of the main areas to explore within the internationalization and localization thesis. It’s a multi-pronged problem, where building machine translation technology that could enter the same ballpark as Google Translate is only step. And Google Translate is less than perfect, to put it lightly. Thus, involving human translators in the loop is necessary to build out enterprise-grade solutions. Lilt is one startup tackling this challenge.
Most of the translation industry is dominated by old-school companies like Lionbridge, which was purchased by private equity firm H.I.G. Capital in 2016. Lionbridge’s business line for regulated translation and localization stood out to me. Translating things like drug labels, financial reports, and legal documents comes at a higher cost due to legal risk if it goes wrong. Perhaps this is an opening for a software business.
Privacy-First Data Primitives: Privacy tech is hot, but it’s unclear what consumers’ willingness to pay or care is. Right now, companies make money and scale faster delivering enterprise solutions. GDPR and CCPA compliance is top of mind, along with the desire to mitigate legal risk. Some products in the space sit as wrappers or analytics layers on existing data assets — BigID or Transcend stick out as examples. Yet what if data primitives, such as databases and ETL tools, were architected to be privacy-first.
Trust and safety are baked into the product, and sensitive data isn’t accessible to engineers or data scientists. One example is Very Good Security, which processes sensitive data at the client and returns a token for the company to store. Other cases include InfoSum and Gospel Technology (both backed by IA Ventures in NYC). Tomasz Tunguz of Redpoint Ventures introduced the cloud prem architecture recently, where he wrote, “The idea behind the new architecture is split a SaaS app into code and the data. The SaaS company writes, updates, and maintains the code. And the customer manages the data.” Products engineered with high trust and privacy in mind not only service a compliance use case, but also enable software to spread into regulated industries like financial services or healthcare.
Design Tool Bundling and Unbundling: People typically rave about Figma’s collaborative features or ease-of-use, which both are about Figma as a product. I like the frame of Figma as a bundle. Before it came along, there were different tools for UI design, version control, asset libraries, and developer handoff. Figma bundled these features, which lowered the cognitive load of flipping between tools, but it also made the tool “easier to consume” for a department running a budget. Rather than paying for another point solution, they pay a single price for the holistic workflow tool.
As design gains more prominence within the enterprise and the part of the role blends in with engineering, will people seek to unbundle a comprehensive solution and potentially even pick up pieces of it that are open source? It’s more theory right now, but for example, what would an open source Zeplin look like?
Trust Proxies in a Network: Within a network of businesses working with each other, how does one signal to the other that it’s a trustworthy and reliable counter-party? If one sells software to a large enterprise, they look for SOC 2 compliance before on-boarding as a vendor. In this instance, the SOC 2 certification stands as a symbol of trust. YC companies work with other YC companies because they know the brand and trust the credentialing. That investment is not just money — it’s also an entry into a network and a stamp of approval.
I’ve seen a few creative ways to bring in proxies for trust. One startup enables professionals in the field to endorse local businesses. Berbix, born out of Airbnb’s trust and safety engineering initiatives, offers ID checks as a service.
Pros and Cons of WebAssembly: WebAssembly helps deploy high performance applications to the web and other clients. Very simplistically, developers write in the language of their choice that gets compiled into WebAssembly code, which is run on the client, and then translated into machine code. For our purposes, it’s best understood as a means to deliver performance-intense apps to the browser or other clients (mobile devices). WebAssembly has been around for a few years, but we’re starting to see more developers adopt this technology.
The technology has received some pushback. Sometimes, it’s been used for malicious purposes. However, I find the promise of the tool to be most captivating. It could open up an ecosystem of simulation tools and environments (the founder of Hash.ai said it helped inform their architecture), browser-based IDEs, games, and even end-user programming.