3 Key Ways a VC Thinks About a Data-Oriented Startup
How does a venture capitalist (aka VC) think about data-oriented startups? I recently interviewed Evan Nisselson, General Partner of LDV Capital, to find out.
Evan has an interesting investment focus on what he calls ‘visual technology businesses.’ It partially stems from his early experiences with digital photography in as far back as the early-mid 90s, when he was one of the earliest digital photo editors on the internet for @Home Network.
Evan’s focus on visual technology is a strategic bias based on the explosion we’re presently seeing in visual images driven by internet connected devices (mobile, IoT, wearables, home security cameras, etc.). This, coupled with the advancements in machine learning and neural networking approaches to image recognition, signal determination and building intelligence from visual assets, is leading to advancements that comprise much of the innovations at growing startups today: VR, self-driving cars, drone piloting and Snapchat’s specs.
Admittedly, I found Evan’s approach to how he considers startups fascinating, so much so, that I asked him to participate on our special VC Panel at DataEngConf this week in NYC. The panel is designed to offer helpful entrepreneurial information to data engineers and data scientists who might be thinking about starting their own companies, including how to attract (and retain) VC interest.
But what are the special challenges a startup faces in jumping into the visual technology fray that VCs often consider? Evan and I discussed three things:
The common challenges with building a founding team should be relatively well understood by anyone presently involved in startups, in my opinion. Having the right balance of technical, product, business and distribution focus in a startup is that amazing chemistry experiment where the successful constellation seems to only manifest itself a small percentage of the time.
But, more specifically, how does a data-oriented business affect the earliest composition of your team? Evan says that if you’re a business-oriented founder looking for a technical complement, the danger could be to see the first data scientist you bring on board as being the ‘tech’ member of the team. In reality, many data scientists, while technical, don’t often come from a strong engineering background and are ill-equipped to build the software systems to wrangle data in the earliest stages of a company (let alone your first viable commercial product).
On the other hand, if you already have a strong technical founder/player on the team, the temptation can be for that person to want to bring in a more experienced data scientist to learn sophisticated techniques from — even when often in the earliest days, a startup doesn’t necessarily have enough ‘data to science’, from the perspective of volume. I’ve seen this in starting my own data-oriented companies as well — sometimes the most rudimentary rules that drive good-enough data intelligence can have the quickest impact in early stages when proving your product and getting traction. And often this can be done by the existing technical players on your team (ElasticSearch configuration params anyone?), without an early company falling headlong into the notion of “we must build our data science team now!”
2. Data Scale
How can a company build data scale — that seemingly impenetrable moat coveted by savvy founders, which has become the hallmark of the most successful and largest internet companies? Well, obviously through usage; but that only truly comes after a company already has many users. In the interim, Evan’s main point here, interestingly, wasn’t as much of a technical one as a personality one — founders need to be able to foresee their own obstacles to acquiring data and be creative in figuring out ways to overcome that barrier. Personally, I’ve seen this occur through data partnerships, scraping public datasets, and digitization of previously non-digital records.
For instance, a data-oriented company in NYC that I’ve worked with was highly successful in building data partnerships early on. They were successfully in this approach because they had sophisticated technology built to ingest, clean, and even add value (new data) on top of the partner’s data — some of which they would return to the partner as part of the deal.
3. Data Uniqueness
Finally, this brings us to data uniqueness — the other form of data augmentation (both in scale and uniqueness) — that comes in an accretive way via any additional data your users add on top of whatever data you started with. If you give your users meaningful ways of interacting with your platform, they can add tremendous value. This is now obvious to us in the data-exhaust-laden days of social networking platforms, but think about how this might affect B2B-based products as well. Interestingly enough, many current data-oriented businesses are in the B2B, not the B2C space. Apparently, investors think one to three social networks of Facebook-scale or Snapchat potential are enough and are now moving away from ‘social’ investments.
For a B2B use-case, medical imaging comes to mind as an example of both points. It’s easy to see how data partnerships are required to access your x-ray images across different medical facilities, hospitals, doctors and even insurance providers. At the same time, any of those providers could add value by tagging or annotating those images depending on their role in the overall process of saving you in an emergency, keeping you healthy or determining insurance benefits.
While these are the most common challenges that VCs take into consideration when looking at data-oriented startups, there are many more. To learn other key ways that VCs think about a data-oriented startup, check out the VC panel at DataEngConf.