Today we cover Part 2 of Algorithms as a Competitive Advantage with Andrew Parker of Spark Capital. In this segment we address:
- You’ve written about your investments in digital healthcare and education. Where do you see the future of these sectors and what areas are still potentially untapped?
- Are there particular sectors or types of startups that lend themselves moreso to an algorithmic-based competitive advantage?
- Do you think investors should use data and/or their own algorithms to better assess startups either pre or post investment?
- Any final thoughts or tips on data and algorithms whether it be for investors or entrepreneurs?
- What’s your take on Title III of the JOBS Act allowing unaccredited investors to invest in startups?
- What are you currently focused on at Spark?
- If we could cover any topic, what do you think should be addressed and who would you like to hear speak about it?
1- On evaluating a technical founder’s expertise
Andrew said that he assesses technical founders by treating them like a black box.
One way he does this is to look at the outcome of their work. What is the raw product they’ve produced? If there’s value in the algorithm, that should be transparent in the product execution.
Another method is to assess one’s thought leadership in code. A primary way to evaluate this is by checking Github to see the frequency of their published code and how well it’s reviewed and cited by other coders.
And Andrew’s last point here is that he is largely language and stack agnostic. He said that the best computer scientists are strongly rooted in programming paradigms that transcend all languages. So if they’re really good, they’re going to be able to pick up the language they need for the task at hand.
2- Data as a competitive advantage during early-stage startup assessment
During early-stage startup assessment, Andrew first asks, is the startup creating a new market or entering an existing one. If it’s the creation of a new market, it’s not clear how data is going to help. There must be imagination, faith and a compelling founding vision, a great founder to be able to evangelize that vision and an audience that believes in that vision. And here Andrew said that the audience doesn’t have to be the customer. The audience could be the investors or even the startups employees. I hadn’t thought of this point before and I think it was a great one that talking to the startups employees and not just the founders, prior to an investment, can give some insight on how much others also believe and are passionate about the vision.
3- Public algorithms, private source data
On this point, Andrew encourages companies to publish what they’re doing openly. He’s found that the more you give away, the more you get back. And by publishing an algorithm open source, others will contribute to it and make the idea better faster. And if the source data itself remains proprietary, while the community is helping improve the algorithm, then the startups competitive advantage is compounds. Here he cited examples including Google’s publishing of their machine learning framework TensorFlow, while retaining their source data around searches and other web activity in their data archive.
4- Continuous Feedback Loop of SaaS
In the era of shelf-ware, most software wasn’t internet enabled and didn’t have a method for improving user experience based on the collective experience of other users. This is why SaaS works so well for data and algorithm focused startups because both the value the product delivers and the innovation on the product-side doesn’t stop upon launch. Rather, after launch the product can increase in value at a compounding rate. And this also allows the startup to increase both the top line and bottom of the business over time by increasing the price/month of the SaaS product, if they are in fact increase the value of the product. Recall the interview with Mamoon Hamid on SaaS investing where we talked about Net New MRR. One component of this metric is expansion MRR, more dollars/month generated for existing customers. Whether expansion MRR is being driven by new feature sets, enhanced capabilities or just a better core product offering, it’s clear that the compounding value of data can be a force-multiplier here.