Gustav Lindqvist brings well over a decade of experience in data science with a demonstrated history of leading teams building data infrastructure, data products as well as uncovering actionable insights. Most recently he headed data science at Medium and prior to that led product and tech analytics at Spotify. Gustav currently consults several SF Bay Area-based startups.
Key takeaways from Episode 04
1. How data science can add value to a business?
“Data science can be split into two parts. One relates to the way you build a product. A recommendation system would be a prime example. you take data, you apply the quantitative method to it, you get something of it and put it back in the hands of users. They hopefully react to it, that data is fed back into that and hopefully, the product gets better over time”, said Gustav. “Then, there is a decision science that serves the same purpose of making something better. But in that case, it’s more about the best possible decision being made within the company. A/B testing is the perfect example of this. Contrary to popular belief, data science can play a huge part in it.”
2. What metrics does Spotify use to quantify its listeners?
“It’s all about the cohorts, the collection of people who sign up or share some milestone and time. So you need to track them to understand how their journey looks like so that you could use those insights to make the journey for the next cohort even better”, said Gustav. “Any metrics that reflect retention, engagement or volume of usage, and the conversion to a paid user. Others are the main ingredients in understanding how are you doing and whether what you are trying to do is actually working or not.”
3. How do you get great datasets?
“Sometimes you have all this data and you try to find some problem to solve with it. That’s where the application of data can be dumb. A great dataset always comes from when you actually have a need and you work backwards to figure out what sort of data could help you find the solution.”
4. How can startups build their own data pipelines and platform in order to harness the power of data?
“On the technical side, there’s no reason to build something from scratch when there’re so many good quality open source tools available. But you need to become good at taking these different components and ducktaping them together into a system. On the metrics side, you need to measure whether you are delivering value to your customers. The challenge here is to define metrics that make sense, and, more importantly, help articulate and explain them.”
5. What mistakes have you seen made by startups when applying data science to their business?
“A typical one is when a lot of time has been spent upfront on creating a complex system instead of doing something simple with a feedback loop to measure the activity results”, said Gustav.
6. What’s the alternative to building a data strategy around hiring top notch data scientists?
“There are too few experienced data scientists around, those who have earned the hard way how to deliver value with data science. Also, the management of data science is really lacking. So, you should invest in education, tooling, and processes for your current team to do that same type of job, applying rigorous methods to how to measure and understand things…”