Data products: introduction
“But how is that useful for anyone?”
I was at a loss for words and tired, standing in a friend’s kitchen after a run in the foothills of Boulder, CO in 2008. I had just spent five minutes giving an elevator pitch for my PhD thesis on ecological statistics to my friend’s dad, the founder of a large telecommunications company, when he asked that question. I bumbled on about the potential applications of what I was doing, but it was a little bit of a stretch.
In the end I settled on something like “the data are really complex and we can do some powerful analysis to get new insights.” Despite his critical eye, my friend’s dad was nice and we agreed that data can have value beyond scientific discovery, even for improving telecommunications products. Existential crisis avoided, I returned to my much-needed glass of water.
What are data products?
Let’s back up... To understand why the question from my friend’s dad is important we need to answer some other questions first: what are data products and why would someone be interested in them?
Data products facilitate an end goal through the use of data. In a corporate setting, as Katie Lazell-Fairman explains, data products need to “solve a market problem and directly generate revenue.” Thus, they can serve two main goals for companies:
- delight your users with an advanced capability and
- protect your business with a competitive advantage.
When data products are done well, one objective will support the other in a positive feedback cycle — a product can attract customers with data, which leads to greater data collection, which can then attract even more customers. This is a business “flywheel” that Gibson Biddle, former VP of Product at Netflix, would characterize as “hard to copy, in margin-enhancing ways.”
For example, Amazon’s Alexa is a data product that completes actions by processing user voice commands with a complex algorithm. It learns and improves with each usage. According to WIRED magazine:
… as more people used Alexa, Amazon got information that not only made that system perform better but supercharged its own [other] machine-learning tools and platforms — and made the company a hotter destination for machine-learning scientists.
Those outcomes are possible if everything aligns.
Why write about data products?
It has been almost 60 years since John Tukey published “The Future of Data Analysis” (sparking the field of data science) and almost 20 years since Amazon launched their recommendation algorithm. We know that companies that “extensively use customer analytics” outperform the competition. So, why write more about data products?
Despite the proliferation of data and analysis, most companies still do not know how to use data well. Earlier this year, the Harvard Business Review noted that many companies “are still information poor, even as leaders have implemented a wide array of programs aimed at exploiting data.” Many U.S. companies are only capturing 10–40% of the expected value of their data and anywhere from 80–87% of data science projects never make it into production.
Launching data products is a risky endeavor. Marty Cagan from the Silicon Valley Product Group outlines four big risks in product development:
- value risk,
- business viability risk,
- feasibility risk, and
- usability risk.
All four of these risks are high for data, which I think can explain the high failure rate of would-be data products. Probabilistically, your data product launch will fail. But it doesn’t have to be that way if we examine the risks.
In the coming weeks I will write about how we can derisk these four challenges. Along the way, we will uncover a strategy for launching successful data products.
A golden rule: Put people first
Let’s get started with a small piece of advice about data products from Google Design:
Put people first.
Data products use relatively novel technology. As a result, it is easy to get caught up in how to use data and lose sight of why to use data. Why do data, algorithms, and models exist? What do people get out of them? In the next article I’ll talk about how to ensure that your data are valuable.
