Monica Rogati taught us that building data products is just as much about people as it is about the data.
By Betsy Mikel (Editor, Women 2.0)
And we’re off with Day 2 at our HowTo Conference in San Francisco! (If you missed Day 1, check out our recap.)
This morning, we welcomed Monica Rogati to the stage for our opening keynote. Monica was one of LinkedIn’s earliest data scientists. In 2013, she joined Jawbone as the company’s first VP of data, where she built and now leads the data team.
How To: Understand the Difference Between Digital Natives and Data Natives
Most people are familiar with the term “digital natives.” This is the generation that grew up surrounded by technology, behind computers, using the Internet.
“I believe what we are seeing today is a transition and a similar revolution of data natives,” said Monica. “These are people who expect their world to be smart and expect the world to seamlessly adapt to their preferences and tastes.”
Digital natives are comfortable programming their own thermostat. Data natives want their thermostats to program themselves. Digital natives use the Starbucks app to order their latte. Data natives want their Starbucks app to already know their favorite drinks. Digital natives are concerned with what they can do with technology. Data natives are concerned with what technology can do for them.
“Digital natives expect a promise of a world that is better, richer and easier,” Monica said.
How To: Think Outside (and Inside) the Data
“Data products provide context and personalization by using data from you, others and the world,” said Monica. A common example is Amazon or Netflix, companies which recommend things you like and predict what you are going to buy next.
“Data science is not about giving you charts and graphs, it’s about giving you deeply personalized experiences,” said Monica. “Ultimately it’s about achieving your own goals. Everyone has a different preference.”
There is a misperception of data and algorithms being antiseptic and mechanical. “Creativity is crucial to data science,” Monica told us. Take the algorithm that “serendipitously” recommends music you might like on Spotify. That very personal experience is entirely built on data. Monica stressed that it takes a creative and communicative data scientist to bridge that gap and build a data product that will resonate with people.
How To: Lay the Groundwork for Data Products
“Start with analytics and exploration,” Monica told us. “At Jawbone, we have this huge influx of data: steps, sleep, food, battery readings, hardware codes, social media. It keeps coming in. Before you move onto sophisticated machine learning algorithms, you need to have this reliable data flow.”
But before you put that data to work to build products, you need to understand how to read and analyze that data. “Before building a smart thermostat, you first need to be able to read a thermometer,” Monica said.
- Good instrumentation: Figure out how to log and report things. You need to know what’s going on to build even the simplest data product.
- Reliable data flow: If your data is not sufficiently tolerant or stable, you won’t be able to build a data product.
- Data cleanup: For your team to work with the data, it needs to be as clean as possible.
- Fast iteration: If you don’t iterate fast, you might look at the wrong numbers without even knowing.
- Good UX: If it’s too difficult to put together data to look at and learn from, you might not do it at all.
How To: Build Data Products
When you’re ready to start build data products and learn from user behavior to make them better, you’ll need to get even more serious about the same steps you took to build your analytics.
- Obsessive instrumentation: It’s not enough to know who clicked on what. You need to know exactly what’s being shown so that you know what choice people made when they clicked on something. What else was there? Why did they make that particular choice? Since most people like to click on the first thing, what was the order and what insights can that info offer?
- Production-grade data flow: If the data is not flowing right, the consequences are severe. If your data flow into the dashboard is broken, you’re losing sales because you’re not recommending items.
- Impeccable data cleanup: You can’t afford mistakes. You can’t have broken code. International characters can’t look bad. An specific example from LinkedIn: You can’t have Darth Vader’s profile show up in “People You May Know.”
- Faster iteration: You can’t anticipate all the possible ways in which data can be broken or which people can play with the data and use your products. So that’s why faster iteration is even more important. You might not even know there is a Darth Vader account on LinkedIn.
- Learning: If you’d like to predict something, you need a model that incorporates many variables. You need machine learning. You don’t want to worry about which parameters to use in your model.
- Speed: You should build data products that react in real-time to user input and learn quickly for a more personalized experience.
- Smooth UX: The UX needs to be near-perfect and robust.
How To: Think Human in Data Products
Monica also reminded us that we need to consider one of the most important “components” of data products: The humans!
“Keeping the human in the loop is the crucial piece of the puzzle,” Monica said. “You need the right technology and the right data flows and fast iteration and tight feedback loop, but what you really need is data natives. The humans are what makes products feel smart and makes your life easier.”
“Great user experience and great data together are what makes products feel smart,” Monica reminded us. “On one hand you have easier-to-use software that seamlessly integrate into your life. On the other hand, all these personalized experience contribute to make it easier to use that product.”
Catch more of our HowTo Conference coverage on Twitter!