October 19, 2017
The Rise of Product Analytics
Analytics is starting to splinter into segments in a way analysts have dreamed about. One space that has been developing is product analytics, and one of the leaders in that space is Amplitude, provider of analytic solutions designed specifically with product teams needs in mind.
Amplitude’s success shows in its client roster. It has serviced companies across a wide number of verticals, such as OkCupid, Adidas, Microsoft, Twitter, Capital One, HubSpot, and Autodesk.
I had the chance to speak with Justin Bauer, VP of product to ask about how the space had developed. Before joining Amplitude two and half years ago, Bauer was CEO of a gaming studio and a principal of Booz Digital.
What trends did the Amplitude team detect that led to its offering?
Software is eating the world. We see instances of this with traditional retail businesses, like Adidas using digital to drive their mission to support all athletes, or airlines like Volaris who believe that in five years we will not have airline companies, but digital operations that fly airplanes.
Thus the importance of product teams inside large enterprises is skyrocketing. These teams are responsible for important business metrics, like revenue and product usage, where traditionally this was on the marketing side or even on the sales side.
But many product teams did not have analytics built for them. We look at Adobe or Google Analytics as great solutions that are built around marketing analytics use cases, not product use cases. A lot of the marketing technology solutions do not answer those questions like Amplitude does. That’s because all these digital experiences that people are having happen now with products. We built Amplitude from the ground up for that modern arrangement.
How was Amplitude inspired by AI or machine learning to deliver product analytics value?
In the early 2000s we had reporting cycles that might take years, or at least quarters, but today we have weekly, daily reports. It’s impossible to comb through every single data-point, and figure out from every single identifier, which trends are emerging.
In fact that was a case Microsoft described to us. The methodology of having analysts setup dashboards and having a product manager review them on a weekly basis — that is a cadence is now becoming so slow.
Because the product team is now driving business revenue, the cost of missing anomalies or trends can be huge. Machine learning is something that can help solve that. So that is what inspired us to build the data insights product line. We centered around the problem and then built a solution around that.
There’s a lot of debate about which tools are right for a given analysis. How do other dashboard and visualization solutions like Tableau, Highcharts, or ggplot for R programming measure up against the needs you see in the marketplace?
Many of our customers do use Tableau for general analytics. But when it comes to product analytics we believe that the speed of exploration is really critical. You can use R to answer any type of question. The answer is going to be very generalized, but when it comes to accessibility, it becomes a huge problem. What we typically see in organizations is the analyst becoming the expert on the customer, but that doesn’t enable the product team see what insights they can derive.
You also have to make sure the data is accurate. With Taxonomy we were able to identify errors and inconsistency and then fix it directly in the product without having to change the underlying code base, thus allowing the product teams to be more self-serving.
Sometimes data science can be time intensive. How does your offering help minimize concerns about resources?
If they have to wait for data science resources to make changes based on the questions that they have, that slows down the development — and more frequently they won’t ask the question to begin with.Technically you can answer in SQL and R. But insights can take minutes or more, and that’s even if you know what query to write. You realize it’s the wrong query, and then it takes longer. In fact it can take 15 questions to get to a useful insight.
If you are having to wait five minutes, ten minutes, or an hour between those questions, you’re not going to get through the exploration cycle. But if those results can be returned in seconds, you can answer 50 questions. We built Nova to do that, and now you can get to an understanding that wasn’t possible before.
Now that companies are moving so much faster, and product is being developed so much more quickly, traditional processes slow down product development. We believe that the solution to that is self-serving analytics, which is what we call product analytics. Give the product teams the capability to make decisions themselves.