I recently caught up with Spenser Skates, CEO of Amplitude, a company primarily focused on driving user growth and retention for mobile-first companies. The analytics stack that Spenser and co-founder Curtis Liu built provides companies with a highly-scalable, cost-effective way to dig deep into all of their raw data and pinpoint the most valuable behavioral patterns. The interview touches on the genesis for his company and their focus on user behavior analytics for the mobile sector.
Daniel – Managing Editor, insideBIGDATA
insideBIGDATA: Why did you choose to focus on user behavior analytics?
Spenser Skates: When my cofounder Curtis and I worked on our first startup Sonalight, a voice recognition app for Android, we wanted to understand what impacted the retention of our users. We spent a lot of time trying various analytics products to attempt to answer that question. A lot of products could tell you what your retention was, but none of them provided insight into how to improve it.
We ended up building our own infrastructure to understand what our users were doing within Sonalight and how their behavior impacted their retention. We ultimately found out that the accuracy of voice recognition for a user’s contacts was a big predictor of retention. A successful first match increased retention long term by 50%. That allowed us to focus on improving the first time contact matching accuracy and ultimately the long term retention of Sonalight.
After spending more time sharing our findings with other app developers, we kept getting asked “how can we figure this out for our own application?” That’s when we decided to start Amplitude. Since then we’ve grown Amplitude to help mobile product managers answer questions about user behavior, find ways to improve their products, and ultimately retain more users and grow their business.
insideBIGDATA: What makes Amplitude Analytics different than other analytics tools?
Spenser Skates: Most self-service analytics tools provide charts including retention, conversion funnels, and daily/monthly active users. These graphs are great at getting a sense of how your business is doing and where your problems are, but are too high level to help you identify critical user behaviors. On the opposite end of the spectrum, you can write queries on raw data, which is time-consuming and requires far more technical resources.
Amplitude helps product managers understand why users stay or leave by giving access to what we call the behavioral layer: a rich middle layer of data in between reporting dashboards and raw user data.
By exploring the behavioral layer using tools like behavioral cohorting and our Growth Discovery Engine, our customers can quickly test whether certain behaviors are more likely to drive retention and engagement than others. This lets them zero in on the most important user behaviors that, in turn, lead to growth for their products.
Before Amplitude, the only way to access the behavioral layer was through costly in-house analytics infrastructure and data science teams. Facebook, for example, has a data science team who found that users who added at least 10 friends in the first 7 days were highly likely to retain long-term. Today, Amplitude enables companies without the extensive resources of an industry giant like Facebook to find this type of “magic moment” in hours.
insideBIGDATA: Rather than building on existing technologies, you chose to create a new stack unique to Amplitude. What contributed to that decision?
Spenser Skates: One major problem with existing mobile analytics tools at the time was the inability to handle data at scale. Mobile users generate far more data than web, and most existing solutions had been built for the web first. This meant that analytics tools either became incredibly expensive, slow, or both when tracking large data volumes. We decided to build a new analytics architecture that would enable companies to track and analyze large volumes of data (in some cases billions of data points per month).
insideBIGDATA: What advice do you have for companies trying to define their mobile analytics strategies?
Spenser Skates: The most powerful thing I’ve seen companies do is to find one unifying metric to center your team around. This metric should be a leading indicator for the company’s success metric, whether that’s retention, revenue, or some other business goal. For example, Facebook’s ‘7 friends in 10 days’ helped focus the team on getting new users to add friends, rather than spending time working on other features that may not have helped improve their main goal of retention.
Having a unifying metric like this is less about finding the absolute perfect number, but rather serves as a guide for prioritization and decision-making. Any time there is a new idea for a feature or campaign, ask yourself: would doing this contribute to our metric?
Another piece of advice is to avoid chasing short-term gains at the expense of durable growth. For instance, Zynga decided to launch a ‘flash sales’ campaign for their virtual goods. The first time Zynga ran a sale on these, it was immensely successful – the data suggested that sales were a great driver of revenue.
Zynga then decided to run these types of sales more frequently. Eventually, however, the demand for the virtual good decreased because gamers had purchased more than they could ever use, and these sales became less and less effective. Although the data indicated that the flash sales were great at regenerating more revenue, they were actually a bad strategy in the long run.
It’s important to realize that data, while incredibly important, isn’t everything. Data might indicate that making a certain product change will increase retention or revenue, but you should always consider whether the change would harm the user experience and possibly hinder your long-term growth. Use data to inform your decisions, but also take into account qualitative user feedback and intuition.
insideBIGDATA: Where is the mobile analytics sector headed? What can analysts and mobile product managers expect?
Spenser Skates: The industry is in the midst of a major shift. In 2015, Americans spent more time on mobile devices than desktops and laptops for the first time in history. We live in an increasingly mobile world, and the analytics industry needs to adapt.
In most cases, people believe their traditional web analytics product should also work for mobile. Unfortunately, it doesn’t usually work this way. There is such a big gap in the ways consumers engage with websites versus mobile apps. While web has traditionally relied on advertising, mobile has shifted revenue streams towards direct monetization of the end user.
The mobile analytics industry is beginning to realize the importance of understanding user behavior, and I expect this will begin to be reflected in the many self-service analytics tools available to product managers today. More of these tools will be able to answer complex queries via simple user interfaces, making it possible for non-technical users to do analyses that were once reserved for data analysts. In addition, data visualization will be more interactive, allowing more exploration within dashboards.
While mobile analytics now seems like a very niche industry, it’s no secret the business world is increasingly data-driven. Product managers want more data, actionable insights, in real time. As a result of this trend, mobile analytics is going to be a necessity for virtually every company in the coming months and years – pushing this “niche market” to the forefront of anyone’s overall analytics strategy.
Speak Your Mind