Posted on by & filed under analytics, Business, growth, marketing, talks.

The contents of this blog post were originally presented at Books in Browsers 2014.

Getting access to Safari content for a limited time couldn’t be easier. In exchange for your email address Safari will give you access to all of our great content for 10 days — no credit card required, no long-term agreement. As you can imagine, Safari generates a lot of trials! We believe in our product and feel that we can win you over when you experience what we offer first-hand.

We know that we’ll never get to a point where 100% of those who sign up for trials will become paying customers, but we always strive to improve our “flip rate” (the percentage of customers who go from a trial to a paying subscription). In marketing speak, there are a lot of levers we can pull to improve that percentage.

  • We can improve the targeting of our marketing to get our product in front of the humans who are most likely to become paying subscribers.
  • We can improve our product by adding features that people love and are willing to pay for.
  • We can add more and better content to our library.
  • Heck, we can do all three!

Great, we’ll get right to work getting better content, adding new features to our site and finding people who are more willing to pay us for our service!  As with the rest of you living in the real world, we need to spend our limited time on what provides the most value. The first step is for us to identify what all those people who sign up for a trial really value. To do that, we need data.

What Content Do Customers Want?

Safari offers about 30,000 titles across dozens of categories. Although we’re traditional known for our tech content, we also offer a boatload of business, design and professional development content. When you sign up for Safari we do something really simple. We ask YOU what topics you’re interested in. We do this to prime our recommendations engine as well as gain insight into what our customers are interested in at a very high level.

Topic Selection

When we go back and look at what customers select, business is by far the most popular topic, followed by design. However, when we break down the topic selections by users who “flip” (sign up for a paid subscription) and those who don’t, we find that business is the top topic in the group that does NOT flip — it doesn’t even crack the top 10 for customers who do flip. That doesn’t mean that customers who are interested in business titles are not valuable to us, but we do need to go deeper for an understanding.

Topics That Pay

Our next step is to analyze the topics they actually read or watch during their trial. Doing so gives us an idea of what topics (and even individual titles) that customers value. It also identifies gaps is what users say they are interested in and what they end up reading. In the case of business titles, is it that we don’t have what customers are looking for? Are they having trouble finding it on our site? Or are is there a bias in the fact that Business is a broad category which generates a lot of “browsing” interest but not as much “buying” interest? It turns out, breaking business down into more specific topics helps, as well as working to improve discovery of business content.

What else can we do based on these findings? For one, we can determine what content we should feature prominently in order to maximize our chances of getting a customer to become a paid subscriber. The value of a Safari subscription is largely in the breadth of our library, but it does matter what content we push to the top, especially to a new customer. We can also work with publishers to identify not only what content gets a lot of usage during trial, but the content that leads to customers paying for subscriptions and furthermore, continuing to pay in the long run.

What’s the best way to improve the product?

What about the features we add to our product? As anyone who’s built software knows, there are a seemingly endless number of feature requests and personal beliefs about what matters. If you add every button, social share option and so on, a product will take too long to ship and when it does it’s bloated and hard to maintain. Thanks to all the data generated by our trial user base, we can examine which features matter based on hard facts, using a number of different techniques including logistic regressions and machine learning. While factors change over time, we find that there are some key things we should focus or product team on.

  1. Getting customers into the reading interface! It sounds obvious, but 24% of people that sign up for a trial never read a single page of a book. That’s a large share, and our models tell us if we get people reading, we can make gains on convincing customers to join us a subscriber.  Based on those findings, our product team has made changes to the onboarding process, homescreen and initial recommendations. Our marketing team can also use those findings to reach out to customers with recommendations and guidance for those who don’t get engaged early.
  2. Not only is it important to get customers to read something, but it’s important for them to do so early and often. We find that getting customers engaged in the first 48 hours of their trial is key to gaining them as a subscriber. It’s also key for them to view multiple titles rather than just many pages of a single title. Again, there are both product and marketing initiatives that have been and will be raised in priority to improve early engagement.
  3. Customers find the fact that Safari is optimized for tablets valuable. While we find no difference between usage on phones between trial customers who flip to paid, we see a significant increase in tablet usage during trial for those who subscribe. Tablets make for great reading devices and we should continue to invest in a site that is optimized for them.
Usage By Device

There are numerous other features that we find lead to paid subscriptions, including usage of social shares and annotations. In fact customers who share at least one annotation are 81 times more likely to sign up for a subscription!

Who are our customers?

Last but not least, what about our customers? Where should we focus our customer acquisition efforts and where are there opportunities for improvement?

We find out a lot about our customers from very basic web analytics. Knowing how they found our site, what country they reside in, what time of day and day of week they first visit us and so on tells us quite a bit. For example, customers that sign up for a trial after visiting our homepage are much more likely to subscribe than those that come in from a Google Search and land on a book preview. There is some bias in that a portion of them that already knew our brand and came to our homepage directly, but there are also many who get to know our brand and offering through our corporate site. A customer who is only exposed to a single title may not understand all that we offer. We take this as a challenge to do a better job in explaining our offering to all customers, regardless of how they enter our site.

We also see major differences in the likelihood of trial customers flipping to paid depending on what country they reside in. There are many factors involved, so we need to dive into the data to determine what we need to do in order to improve our value to each geography. In some cases, our price is high relative to local conditions. In other cases, our content might not match up with regional preferences. It could also be that finding the right content is difficult with a search engine that is primarily tuned in English (though we have made tweaks for other languages).

While the findings can help guide our product and engineering teams, it’s also important that we don’t try to solve for all markets at once. The web makes it possible to sell worldwide, but it’s often best to identify and tune your product to your target markets first. After those markets are doing well, work out to the next set of most similar geographies and repeat.

Safari Tweet

Finally, we’ve found it important to remember that at the end of the day our customers are humans! It’s something that can get lost while sifting through millions of rows of data, but it should not be overlooked. We like to set up phone calls (yes, we still have phones) with customers to learn about their experience with the product, include customers in beta programs as we build our mobile apps and run surveys on site to collect qualitative feedback. It all further supports the insights we mine from our data and ensure that we aren’t overlooking what might otherwise be lost in the noise.

Tags: analytics, Big Data, customer value, Data Science, pie charts aren't always bad,

Comments are closed.