The Guide to Data Export

FullStory has a firehose of information. Use Data Export to put it to work

In today’s world, we are told data is all around us and if you aren’t using data to make decisions, you are falling behind. But there’s a problem: in the vast sea of information available, which pieces should you analyze? And once analyzed what should you do next?

Like technology, merely having more data does not create any value; rather, it's what you do with the data that matters.

Value is created when data empowers decisions and strategies to improve your business processes. In this way, data acts as a catalyst of behavior change.

When using data to make decisions, identifying causation is king, and the best way to determine cause and effect is to run controlled experiments. By making slight changes to UX and measuring effects across KPIs, data scientists can identify opportunities to improve the experience for every user on your website. When armed with information, key decision makers can confidently make bold, but positive changes to your UX.

Of course, figuring out what to measure and what to test is hard. Raw numbers fail to tell the full story of customer experience. For that, you need a system dedicated to measuring and understanding customer experience. You need a way to understand the user experience beyond the abstract raw numbers.

And that's where FullStory Data Export can help.

☞ How to use this guide

The Guide to Data Export comes in two parts. First is an Overview of Data Export (You are here). Second is a Walkthrough/Case Study on how to build a Feature Usage Dashboard.

The Overview (You are here) will discuss common pain points observed in traditional CX data analytics. Then, it will leave you with a sense for how FullStory Data Export can overcome those challenges.

The Walkthrough will show you in detail how to put the Data Export Pack to work to build a Feature Usage Dashboard. It includes detailed descriptions of how to use our API to load data into an analytics tool and then use SQL or Python to transform indexed event data into inputs for statistical analysis or dashboarding.

If you are a consumer of dashboards rather than a creator, we recommend sending the Walkthrough to someone in a data analyst position at your organization.


The real problem(s) with current CX data

You know that everything users do on your website or app creates information that can be analyzed. You also have tools upon tools to analyze and diagnose that information. So why don't you have enough information to run the right analyses and get the answers you so deparately need?

You have incomplete, imprecise data

Chances are you are already using tools or gathering data to understand how many users engage with a feature or add to cart each day. If you are, congratulations on taking the first steps to better understand your product through the eyes of your customers.

However, if your tools serve up pageviews, click paths, and heat maps as a proxy for user engagement, your understanding will be necessarily limited, abstract, and lacking in empathy. While these types of metrics are great at answering questions around feature discoverability, they fall short of conveying the breadth and nuance of the user experience. They also can easily fall into the trap of being vanity metrics or analytics theater.

What's more, without the appropriate context around the numbers, data scientists and others who rely on analysis will lack the necessary confidence required to suggest bold changes to improve the customer experience on your site or app.

What's needed is a comprehensive, all-inclusive set of information about the customer experience online—a dataset with enough detail and nuance to act as a foundation for analysis.

Is such a thing even possible?

You're limited by a “bucket of events”

Too many analytics tools charge you by the number of events you want to track. You get a "bucket of events," and that's that. This approach can be frustrating because it's nearly impossible to predict accurately, in advance what customer behavior is going to be significant or not. For example, If you’re an eCommerce site, you can’t predict which interactions taken on a Product Details page will correlate with Purchase.

Why does your analytics software expect you to have advance knowledge about the interactions you want to record? If you already had these answers, would you even need the tool? Managing a “bucket of events” is fundamentally broken.

Another problem with the “bucket of events” approach is you need to take time to instrument and implement the events on your website before you can record them. Depending on your deployment cycle, how many engineers you have sitting around twiddling their thumbs, and the complexity of your site, this process could take months. Even if you are able to deploy the scripts quickly—say in a day—every time you make a website change you’ll have to implement your analytics tracking all over again. This is one of the hidden costs of analytics that quickly adds up.

There are millions of potential interactions on your web property. Why look at data that knowingly excludes 98% of it? Why be forced to sit in the back of an engineer's queue to re-instrument events just because you made a simple site change? Without capturing all of the data, all the time, can you ever be confident you have the capability to thoughtfully answer new business questions?


FullStory Data Export Pack Has the “Right Data”

FullStory solves the problem of CX data by offering all the data without limits on what is captured.

Our Data Export Pack not only has all the data, it has the right data. The Data Export Pack from FullStory contains ALL of the indexed customer interactions on your website—the “bucket of events” problems mentioned above simply do not exist. But what data gets collected and indexed?

The types of events FullStory indexes are:

  1. Clicks — When the user clicks on something on the page
  2. Navigates — When the page / url changes
  3. Changes — When text changes on the page (i.e. typing into a form field)
  4. Abandons — When the user abandons a form
  5. Thrashes — When the mouse moves erratically or in circles

Anything that you can search for in the FullStory app ties back to one or more of these events. The Data Export pack provides these events to you in an easily readable JSON file. Indexed events are considerably more valuable than raw data or log files as both formats are difficult to make sense of and expensive to query.

Each event recorded in FullStory includes with it other information besides the Event Type, including:

  • User / Session Fields — data that uniquely ties an event to a specific user and session that can be looked up in FullStory.
  • Page Fields — all the information about the event, including the event type, the URL it was recorded on, and the time the event was recorded.
  • Environment Fields — information about the user’s client, including the browser, device, and operating system.
  • Custom Variable Fields — information from your other systems can be passed in and stamped on the indexed events (like A/B test variables and account information). More on how to do this here.

If you are curious about which specific attributes are available in the pack, this support documentation outlines every field that will be included in your export.

How the JSON file works

What does all that mean? It means that you are not just looking at a ‘click’ on a button, you are looking at a click on the purchase button of a specific product from a mobile device from a customer that has never purchased before and came to your site from a new AdWords campaign.

The depth of data provided by FullStory Data Export will have your data science team chomping at the bit to get started.

Example of a single event in JSON format.

If you have not analyzed data in the JSON format before, it may seem scary. Worry not, it can be deciphered quickly or converted into a more familiar format if desired (E.g. CSV).

For example, take the picture above. This “JSON blurb” would represent a single row of data in an Excel sheet. Purple text represents the “name of the column” and the orange text is the value for that specific row. Therefore, a single session will be represented by up to hundreds of these JSON event blurbs, but they can all be easily tied back to a sequence of actions taken by a specific user in a specific session.

Custom variables in FullStory

The attributes that come standard on the Pack (those listed above) are a powerful source of customer experience insights in their own right, but you can extend that power through enriching the data with your own custom variables. Custom variables are the meta data about your users that is unique to how your team analyzes and understands your business. They provide added context to your data analysts when viewing CX data.

Custom variables work through our API call FS.identify(). Using FS.identify() customers can pass data to FullStory from their CRM systems, attaching that data to all events from a certain user. While custom variables could be just about anything you could imagine, here are some common custom variables/categories to help you imagine the possibilities:

  • Customer/User ID
  • Account information
  • A/B test information
  • Customer aquisition channel
  • Lifetime spend

For each custom variable you pass through, you'll find key:value pairs in the JSON blurb. E.g. in the JSON blurb above user_segment_str, user_testGroup_str, and user_lifetimeSpend_real are all custom variables.

Attaching custom variables to your FullStory data ensures the customer interaction data captured can be segmented and analyzed in accordance with your team’s needs. For example, you can prioritize your analysis on the accounts/customers that matter most to the success of your business.

Getting your data exported

Lastly, you probably want to know how you get all of this great data once you bought the pack. The two broad means for getting your data exported come down to:

  • Download the data directly within the app (Anywhere from every 30 minutes to every 24 hours)
  • Utilize our API and export the data into an appropriate tool (E.g. Amazon RedShift or Google BigQuery) to prepare it for analysis.

NOTE: For in-depth how-to on using the above methods, see the Data Export Walkthrough.

Who is the Data Export Pack for?

In a sense, everyone at your company can benefit from the Data Export Pack, but not everyone will be actively using it. This Pack is geared toward companies that have a business intelligence or data scientist resource—someone who is comfortable with SQL and/or isn’t afraid to get their hands dirty with Python/R.

That said, your Data Export resource does not have to be dedicated to it full-time. There are good use-cases for doing project based work. You could, for instance, build a dashboard to help your Customer Success team see how their big accounts are using the product. This would require about a week for a data analyst to set up but would provide value on an ongoing basis.


5 Ideas to Put Data Export to Work

Use cases for the Data Export Pack are limited only by your team's ability to ask testable questions about your customer's experience. We encourage you to analyze and optimize for the things that make your company successful given your industry.

However, if you want a few examples for how you could get started with Data Export, consider the following and let your imagination do the rest.

1. Find what causes conversions

Data Export can be used to identify the critical levers that drive results.

Many of our customers express interest in wanting to use statistical analysis or machine learning to uncover better insights from their customers. A sample question they have is “How do actions taken on my homepage affect user conversion?” With the Data Export Pack, you have access to not only the interactions customers take, but also the characteristics like device type, active time on page, and other information that can help feed your machine learning engine.

Using a “conversion” as your desired outcome, you can attribute characteristics to a user session (E.g. came to the site from Google, # of Product Pages Viewed, # of Visits this week, etc.). From there, you can run logistic regressions to determine which of those characteristics are significant in driving conversions. This regression analysis will uncover the characteristics correlated with conversion. Using your findings, you can run experiments to confirm hypotheses getting you closer to inferring causality. Even without experiments, in many cases, correlation can be helpful in your journey to making better business decisions.

2. Fraud/suspicious behavior detection

Data Export can be used to identify potentially criminal behavior.

Data Export can also be used to identify fraud/suspicious behavior on your website. Use the Data Export Pack to build or augment fraud monitoring algorithms.

Customers who use Data Export to look for suspicious behavior identify things like rapidly switching tabs to complete tasks or filling out form information at superhuman speeds. Data Export can be more efficient than raw log data here because we also have the IP Address, User Agent, and Lat/Long, all of which can help build more accurate fraud detection models.

3. Building behavioral cohorts

Data Export can help you identify from all your sessions the cohorts you need to follow.

Consumers are frustrated with the lack of personalization they receive from retailers’ marketing emails, and as this WSJ article points out, only 2 out of 5 shoppers say the information they get from retailers is relevant to their tastes and interests. Bombarding people with irrelevant emails they don’t want is a quick way to create a bad customer experience.

Using the FullStory Data Export Pack you can build segments of customers based on the actions those customers take on your website or app. Using these customer segments, you can better target your marketing and create the kind of dynamic experiences customers want—the kind of customer experiences that will improve your brand.

4. Task completion time

Time is everything for your users. Data Export can help you analyze opportunities to improve task completion time.

Another ecommerce use for the Data Export Pack is “task completion time” analysis—i.e. analyze how long it takes users to add to cart and how long it takes them to complete the checkout flow. Task completion time analysis can help you determine how changes to your website are affecting the time it takes users to complete certain tasks.

Task completion time is useful right out-of-the box, but where it becomes even more helpful is in measuring those numbers over time, across A/B groups, and between mobile and desktop devices.

You can use task completion time analysis to prioritize engineering and design time for site or app improvements (E.g. if it is taking mobile users forever to check out). Other things you can use this analysis to determine:

  • What are users who spend a lot of time on my website but fail to purchase doing?
  • When users want to purchase quickly, how easily does my website facilitate that transaction?
  • How long into a customer’s session before they turn to chat for help?

There are many other use-cases for task completion time, and using Data Export, you can build out these analyses whenever they are required.

5. Build a feature usage dashboard

Have you ever wanted to know just how many of your users are adopting your latest feature? Or how feature usage breaks down across your user base? If so, Data Export can help you build a "Feature Usage Dashboard." In fact, we've already built one that we use internally at FullStory.

If you want to get an understanding of what's possible for this use-case, hop over to our Walkthrough.

A glimpse at our Feature Usage Dashboard. You can learn more about this by heading over to our Guide to Data Export Walkthrough: Building a Feature Usage Dashboard


Conclusion and Next Steps

You should now understand conceptually what FullStory Data Export provides—and have a few ideas to put it to work. That may be all you need to get started.

If you're ready to go deeper, you'll want to take a spin through our Data Export Walkthrough, which is effectively the advanced "Part 2" of this Guide. Therein we demonstrate how to take FullStory's raw data and convert to a readable format, how to identify unique sessions from the raw data, and, finally, how to transform the data for inputs into data visualization tools and statistical analyses.

You might not need that level of detail. If that's the case, send the Data Export Walkthrough on to the right person on your team who can get you up and running.

What CX questions could you answer with Data Export?

Taking the first steps in analyzing customer experience data can be difficult. Our hope is that this overview will inform you on the type of data available with FullStory and our Walkthrough will help your data analysts get their feet wet in extracting value from said data. From there, they will be able to ask their own CX questions, transform data appropriately, and test their hypotheses.

Data analysis is a journey, not a destination. You will discover different aspects of how customers experience your website than you set out to test, and that is okay. As a data scientist, the computer is doing the number crunching while your task is to constantly refine your inputs and ask the right questions.

Now, with Data Export you can ask the right questions and know you'll have the data you need to answer them.

Last updated November 2018.