Conversions - Tips & tricks

Looking for a deep dive on Conversions? Check-out the 5 part series designed to provide more details about the Conversions product. Use this article as a high-level to review some Tips & Tricks to help you get started with Conversions. 

In this help article, we'll review the following:

• Segments vs. Conversion Funnels
• Best Practices for Custom Events and Watched Elements in Conversions
• Do's and Don'ts

We'll also answer some of our most frequently asked questions, including:

• What happens if there are no Opportunities in the table?
• How should I make sense of all the math?
• What should I do with these Opportunities?

Segments vs. Conversion Funnels 

Segments should be used for defining audiences of people that have done a certain thing or have common properties, while Conversion Funnels are meant to specify a series of events that occur in a specific order. Conversion Funnels need at least two events. Best practice is to have four to six events in your funnel.

Existing Segments in Fullstory can be used to define the cohort of users in Conversions using the Performed By filter. If you have a Segment of users that you want to use in the Performed By filter, make sure to save the Segment before using Conversions. 

As you are building out the events in your conversion funnel, keep in mind that users can fall out of the funnel after any event, so make sure your first step in the funnel is not too broad. As you are investigating an opportunity and watching sessions of users who had a specific experience, but failed to convert, Fullstory will highlight where in the session the opportunity occurred for the user. 

Best Practices for Custom Events and Watched Elements in Conversions

If your Conversion Funnel uses Custom Events to define steps in the funnel, be sure to also include the “Visited URL” steps so that Page Performance on that URL will be considered as a signal for that funnel. In other words, if you define your entire funnel using only Custom Events without any specific page navigation, you will not see any Page Performance Opportunities for that funnel. 

Additionally, Custom Events and Watched Elements can be used as signals to pinpoint errors or interactions that the user may see or experience throughout their session.

For example, you could use either of these to look for opportunities related to a phone number field validation error or an “out of stock” alert at checkout.

Sometimes Custom Events can be related to an experience that is not inherently negative, such as seeing a coupon pop-up. If you want to understand whether that Custom Event had a negative impact on the Conversion Funnel, you can configure it as a signal to test that hypothesis. Conversions supports string and boolean property types for Custom Events.

For Watched Elements, there is a limitation to the types of Elements that can be tracked. They must have text and must not include dynamic text. If not, these elements will not show as Opportunities for the conversion funnel.  

If you don’t see the option to add Custom Events or Watched Elements in your Conversions Signals configuration window, this means you don’t already have any set up in your account. Check out these guides for how to get started setting up Custom Events and Watched Elements.

What happens if there are no Opportunities in the table?

If you don’t have any relevant Opportunities listed, you might need to adjust your funnel and analysis configuration in multiple ways to find the most useful setup for your data. Here are some tips that can guide your experimentation:

• Review and adjust your funnel steps - Check to ensure the number of users in each step of your funnel seems right, and adjust the step criteria if needed. We recommended using 3-5 steps that cover a linear flow. If there is an exceptionally large drop-off between two steps, split it into two separate funnels. For more ideas, check out our in-depth guide to creating funnels for Conversions. 
  

• Narrow the analysis down to a particular Segment - Your site may have a very different user experience for different Segments of users, like desktop vs. mobile, iOS vs. Android, Chrome vs. Firefox. Try limiting the analysis to a specific Segment to find signals that only affect that sub-population of your users.
  

• Change your date range to be smaller or larger - You may need a longer time range to capture enough data to find statistically significant opportunities, or you may need a shorter time range to pick up a signal that was recently introduced but hasn’t been happening across your full time range. Keep in mind, if you select a custom date range in Conversions, it must be 30 days or less.
  

• Add custom signals - our out-of-the-box signal types can be great to get you started, but what if your site already loads fast and is free of Rage Clicks, Dead Clicks and Error Clicks? Try adding Custom Events and Watched Elements to pick up signals unique to your site, like checkout errors.
• Broaden your funnel - If your analysis configuration is too restrictive, you may end up with very small volumes in your funnel, particularly in the final steps. A good rule of thumb is that if those numbers drop below 100, it will be more difficult to determine confident correlations for Opportunities. In this case, try adding a larger number of users to your funnel by broadening the date range, segment definition, or criteria for the last steps in your funnel.  

In general, iteration is key when creating funnels in Fullstory. It's not uncommon that you might have 3 or 4 different iterations of a single funnel aimed at identifying what’s inhibiting progress from one step to the next vs. through the entire funnel (in fact, we suggest this approach). Consider both micro and macro conversions of a user’s experience, like selecting a plan (micro) and completing checkout (macro).

How should I make sense of all the math?

The conversion impact number summarizes how much worse the affected group converts compared to the unaffected group within the steps of your funnel where the signal was found. Something to keep in mind is that the conversion impact is a relative percent change and not an absolute difference (i.e. subtraction) in the conversion rates. Opportunities are based on calculating statistical correlations.

Next to the conversion impact, you will see the confidence level. If the confidence level is 95%, that means you can be 95% confident that the result is due to an actual underlying difference in the two groups (affected & unaffected), not just random chance. But correlation may not mean causation: it is important to consider what upstream behavior could be the root cause of the impact.

By default, we prioritize the Opportunity list by the total number of conversions lost. However, you can adjust to sort by affected users, conversion impact or confidence.

What should I do with these Opportunities?

Try getting started by watching a related session to get that additional context and clarity for each of the listed opportunities. From there, you can note and share sessions with related teams when you’re ready to make improvements based on the opportunities you discover. Come back to Conversions as often as you are shipping code to see new opportunities. Adjusting the date range will help you see how Signals were impacting conversion rates before and after implementing changes on your site. 

Do’s and Don’ts:

Do use Conversions when you …

• have an unknown unknown and don’t know where to start making improvements
• have a specific hypothesis that you want to test
• need to see what is having the biggest negative impact on your funnel

Don’t use Conversions when you’re trying to understand …

• positive impacts on conversion—it’s currently only for negative impacts
• single, specific examples for a single user