Importance of Analytics
The magic of Frase Answers is not that it's capable of answering 100% of the questions your site visitors have (though don't get us wrong, that's still awesome).
The magic is that Frase can help you analyze and understand questions your audience is asking, where they're asking them, and when.
Every interaction is a chance to learn more about your site visitors. In this guide, we'll touch on how to use Frase to analyze the intent of those site visitors.
More specifically, we'll cover:
- Understanding the Activity Overview Data
- Tracking Self-Service Resolutions
- Upvotes/Downvotes and Topic Gaps
- Understanding What Your Audience Is Asking About
- Pinpointing Where on Your Site Questions Are Being Asked
- Filter Data and Pull Reports
Understanding the Activity Overview Data
The Activity overview shows total questions, total unique visitors that asked questions, click-through rate, email capture rate and more. The dropdown next to "logs" at the top of the page will allow the selection of a certain timeframe -- "last month", "last day", "last week" or "last quarter".
Navigate to the "Analytics" tab from the left side menu to access the Analytics Dashboard for your chatbot. By default, your Analytics Dashboard will open the "Overview" tab.
This view allows you to understand the high-level performance of your chatbot over time.
These performance metrics include:
- Total users: the total number of unique visitors that asked a question
- Total questions: the total number of questions asked
- Answer Click-Through: the number and percent of Assistant Answers where the visitor clicked on a call-to-action
- Emails captured: the number and percent of answers that captured a visitor email
- Live chats: The number and percent of answers that resulted in a live chat
Tracking Self-Service Resolutions
Frase helps you track what we call "Self-Service Resolutions." We define this as anytime your Assistant provided an answer that was trained as correct or upvoted by the visitor.
Using an industry-standard $5 cost per resolution (which is defined as the cost it would have required for a human to answer the question), we calculate your approximate total cost savings over your selected time frame.
Upvotes/Downvotes and Topic Gaps
Your Assistant won't always get it right. Oftentimes, the negative ratings will be higher than positive since it's much easier to downvote an answer that looks incorrect, whereas if the answer looks right, you'd navigate to the page (without giving an upvote).
Frase makes it easy to track this:
- User Ratings: After every answer your Assistant provides, your visitor has the opportunity to upvote an answer if its helpful or downvote it if it's not.
- Topic Gaps: Depicts the % of answers that your Assistant (and by extension, your website content) wasn't capable of answering confidently. When this number gets too high, it's a clear indication that your assistant needs a little training (or you need to create more content on your site).
Understanding What your Audience is Asking About
This is where it gets fun. Frase uses Natural Language Processing (NLP) to analyze and categorize the questions your audience is asking.
The list of top topics is sorted by occurrences. You can break down the list by topic on the right. Click on the pencil icon to quickly train an answer for a specific question.
Pinpointing Where on Your Site Questions are Being Asked
Frase makes it easy to track both where on your site questions are being asked and which pages your visitors are clicking through to.This can shed critical insight into what your highest-value web pages are.
URL Locations are the pages where your assistant gets most engagement. CTR Destinations are the pages your assistant sends traffic to as a result of a click-through.
Filter Data and Pull Reports
At the top of Analytics, click the dropdown next to "logs" to choose the timeframe you'd want to filter by. Click the download button (boxed in green in the screenshot below) to export a report to Excel. Click the filter button (boxed in purple) to apply more filters to the data - note: these filters (except keyword filter) applies to the view itself, not the report.