A large number of websites have a chat feature to interact with their visitors. This allows them to
- Offer real time assistance and advice
- Promote a service/product to the visitor
Some chat applications offer integration with web analytics tools. The Google Analytics blog gives an example of such a tool here.
We tracked some chat conversations for a client using Google Analytics and analyzed the data obtained. Here is what we found.
Visitors to a site (xyz.com) can discuss their technical problems by chatting online with a support executive. The goal of this chat application is to solve technical queries as well as persuade the visitor to sign up for an annual service package.
Why is tracking important?: We recommended tracking ‘chats’ to analyze
- Paid/Organic keywords resulting in chat
- Locations from where people initiate conversations
- Contributions of chat to conversions (sign ups)
Implementation: Clicking on a chat button on the website opened a new window, on a third party domain. This domain did not allow placing the Google Analytics Tracking Code (GATC) thereby ruling out cross domain tracking.
We recommended putting virtual pageview code linked to the ‘on click’ event of the “Chat Now” button. “Virtual pageview” was suggested instead of “event tracking” as we wanted to carry out a funnel analysis on the conversations.
Analysis: A funnel was set up in Google Analytics which tracked clicks on the chat button – from initiating a chat conversation to signing up for the annual service package. A majority of the conversations originated from a particular segment of the visitors, for which this analysis was done.
On analyzing this funnel we realized that the first step itself was a bottleneck in this funnel as only 9% of visitors proceeded to the next step (Registration Page) after the chat session. (First two steps Shown in Fig 1)
Fig 1: Funnel Report – Pre Analysis
A high drop rate could also indicate a technical issue with the chat application like,
- Long loading time.
- Long response times while chatting due to issues with the chat server.
Both the possibilities were evaluated and discarded after conducting several dummy chats from our side.
We investigated this matter and it was concluded that an inefficient sales process at the client’s end (during the chat) was causing the visitor drop off.
We recommended better sales training to the online support staff so that chat visitors could be converted at a better rate. Here are some of the inputs shared with the client:
- Asking for a phone number does ensure that the visitor initiating the chat is genuine, but this information should be asked only after a certain level of trust has been built during the chat session. The visitor needs to be assured of company credentials, only then he/she will be comfortable in providing their phone number. Building trust with a visitor could be a time consuming process.
- Analyze the chat scripts of all successful sales and then build a “model script” based on that.
- Identify the best chat representative (in terms of conversions) for managing shifts.
Results: Here is the funnel report after implementing our suggestions. (The duration of the funnel is same as Fig 1)
Fig 2: Funnel Report – Post Analysis
The results are tabulated below:
|Metric||Fig 1||Fig 2||% Change|
|Proceeded to Registration Page||9%||16%||78%|
The average time spent on chat increased by 30% after implementing our suggestions. It indicates a better engagement with visitors (as building trust takes time).
The above experiment demonstrates the use of web analytics for finding and fixing issues in the chat conversion process. Here the issue is not related to website design or usability but can be addressed using insights from web analytics.
The analysis helped the client to increase their ROI. We look forward to sharing similar experiments in the future.
Contributed by Ravi Shukla, Analytics Team