Investigating User Behaviour with Website Vs. WhatsApp Chatbots.

Messaging applications and platforms are probably the most utilised features on our devices, therefore it’s only natural that Chatbots have now migrated and become a prominent fixture on these platforms. In April 2020, WhatsApp became the 4th most downloaded application in the world, so it comes as no surprise that more than 50 million businesses are already using WhatsApp Chatbots to interact with their customers (Source: FreshWorks). Automating processes and growing a digital presence has become a priority for most businesses as a result of the pandemic, and these messaging platforms have been a key channel for this.

Earlier this year, we conducted user research to uncover users’ platform preferences and behaviours when it comes to customer service chatbots. Through the research, we were hoping to gather deeper insights on users’ engagements with chatbots and the viability of implementing WhatsApp as a channel into our chatbot console. We ended up holding two rounds of user research around this topic to help us gain a comprehensive understanding of this behaviour.

The focus of our research was on the comparison in user behaviours among different devices and preferences, specifically between the traditional website chatbots and WhatsApp chatbots. A simple banking service chatbot was used as the premise for our testing scenarios with the participants. Testing the website scenarios was straightforward: we built a simple chatbot prototype for participants to use. However, for the WhatsApp scenarios, instead of building a WhatsApp prototype bot, we used the “Wizard of Oz” technique which involved a team member acting and mimicking the bot interactions unbeknownst to the participant.

The primary objective of the first round of testing was to collate data to determine if users preferred to communicate with a live agent on a website or a WhatsApp bot. Our secondary objective was to discover which of the two chatbots were more intuitive for users during the chatbots interaction. The testing session further improved our understanding of users’ expectations of chatbots and occasions when they would prefer to engage with a live agent.

In each session, participants were required to go through three scenarios that was drawn up with these objectives in mind. Participants experienced using a chatbot on both the website and WhatsApp in the first two scenarios, before giving them the option to choose between the platforms in the last scenario.

The second round involved participants solely using a mobile phone, and focused on understanding behaviour when using a bot on a mobile app vs a WhatsApp bot. We also wanted to observe whether users preferred to be contacted back by a live agent via a phone call or text. Once again, participants were required to go through three scenarios that involved interacting with bots that were both in an app and on WhatsApp.

Findings

Prior to each session, we had team discussions about the expected user behaviour based on our personal experiences and chatbot work, and drew up a list of hypotheses. We then analysed the research by evaluating the research findings against these hypotheses to validate the results and highlight areas for further research.

Our first round of research highlighted the fact that convenience was a driving factor for the users’ platform choice, specifically with regards to the device, location and waiting time. Without specific instructions on the choice of devices, all participants made the same decision and chose to use a laptop. This influenced their decision on which platform to use, as they all chose to use the website bot over WhatsApp when presented with both options. All participants stated that they decided to use the website bot because they were using a laptop, and it was more convenient. This highlighted a fact we didn’t consider before, which was the importance of the device used as a deciding factor.

Our second round of research built upon the initial findings by making the participants use a mobile phone throughout the testing. Once again, it was observed that users preferred to do their own research before reaching out to the bot, which aligned with the previous findings.

The main findings from this were:

• Users prefer to be contacted back via text instead of a phone call.
• When using the web chatbot on a mobile browser, users now preferred to move the conversation to WhatsApp when being told to wait for live agent assistance, as it was now more convenient.
• The in-app bot was the more popular option vs the WhatsApp bot, with the reason given by users was that the app was already downloaded.

What this all means

The findings and analysis from our research sessions proved that WhatsApp was definitely a viable platform for us to develop and integrate into our bot builder product. WhatsApp is a dependable platform that offers convenience, and as it is now widely used across several age groups, it is a familiar and trusted interface.

Although in-app bots were shown to be the more popular choice, it’s important to note that not all users may have the app downloaded, in which case they would choose to use alternative platforms. Also, as mobile apps are expensive to build and maintain, depending on the businesses resources it may be more suitable to use a chatbot on a messaging platform.

Looking into WhatsApp further, we also uncovered unique features of using Whatsapp over a website bot. For example, it’s possible to send out automated outbound messages, which can be used for promotional messages or personalised notifications for the customer.

Additionally, we also decided to implement and enable the live agent feature for WhatsApp. This means that a live agent can go to the Chat Agent page on our console and interact with the customers through direct messaging. This allows for businesses to offer a more personalised service with direct customer support.

Identifying the right cases

It’s important to note that while numerous statistics show an increase in chatbot usage and adoption, many users are still sceptical over chatbots’ practicality, stemming from their own poor experiences. Numerous companies are implementing chatbots with content that don’t serve users effectively, resulting in users’ frustrations.

Taking FAQ bots as an example, it is critical that the bots’ responses could directly address their enquiries, reducing the reliance on further assistance from live agents. Whereas if you want to implement a customer service bot, the live agent assistance should be included earlier on in the process flow to improve efficiency.

Finally, although this research steered towards investigating WhatsApp as a chatbot platform, these findings draw parallels with the Facebook Messenger platform due to the similarities in capabilities and customer usage. Therefore, it is both crucial and necessary to understand the demographics, habits and preferences of the target audience to effectively determine the right platform to situate the chatbot.