CARE Bangladesh launches chatbot on WhatsApp to promote immunization

Creative process

Content development for the chatbot began with an assessment of what information surrounding immunization was most important for people to access. All content was locally designed by country team staff in Bengali, the local language.

“We followed the approved messaging from the long ongoing successful EPI campaign of the Bangladesh Government,” said Priyotosh Das, CARE Bangladesh’s Communication Officer. “We planned to inform the audience about the main topic and then gradually move on to the importance and access-related messaging.”

The chatbot launch was supplemented with a Facebook ad campaign because, as CARE learned through a previous project, chatbot adoption and engagement can be significantly boosted through social media advertising.

The ad campaign consisted of locally designed static and video ads that used engaging animations, trusted messengers, and persuasive copy emphasizing that immunizations were free. Each ad was linked to the chatbot.

The vaccine calendar ad pictured below provided valuable information on when children needed to be vaccinated and generated the highest number of post engagements (totaling 249,800) and shares (13,200) throughout the entire campaign. It also led to the start of 47,700 chat conversations.

The ad below, which gives examples of locations where vaccinations are available, drove the highest conversion rate from ad to chatbot conversation.

A survey feature within the chatbot was designed to serve as an in-platform measurement tool; however, in the initial weeks of the campaign, user engagement with the survey was low. “Also, initially, we noticed users mostly interacted with just the first message and did not go into deeper conversation with the chatbot,” commented Das. “We addressed this by developing instructional videos which showed step by step on how to use the chatbot optimally.” An additional ad campaign with a survey call-to-action was also developed.

Results

While ads for this project ran from December 13th, 2022, through to January 22nd, 2023, CARE Bangladesh’s ad campaign was divided into three phases. The first two phases tested the effectiveness of different Facebook advertising campaign objectives and the final campaign retargeted the previous audience and promoted the chatbot’s survey feature. Campaign durations were as follow:

• Phase 1: December 13th, 2022 – December 23rd , 2022
• Phase 2: December 23rd, 2022 – January 29th, 2023
• Phase 3: January 11th, 2022 – January 29th, 2023

Notable metrics averaged across the three campaigns included:

• Total reach (Phase 1 + Phase 2) was 34.2 million people. Within this group, we reached 1.1 million users in our Phase 3 retargeting campaign.
• Post engagements totaled 662,722 and had an overall share rate of 0.083% (which is higher than our benchmark).
• Overall conversion rate (CVR) to chat was 6%. This is lower than our benchmark possibly due to the audience’s inexperience and hesitation with communicating with an automated chatbot.
• Each campaign phase had varying CVR rates, with the strongest CVR observed when chatbot users were retargeted in the final campaign.

Results from the chatbot and its user engagement include:

• A total of 123,918 conversations were started in the CARE Bangladesh chatbot
• 14% (17,318 people) opted in to receive future immunization-related messaging from the CARE country office
• Conversation lengths averaged around 5.7 messages per user

The screenshot below shows an example of CARE Bangladesh’s opt-in message enabling users to choose whether they want to stay connected for future communication from the organization.

What did users think about the chatbot?

CARE Bangladesh’s in-app survey yielded important insights into how users perceived the chatbot service. The survey had a completion rate of 77%, meaning that of the 2,057 users who accessed the survey, 1,590 fully completed the survey.

Some results include:

• 89.7% of people reported that they would recommend this service to others, while 4.8% said they would not recommend it (the remaining 5.5% were unsure)
• 75% of people reported they were fully satisfied with the chatbot service
• When assessing understanding about immunization importance, over 75% of people expressed an increased understanding through using the chat service

CARE Bangladesh’s team also conducted an in-person endline survey in Gazipur with healthcare facilitators and target users. Some of the qualitative responses that were collectively shared included:

• Facebook Messenger is more widely used among the population, so some people either didn’t have WhatsApp or didn’t want to download it when they already had the Messenger app.
• There was a sense of initial distrust with the chatbot, given the novelty of the service. Women commented that they used their husband’s phones to message the chatbot since they didn’t trust the service initially.
• There was also a technical barrier to learning to use WhatsApp in an automated chatbot format as opposed to more personalized 1:1 messaging that users are used to on messaging apps. Many people thought that they were conversing with a human being in a real-time live chat and were expecting a response from the other end to their personalized questions.

“There was a 24% increase (from baseline to endline) in the understanding of EPI which indicates significant increase in understanding the importance of EPI and its accessibility,” said Tanzin Labonno, Senior Technical Officer-Monitoring & Evaluation, Urban Health unit of CARE Bangladesh. “Qualitative assessment could be helpful to understand the effectiveness and acceptance of the chatbot in the communities to improve it further to cater to people’s needs.”

What’s next?

CARE Bangladesh is committed to sustained deployment of its chatbot on WhatsApp and hopes to integrate diversified program activities and messaging into the platform. The country team, supported by CARE’s Digital Inclusion team, plans to onboard contact details of 60,000 garment factory workers from its Emergency Communication Management (ECM) project to be included for future messaging services in maternal and nutrition health information.

Over the course of the next year CARE Bangladesh hopes to launch monthly content audits and add diversified content subject areas (i.e. maternal health, climate response, feedback mechanisms) into chatbot content workflows.