Verybuy | World’s 1st Bot with Multiple Brand Personalities
Execution & Project Management
#eCommerce #Bot
Creative Team: Kenie Kwok, Frankie Luk, May Seow
Bot Developer: Bot Bonnie
This is a self-initiating project. We’ve seen more and more eCommerce adopting Messenger bot as a part of their customer servicing. Yet, unlike programmatic ads which deliver most related content to individual target audience for optimised performance, everyone receives same responses, with identical tone and manner, no matter who they are.
Our hypothesis is that bot with personalities would outperform a generic robotic tone, driving more conversions & engagements throughout the messaging experience. And we decide to test it out with Verybuy, a Taiwan fashion eCommerce business.
Get to know the shoppers
In the time when we can literally buy anything online, experiences is the reason why we still go to bricks and mortar. The window display, the smell, and most importantly, the human interaction. Shopping feels better when the salesperson understands you well. And when you feel good, you spend more. Thus, we kick off the project by digging into VeryBuy’s existing data and identified 3 types of customers who have the highest potentials to convert. We worked with client and built up their distinctive persona with MBTI tool. Fun and useful exercise.
Shape the bot sales with people-first strategy
Based on the TA persona, we explore different tailor-made brand tones in responsive to the 3 types of customers. Instead of traditional advertising copywriting, we turned to seek help from clinical psychologists to deep dive into the personalities of these communities— eventually defined 3 corresponding virtual salespersons to best represent the brand.
Construct user experiences with 3x tones
Based on the BAU Bot, we asked 2 award-winning Taiwan screen-writers Li Nien-Hsiu (李念修) and Chang Yi-ning (張逸寧) to design and craft three different steam of craft the dialogues in the Bot to be more human, authentic & compelling.
Compare BAU bot with personalised bot
With conversion as the KPI, we set up a spilt test to test 3x personalised bot (as Cell 1) against the BAU (as Cell 2), and we rain into epic fail right away.
After studying the results, we realised the test cell has more steps than BAU, which created frictions for TA and hence the high drop-out rate. We quickly optimised the user experiences and creative assets, and kept variables other than the testing item the same.
We quickly see positive results and all personalised bot ended up outperformed the BAU in terms of engagement and conversion: both CPA and CPC lowered by double digit. Our hypothesis is proved. This test went beyond just a campaign and influenced the product.