Explore

Labina's Portfolio

Improve the capability of Tiktok Shop Chatbot

Anchor

Description

My roles

Product Designer & Research, Planning & Strategy, Prototyping, Leadership

Teams

3 Product Managers, 1 Operational Teams, Tech/RD Team

Timeline

1 - 1,5 months, Q3 2024

Summary

By this initiative, I working closely with Principle/Staff in defining the design strategy, led all the design process, prototyping to decrease bad experience and redesign overall tracking page. I breakdown the problems, identify the spot of improvement and translate it into actionable design goals and present the design proposal to VP of design and Logistic stakeholders. Successfully decrease 12% fulfillment customer service cost, decrease 11% order delivery complain rate, and increase 3% order delivery customer satisfaction.

There are no rows in this table

⁠

Why is this important?

TikTok Shop has implemented a chatbot service to support customer inquiries and reduce the load on customer support agents. The service is supposed to be helpful and relevant for the buyers.

What went wrong?

Although the Chatbot is already established in Tiktok Shop service, however, some issues still arise on some chat topics or modules such as logistic and delivery, also the communication confusion between buyer and seller which these add more dissatisfaction and affect our overall CSAT.

17% of dissatisfaction reasons were due to difficulty of the buyers reaching a live agent.

Logistics BOT consultation satisfaction is about 40%, which is about 15-20pp lower than the average level.

About x% of the users are dissatisfied because sometimes the BOT is not answering what they really need.

What should we do?

How can we improve the chatbot's design capability and reliability in to increase CSAT?

What is the solution?

Created scalable yet helpful BOT card design for various cases in Tiktok Shop.

⁠

The screenshot or gif of the overall flow

Plan & resolution

My design process typically different, depends on how complex or big the scope of the projects. In this case, the design iteration should go fast and well-targeted, so it can generate good metrics.

Thus I simplified my process to be these 2 mains activities,

Understanding → The problem/gap, the targeted users, the project scope, the team’s vision

Plan → This part is also quite crucial for

Design → Involving the individual work such as exploring the design, benchmarking to other similar apps

While in the middle of the process there will be always a collaboration since this project involved crossed functions (XFN).

⁠

00 Who are the users?

TTS

01 Audited real cases to find gaps and identify what matters most to users

02 Aligned with PM to refocus on top 3 pain points

...in the chatbot experience. Initially, the PM pushed to prioritize a new AI suggestion feature for higher business visibility.

⁠

However, I noticed from chat logs and CS feedback that users were still frustrated by basic interaction issues, like the bot misunderstanding intents or getting stuck in loops.

I presented this data, and we realigned to tackle the top 3 experience pain points first. It wasn’t the flashiest choice, but it improved retention and reduced CS tickets.

Top 3 pain points.

⁠

Users feels that shipping status tracking is unclear via chatbot → caused user’s frustration and repeated questions

Users feels that shipping status tracking is unclear via chatbot → caused user’s frustration and repeated questions

⁠

Chat access to Live Agents was hard to find → delayed issue resolution for complex problems

Chat access to Live Agents was hard to find → delayed issue resolution for complex problems

⁠

Chatbot gave irrelevant replies about products/sellers → BOT asked user to chat with Seller directly manually

Chatbot gave irrelevant replies about products/sellers → BOT asked user to chat with Seller directly manually

⁠

03 Designed scalable solutions driven by user intent

Because the chatbot reply is really depending on the user intents

Found some design opportunities by doing design benchmarking from other apps.→ Applied to some design (include features) explorations

Create user flow based on identified users behavior and emotions especially in situations like:

Rushed, anxious: they would track the shipping status especially for delayed delivery,

Curious: For new users they asked about the products to chatbot, not to sellers

Rushed, irritated, for complex problems user needs to easily contact CS to resolve their issues.

Conducted quick testing to few real users to gain insights (show insights later) → this influences the design iterations

As previously I already established the design principle, I revisited it to be the design compass.

04 Iterate: Discussion, Design

Aligned overall progress with the PM, Design Managers to exchange POV, receive feedback

Show design explorations for end-to-end cases

Plans changed, we needed to launch sooner. Here's how I handled it:

Reprioritized card feature launch -> speed up the development process and see significant impact. If not -> then iterate

Locked decision, we go for #problem 1 go live first for A/B -> lowest CSAT, low development effort

05 A/B Test

Dashes point -> supporting data, so what, how can this help

06 Apply to All Requirements

Next, reused the design pattern (based on the feature that works) for more scalable design & development

The #problem 1 design is performing good

Follow the design patterns

Impacts