top of page

Revamping Traveloka's Bot (IVAN)

Project info

Role in the project: Design PIC, Sr. Product Copywriter

Project nature: In the fog

Project period: 8 weeks

Team members: Product Manager, Interaction Designer, Visual Designer, FE, BE

Project involvement

How did I help as a PIC and Sr. Product Copywriter?

​

  • Create design brief & present the design brief during kick-off meeting

  • Lead initial discussion with product team

  • Assess project scope and limitation with product team

  • Hygiene/administrative tasks (set up checkpoints, lead checkpoints, manage stakeholders, etc)

  • Lead brainstorming session for bot conversation guideline

  • Lead design direction approval

  • Create bot conversation guide

Background

CHANGES FOR BOT IS LONG OVERDUE. MULTIPLE ISSUES OCCUR EVERY DAY.

Screen Shot 2024-05-19 at 13.33.57.png

How might we improve IVAN's flow and response and migrate all of IVAN’s visual components onto the new platform?

DESIGN CHALLENGE

Timeline & design deliverables

As a Design PIC, I was responsible to manage the timeline given by the product team.

 

Looking at the given time and the nature of the project, I decided to split the work into two streams:

​

In the fog: Stream 1

Focuses on creating the conversation framework (that includes new flow, copy, and visual) as a way to improve IVAN's flow and responses.

 

Paint by Number: Stream 2

Was more straightforward — focusing on how can we migrate and integrate our existing visual component to the new platform.

Screen Shot 2024-05-19 at 13.49.35.png

Design process

  • Define design principles

  • Desk research (book & internet)

  • Craft the content

  • Refine

  • Review & iterate

Design principles

Screen Shot 2024-05-19 at 14.00.26.png

Desk research

Since the team was quite unfamiliar with bot (IVAN was not made with ideal solution in mind), each role decided to do their own desk research. As a Copywriter, I did desk research in relation to conversation design (this is also new in Traveloka, right?), conversation framework, and how bot conversation is ideally designed.

books_bot.png
🤖 New Bot Implementation (Wall-E) _ Working Doc.png

Other than books, I gathered information about conversation design and bot from existing material for IVAN — even though they were not that ideal. I read Bot Content Guide and Bot Content Improvement by Abel (previous CW who worked on IVAN).

​

 

Upon reading the existing content guide by the previous Copywriter, I found that we didn't have comprehensive conversation guide especially on how IVAN should "talk" to our users. The existing materials only talk about how bot content PIC should write answer to users and how bot content PIC should document the proposed content or copy.

​

 

Therefore, my design goal became clearer ⤵

How might we create a comprehensive conversation guideline which complements the proposed improved flow, visual, and copy accompanied by examples?

Screen Shot 2024-05-20 at 11.18.04.png

Above is the table of content of the conversation framework for IVAN, our dear bot

While crafting the content for the conversation guideline, Daru focused on section two (IVAN's new personality), and I tackled the remaining sections. I made this call based on perceived scope, that defining the personality would be a lot more than other sections. Turns out, I overestimated the depth of crafting IVAN's new personality. What initially seemed like a a multi-layered exploration, eventually it could be done within days.

 

Here's the current bot conversation guideline. (unlinked, confidential).

The next step is to do roadshows to business units where I will introduce the guide so the SABP and Ops can craft better content for IVAN. Simultaneously, I will also collect feedback from the stakeholders regarding the guideline so I can iterate the current version, finalize it, then craft the ID version (yes, the current one is only in EN).

Impact

Note: data compares the week before release (21 Apr to 27 Apr) and the week after release (5 May to 11 May) – Release week (2 May) is not considered.

 

Solve by Bot (higher better)

  • Increased 5.01pp from 38.00% to 43.01%, increasing the baseline by 13.18%

 

Recontact Parent Case (lower better)

  • Decreased 4pp from 31% to 27%, decreasing the baseline by 12.90%

 

% Solve by Bot (not Recontact) (higher better)

Increased 5.40pp from 26.10 to 31.50%, increasing the baseline by 20.69%

 

Overall, in all categories, the post-release results have shown good performance, moving the needle towards favorable outcomes.

bottom of page