Adapting to AI: Redesigning Gengo's Translation Workflow
Timeline
2018 - 2019
Company
Founded in 2008, Gengo has translated more than one billion words for 65,000+ customers. A Tokyo-based, leading edge technology company providing crowdsourcing, machine learning and localization services to global customers.
The company’s primary goal of 2018 was to transform the service from human-based translation to machine-based translation. As a typical SaaS company, it serves the needs of multi-stakeholders, which requires sustainable and efficient design and product strategy.
Main Tools
Sketch, Keynote, Invision, Zeplin, Jira
Roles
UX Research & Design, UX Copy, Product Feature Specs
Key Capabilities
Order Form Redesign
The Problem
New business model to offer double-pass translation service: first translator, then proof-reader.
Current order form cannot afford the new business model, plus almost 96% users skipped the added service/ customize function on the order form (see chart below).
The Challenge
Designing a simple ordering experience of selecting the right type of service, yet with navigation flexibility, given the complexity of Double-Pass business model.
The Solution
After 2 iterations of business model designed with prototypes, my PM and I came up with:
2-Step-Form that coordinates with Double-pass. We recommend the user a type of service based on their needs:
2.Choose translation purpose
-> Plus Proofreading Service
1.Choose the content type
-> Standard or Advanced
Iterations
Before & After
Before: 2 service types -> Single-Pass
After: 4 service types -> Double-Pass
Learnings
Product Strategy and Design Strategy evolve hand in hand. Using designs to navigate through ambiguous feature request to find the optimal solution.
Design Principle helps greatly for a time-efficient decision between 2 good designs.
Pre-Edit Machine Translation (“PEMT”)
The goal was to add Machine Translation function into the current translation work tool such that:
Users know what is Machine Translated content (MT), what is Translation Memory-based suggestions, and what they supposed to work on.
But first, what are the existing problems?
How are the users using MT now?
The goal is to figure out what information users values the most and their task prioritization when adding Machine Translation (MT).
*TM: Translation Memory
Redesign with 3 focus areas
Pre-fill machine-translation results with tags for clarity and ease on the eyes.
Utilize “inactivate” and “activated“ status to direct attention on useful sidebar information.
Intuitiveness
Learnings
Whiteboard of feature focus and questions, it’s clear and linear, yet lack the connections to design initiatives.
Using a decision tree give a clearer structure.
Decision tree of design progressing questions help in terms of pushing design to serve all needs.
An interface with abundant functionality and visual hints requires prioritization on visual attention.
If time allows, prototype works a lot better than static designs with lengthy function specs both for documentation and communication.
Reference: https://www.nngroup.com/articles/which-ux-research-methods/