Process Improvement

Streamlining A Complex Process: Localization Handoff

Spring 2021

Background

Below are the steps I took in my role as the authoring team's liaison to the localization team to drive a difficult process change. This showcases my UX Research and UX Design skills and empathetic nature to understand what the team needed and design a successful solution.

Context and Environment

• Structured-authoring natural language generation (NLG) content management system for a large online travel agency.

• “Producers” author dynamic content products in English & prepare them for localization in 30 - 40 different languages. 

• Most English authors live on the West Coast while most linguists involved live in Europe. Given the complexities of the NLG authoring tool, product handoff reports are necessary to assist linguists, especially when authors are offline. The reports used for localization handoffs have gotten increasingly complex.

Opportunity: Streamlining the Process for Localization Handoff 

A project manager tracked the metrics of queries raised by linguists across each month and project. These metrics, along with specific linguist feedback, suggested that we could reduce the number of unnecessary linguist queries if we reformatted the localization handoff. 

Hypothesis

A streamlined handoff process could save author, project manager, and linguist hours (which are lost while linguists waited for answers on a project due to the time differences). 

My Role:

I worked directly with the “users” on this project, organizing, collating, and designing everything we showed to the linguists. I collaborated with my manager and another colleague on the process to make sure everything was watertight. 

Tools Used

Excel, proprietary content authoring tool & content management system, Jira, UX principles

Process

Instead of focusing on a solutions-first approach, we worked to clarify the user needs of the product–i.e. the handoff, before focusing on solutions:

1. Collect User Stories: 

a. We collected user stories from all linguists and the localization project manager in the following format: 

As a _____, when doing [task], I need _______ so that I can _____.

As a project manager, when receiving a localization handoff, I need to know what clients are in scope so I can keep track of translation hours per client for billing purposes. 

As a project manager, when running a completion audit, I need to know which units are in scope so we can filter out data properly and notify linguists which units are still pending. 

b. We did some user research by watching linguists work in the proprietary authoring tool and taking us through the process of receiving a handoff and localizing the product. 

2. Reflecting User Needs Back: 

1. We received a few dozen user stories and ensured they were formatted as requested (problems, not solutions). 

2. We then organized them based on topic or area and reflected them back to the linguists to ensure we were on the same page, asking followup and clarifying questions as we went. 

3. We also asked linguists whether each need was “required”, “nice to have”, or “not required. 

3. Design and Redesigned Solutions:

Based on user stories and the existing localization handoff, I and two others redesigned the handoff in a way that would better meet users needs. This involved a discussion of the linguists’ needs, where the knowledge should live (whether it should be provided by authors in the handoff, sit on the authors’ project page, or be self-serve) and whether additional training was needed.

Challenges

• Linguists’ process and terminology were unfamiliar to us (the authors), so we had to follow up and ask many clarifying questions.

• Linguists were eager to move faster through the process and in some cases wanted to provide solutions rather than their user needs, so we had to push back against this. 

Learnings

• Linguists were confused about the blueprint, which authors use to begin authoring but which was not an accurate representation of the completed project. 

Next Steps

• Worked with developers to make the paragraph reports available in linguist view

• Conducted a linguist training on how to use paragraph reports

• Removed blueprints from project pages as they caused too much confusion 

• Assigned project manager to pull this report and include it in linguist assignments 

• Various other small tweaks that improved the process

• Redesigned and simplified the project page template

Outcomes

• Resulted in a reduction of 8 hours per handoff. 

• Resulted in a reduction in unnecessary questions per project.

• Reduced frustrations & improvement in rapport between authors and linguists and in morale of both authors and linguists

Conclusion

• This process was slow and methodical, but it worked quite well. 

• Our hypothesis was correct, because this saved 3-6 hours per linguist per project. 

Reflections

• I would have liked to do more user research, spending more time watching linguists use their interface and understanding their processes, before designing a solution. 

Appendix

a. What is dynamic content? 

Dynamic content uses templates to create flexible content, in which the content changes based on the available data for a specific topic, such as a hotel's amenities or a city's landmarks. Another way to understand dynamic content is to think  of the childhood game of mad-libs, except the data that you fill in the blanks should actually make sense. Data management is the other side of the complex creation of dynamic content.

b. Who are the “users” in this scenario? 

The localization team (specifically, the linguists and localization project manager). We collected stories from those with the greatest experience in the NLG tool.

b. How does dynamic content tie into localization?

Localizing dynamic content causes increased challenges, compared to localizing static content.