Collaboration and Scale

Making products that work in multiple languages isn’t just about words. It’s also about understanding users and their communities — and your collaborators.

Facebook places a careful emphasis on building collaborative relationships that help us serve billions of users. We’ll look at how we think about design relationships, how they result in products that scale globally.

Ben Barone-Nugent & Jamie Kimmel – Collaboration and Scale

Collaboration – teams working together
Scale – teams working together not just locally but internationally

When we build, we start with what we know – it’s our own bias! We think about things we’ve used, people we know; we draw on our experience and view of the world.

Problem 1: empathise and get beyond your biases, nationally
Problem 2: empathise and get beyond your biases, internationally

Billions Building Better … Facebook does put a lot of emphasis on collaboration. Not just to decide what to do, but also what not to do.

This talk will refer a lot to challenges in localisation; and the collaboration that goes on to make it happen.

How do you work out what works well in both NYC and Bangkok? How do you go about understanding the cultural nuances of language and features?

Ben and Jamie work in a group that helps people discover things to do wherever they are; and to connect with businesses so they can make money. The idea is “social at scale” – pairing social information with search results, to make them more compelling than raw search results. If you see a recommendation for a burger bar, it makes a difference if someone you know and trust has said it’s good.

Looking at the way collaboration manifests in team structures… Teams typically have three design roles: designer, writers, researchers. This does mean that everyone has responsibility for UX, even though nobody has the title specifically. Product design is a process of how the roles interact, the healthy tension between them that drives a good result.

To put it another way, collaboration is about having conversations with other people. That’s obviously much easier if you can co-locate. Being able to spin your chair and talk to someone makes it much more likely you’ll do so on a regular basis.

We blur the boundaries of what we do. Most people are good at more than just their supposed core role.

Collaboration in action… the tension between views is essential. It can be hard to be humble, but we will always be humbled if we become overconfident and silo ourselves.

When the recommendations feature was launched in Facebook, it was added in two ways

  1. an intentional choice to post a recommendation
  2. natural language processing would pick up relevant phrases, then suggest the user turn their post into a recommendation request

That means not only does the UI have to have well-localised text and labels, the underlying NLP engine needs different data and training. Simply translating the English into Thai doesn’t work! Slang alone will be a major problem.

Design is incomplete without language.

They researched the literal Thai equivalent to the word “recommendations” and found it had a dual meaning, one that matches closely to the idea of suggestions; but another that had connotations of seeking more-personal help. That made it weird to broadcast to their whole Facebook feed.

When Ben joined the team he needed to build credibility as a content strategist – the new “word nerd” needed to show how they were going to fit into the team and add value. A good place to start is the golden rule – do unto others… Give and take feedback, open up your process to others, get to know your coworkers as people. If you don’t know people at least a little on a human level, you won’t have the context to interpret feedback you get from them.

A useful phrase is “hey I could use your help on this”, to actively bring people in. Similarly ensuring that you acknowledge when you want to go with someone else’s option for something.

But it can be truly as simple as asking people about their lives (“How are the kids?”). Don’t be pushy or invasive, but open the way to talk about more than just work.

The cultural nuances of features… understand “extreme response bias”. If you give the same survey to people around the world, you’ll see very different response rates and content. In Brazil people are more likely to answer a 1-5 scale at the extreme ends (1 or 5); while many asian countries are more likely to use the middle range (2-4). So how would you provide a global rating system for movies? Does a five-star system even make sense? This is one reason some systems have gone to simple ‘thumbs up or down’ style systems.

What’s in a name? It’s often either the first or last thing in a project. The idea may start with the name, or it may be built and prior to launch a naming decision has to be made. But what value does the name bring?

Naming is really REALLY important. It says a lot about how you want to engage with your audience. A good name sets clear expectations for what your product is. It shows empathy, shows that you understand the customer.

When you are releasing something globally, you run the risk of losing people with a one-name-fits-all-markets approach. The “Local” feature was well-named for America, but didn’t work everywhere. In Thai the translation of “local” was a bit negative.

A naming process:

  • namestorm – throw lots of ideas around
  • wording – put some contenders into context, write some copy using it
  • locals – check with locals, talk to people who work in various regions to give feedback
  • testing – run experiments to see if the names work

The localised versions of the product did show good metrics; but ultimately it was worth doing to simply have a stronger product.

Last thoughts…

The intuitive is different from the actual – you think that simple translation will be enough to roll out a product or feature in a new region, but it just doesn’t work out that way. You need to understand cultural nuances of the target market and then keep testing and iterating as you learn more.

Ultimately in some markets you will need to solve quite different problems to get to the same result. American users were comfortable asking large groups of people for a recommendation; but Thai users preferred to ask much smaller groups, often with something like messenger rather than the open news feed.

The actual requires collaboration. Get to know your coworkers as people – people don’t have brick walls between their personal and professional life. Meet with people a lot. Co-locate when you can, to the product and not to the discipline. Be open and honest and human. When people know each other well as people, they can have much more honest conversations.

Names are really important. You need to think about them carefully, and not just through your own prism.