Crowd Translation: A Beginner's Guide

As more companies recognize localization as a fundamental ingredient to reaching customers around the globe, the need for translation has risen dramatically.

And the localization industry has grown accordingly. Often, traditional translation processes aren’t up to the task of producing global-ready content at the increasing speeds and volumes necessary for companies to stay competitive. This has led the industry to develop new and more flexible approaches to translation than working with single-language vendors (SLVs).

You may have heard the buzzwords: crowds, communities, curated communities. They’re often mistaken for each other, but what do they really mean? How do they work? Are they better for your business than going down the traditional route of sourcing translations from in-country, specially educated linguists through agencies?

For this post, RWS Moravia Group Project Manager Marina Pantcheva is helping us break it all down for you.

The first thing to note: ‘crowd’ is an umbrella term for a few different methods of translation management. We’ll compare those later. Let’s start by comparing the general concept of crowd-based translation with its traditional counterpart.

What’s the difference between crowd-based and traditional SLV-based translation?

From the table, you can get a quick sense of how the two models differ, but let’s unpack them.

At its core, crowd-based translation allows you to break up large translation projects into smaller chunks and distribute them directly to a large global community of freelance translators to work on concurrently. These resources may or may not be professional linguists, but the assumption is that they have mastery of the language(s) they translate from and the language(s) they translate into.

Under the classic single-language vendor model, on the other hand, you contract specialized services from professional translation agencies or vendors that partner with their own in-country resources who have proven to be qualified in specific language pairs. SLVs might also divide up projects to handle larger volumes, but they use their own dedicated systems for project management and resourcing.

And we can separate the differences even further:

Crowd translations are based on open-call work distribution

Working relationships under the crowd model depend on translators picking up the work. Individuals are free to claim translation tasks if and when they choose. SLVs, on the other hand, typically have contractual obligations to deliver, guaranteeing a certain throughput per unit of time. This means they are usually long-term partners that know your brand and localization program inside-out.

Crowds have hundreds of resources for each language

The number of resources you work with under the SLV model, true to its name, is often commensurate to the number of languages in your localization project: usually no more than a few resources per language. One agency may not be able to cover the entire scope of your project. Under the crowd model, a large bench of resources will be available, across languages, to ensure coverage of the work. It could amount to tens, hundreds, even thousands of translators per language.

Crowds provide flexible and fluctuating capacity

The capacity of crowds is calculated based on the number of translators in the crowd. But there can be big discrepancies between expected and actual capacity for two reasons: crowd translators can choose not to work, and crowds typically contain lots of “lurkers”: translators who are part of the crowd but do not actively contribute. Conversely, because of its contractual nature, a vendor’s or agency’s capacity is fixed and often guaranteed.

Crowds have better coverage of time and geography

Flexibility of coverage is greater under the crowd model. In general, crowds are active 24/7 because they work all days of the week, both within and outside of their native countries, allowing for better coverage over weekends, holidays and time zones. SLVs typically work with translators located in-country, in their own time zone, and national holidays can pose scheduling problems.

Crowds offer excellent transparency and data

Transparency is lower within the SLV model. Language service providers (LSPs) send projects to vendors and get them back with reduced insight into what happens in between. In the crowd model, project managers and clients can extract data on the way each individual resource works: their average turnaround time, quality scores, the types of projects they pick up and so on. This opens up an ocean of possibilities for the smart management of these communities.

And here’s where crowds can be further categorized into three distinct levels of management, from unmanaged to carefully curated.

What’s the difference between crowd, community and curated community?

Crowd (or ‘crowdsourced’) translation usually involves a large network of resources with certain skills. You’ll find these freelancers through online crowd translation platforms. These crowds are public groups—the entrance is free for anyone who believes they possess the skills necessary for the crowd tasks—though crowd members may still be asked to take a test to qualify for the work. Crowds are best for tasks that do not require special knowledge or training, such as translating user-generated content.

Translation communities (not to be confused with volunteer translator communities, which translate things like Wikipedia articles or TED Talks for free) are built by selecting, qualifying and engaging a subgroup of the crowd to form a network of resources assigned to a given project. These people may come from the crowd and become a dedicated group for you. There is a higher level of engagement and communication among the group—who, together, have a vested interest in the project’s success—plus more solid testing and qualification of resources. Communities are great for translation of more specialized content, in which brand terminology and style guides must be adhered to.

Curated (managed) communities are hierarchically structured communities. The structure comes from assigning different roles to the community members (translator, editor, reviewer, etc.) and different levels of expertise within each role (junior versus senior reviewer, standard- versus premium-quality translator, etc.). Mirroring the same level of specialization you’d get with an SLV, curated communities can deliver complex tasks as well as an SLV can.

So, what’s the best model?

Each has its pros and cons. There’s no “better” model per se, but different models are best suited to certain situations. For example, the flexibility of crowd models lends itself well to localization projects of large or rapidly changing scale: an increasingly common need these days as content volumes increase alongside companies’ awareness of how important it is to localize their products.

But stay tuned—we’re diving deeper in an upcoming blog post into the best approach to take for different scenarios. For now, you can always reach out to us for resourcing advice.

Thanks to Marina Pantcheva for lending her considerable expertise to this blog post!