Tarjimly: Crowdsourcing Translators to Help Refugees Communicate
How can you leverage crowdsourcing to help refugees in a scalable and financially sustainable manner?
Tarjimly is an early-stage non-profit organization that is helping to connect refugees and aid workers in real-time with crowdsourced translators. Based in the Bay Area, its lean team, is part of Y Combinator’s latest batch of promising startups.
Value Creation
Cofounders Atif Javed and Aziz Alghunaim, close friends since they met as prefreshmen at MIT, wanted to do something to create a significant impact on the worldwide refugee crises. They realized one of the biggest problems for refugees was communication. Not only did refugees need help communicating with aid workers and those around them, but also in translating vital documents like medical records. Aid agencies and NGOs sometimes hire translators to help with this, but this can prove very expensive, and any given translator can only serve one person at a time. Moreover, refugees need translation services all the time, even when aid workers or translators aren’t around, and existing digital solutions like Google Translate are simply not sophisticated enough for such translation purposes.
Tarjimly works to solve all of these key problems. The service connects those in need of translation services to crowdsourced volunteer translators, who now have an easy, tangible way to help refugees around the world. Tarjimly already has a community of 2800+ volunteer translators, 800+ refugees, and 600+ aid workers. The service is currently available through Facebook Messenger, but the team is working on building its own platform since customers have expressed a desire to use a separate channel for such communication. Whenever a refugee or an aid worker is in need for interpretation services, they can request a translator for the desired language through the Tarjimly chatbot. The service then uses proprietary algorithms to figure out which volunteers are most likely to be available to help, based on parameters such as their geographic location (and thus time of day), online activity (for example, whether they are currently active on Messenger), etc, and pings several of them until someone responds to the request. Their algorithms have become so good at identifying the best people for the job in real-time that the average time between a request and a response is just 60 seconds! For reference, it sometimes takes a ridesharing app longer to connect to a driver in Boston. Parties on both ends can then interact anonymously with each other via text or voice and can transmit photos, videos, and documents to each other over the platform. Users can also rate their experience after the interaction, which can help weed out any fake volunteers.
Since Tarjimly is still very young, it has relied on grants and philanthropic donations for financial support so far. But going forward, the team wants to set up a business model that will allow them to be self-sufficient. The founders are currently considering three alternatives for revenue generation:
- Charge NGOs and Aid Agencies to use service – Perhaps the most intuitive revenue generation method is to charge NGOs and aid agencies to use Tarjimly’s services. There are, however, quite a few barriers to implementing this. Firstly, aid organizations could be difficult to penetrate since many of the organizations prefer to use vendors they already know and trust. Additionally, many such organizations have had bad experiences with clunky, unsophisticated tech products, and are not inclined to be the first adopters of new technology. Finally, since the service available to refugees free of cost, aid agencies can get around paying for Tarjimly by having refugees initiate the translation requests.
- Upsell data to Governments and Organizations – Another way to capture value could be to make the translation services free but to sell the data generated from these interactions to aid agencies and/or governments. Tarjimly could consolidate information regarding how many refugees are being tended to every day, what topics are top of their mind, what services they seek the most, etc, from usage patterns. Such information could be extremely valuable to these organizations in understanding their impact, gaps in their reach, and efficiency of fieldworker deployment. While there may be some privacy issues to be navigated in this model, most consumers of the web have become accustomed to this method of data collection.
- Charge translators – Tarjimly wonders whether its translators would pay to volunteer. This does sound like an unrealistic proposition for most platform businesses trying to build a network. However, since Tarjimly’s translators are people who are actively volunteering their time to help a cause, and since they will know that their money is helping keep the platform up and running, Tarjimly’s founders think their translators may have a higher willingness to pay compared to customers of other platforms.
Based on the aforementioned considerations, my personal view is that Tarjimly should charge NGOs and aid agencies for using their services through a subscription model (instead of SaaS). Aid agencies currently pay an average of $80/hr for translators, costs that could be reduced by over 60% if they switch to Tarjimly. Also, charging a yearly subscription fee would give Tarjimly revenue certainty, making them less sensitive to the fluctuations in usage through the year and less prone to disintermediation. Lastly, since the service is so reliant on having a large network of active translators, it is best to avoid charging these volunteers so as not to alienate the more price sensitive members of the community.
Final Thoughts
Many resources become available to the world when language barriers are removed. And thus while Tarjimly has started with its core translation product, its founders are already thinking about additional services that could be offered through their platform, making their core product stickier while further helping those in need.
In terms of leveraging digital technology for the public good, Tarjimly is certainly a breath of fresh air. While there is a long road ahead of this young startup, the opportunities are also large and profound. Here’s to hoping Tarjimly phenomenal success!
References
- Personal correspondence with Co-founder Atif Javed on March 23, 2018.
- https://techcrunch.com/2018/02/13/bilingual-tarjimly-lets-you-help-a-refugee-or-aid-worker-right-now/
- https://www.forbes.com/sites/toriutley/2017/06/20/how-three-millennial-entrepreneurs-created-a-messenger-bot-to-help-refugees/#4a4cdd7c6be9
- https://www.bustle.com/p/tarjimly-connects-refugees-with-translators-the-service-couldnt-be-more-needed-8240902
- https://www.youtube.com/watch?v=NZp0a7_zTUg
- https://www.tarjim.ly/en
Such an inspiring story! I now many people who want to contribute to the refugees’ plight, but who have struggled to find something actionable. This is very impressive. A couple of concerns:
1) Apps like Google Translate are not currently good enough to serve this function (per your post), but they are improving dramatically, quickly (e.g., able to live translate japanese character on your phone screen) — how much longer will humans need to be in the loop for translation (especially for basic text etc)?
2) Regarding your monetization model, I think it is likely the best route, but I’d advise to walk it very carefully. Thoughtful communication will need to be conducted with the translators (who to this point have viewed there time as a gift, but may either a) want some cut of the profits, or b) disengage if the organization starts to charge end users fees). Given the fact that many other organizations out to make a profit are able to get crowdsourcing support without paying contributors it seems likely that this app will be able to do the same. However, navigating the change thoughtfully will be critical to not harm the contributor base.
Thank you for such an inspiring article! I wonder how learning effects will fit into this model. I assume there are some overlaps in documents required for translation (e.g. many refugees seek to get their medical documents translated). If we rely on crowd-sourcing model in which people who will translate documents are different every time, such learning effects may not be accumulated effectively. Is there any function to avoid such loss, e.g., matching certain type of documents with people who have translated similar documents before?
Taka, I had the same thought as I was reading the post. I would think that there is a high overlap in type of concerns and documents. Implementing things like tutorial videos or chatbots for the most frequent topics would help the platform scale. However, given that there seems to be an oversupply of volunteers I don’t think that this will be an issue in the short-term.
Thanks for a great blogpost! I really like the social impact of this project and I believe there will be lots of volunteers willing to contribute. My concern is that it might be difficult to scale since one volunteer can help only one refugee at one time, and repeated common questions/requests might be asked again and again. In addition, I am wondering how to control the quality. I found most topics consulted by refugees are legal and medical related, and these areas require some expertise (not to mention that the legal/medical system is very different in different countries). How can the company make sure volunteers provide correct information/translation for refugees?
Ting, I had the same thought. Given that it sounds like the documents that refugees are translating might be relevant to the much larger community, I wonder how they can leverage that single translation to share with the broader community. Or maybe this could even turn into a monetization strategy where governments (or other organizations) pay for documents to be translated such that they are relevant to the refugee population and then they can get the data on the usage of that document etc. Since they are building their own app, I wonder what forms and such can be brought directly into the app – maybe with a crowdsourced translator as someone to help explain the form and its use? There seems to be a lot of opportunity here to help the refugee population.
This is such a cool idea. One this that shocked me is that there are far more translators using the application than refugees themselves. I love that there are so many people who are excited to volunteer their time for this purpose. But also curious about how to increase the scale of impact that this can have by increasing the number of the refugees using the app. One question is what percentage of refugees have smart phones than enable them to use this service (perhaps most do?) Or perhaps this is where aid workers come into play for the more immediate translation needs, but what about the continued translation needs of refugees if they do not own their own smart phone.
I think it is great that it is possible to connect with a translator within 60 seconds, but I wonder if this is one of the most important metrics. Other key considerations should be the quality of the translation, the time required for each individual translation, etc.
It is wonderful to see technology being leveraged in a way that can really help a massive problem facing society today.
Great post! It is amazing to see a use of the crowd sourcing /community model for non-for-profit organizations! I would assume that Tarjimly will not struggle with motivating people to contribute, neither do they struggle much with the quality of content. Everyone contributing is naturally incentivized to produce high quality work given what’s at stake. I personally would love to be able to have a platform where I can help other so easily, i.e. from the comfort of my sofa. I hope that Tarjimly will inspire more businesses like this.
Thanks for an interesting read! I had a post about Duolingo’s translation services – in which Duolingo monetizes the translated content crowd-sourced from its users.
One of Duolingo’s challenge going forward is how to best meet the demand with quality translation in a timely manner. I think they could learn from Tarjimly’s proprietary algorithms, not the piece about geo-locating translators or enabling real-time chat, but how to better match demand of translation with the user who has the best suitable linguistic capabilities and participation rate.
I think Tarjimly can also learn from Duolingo’s monetization model – free for users (in Tarjimly’s case translators and refugees) to encourage participation; charge organization who will benefits or have cost-savings by having the translation done by Tarjimly.
I love what this nonprofit is doing and am so glad you shared it with us. Like you, I agree that the NGO option is really the only viable option of the ones you laid out (I think there are some potential privacy issues when it comes to using the data elsewhere – especially for refugees), but the issue you pointed out to circumvent the fees is a real one. If the platform can think of ways to make the use of the platform relevant to the NGOs (i.e. refugee management, a way to deliver and submit the filled out forms, etc.) that could be one potential avenue that would make the NGOs willing to pay for the services rather than have the refugees look for translation directly. Another option of course is to continue applying for government grants or solicit donations, as this is a nonprofit venture.
I agree with everyone so far: great post! I share JPW’s concerns regarding advancements in products like Google Translate that could render this platform irrelevant in several years. This company has a great mission as you highlighted, but I am also concerned about its ability to capture value without having to share some of that value with the volunteer translators. As someone previously noted, the translators currently outnumber the users, but I wonder once Tarjimly scales if that ratio would remain true with 65.3 million refugees in the world today. I think it is possible in the growing “gig” economy, but I wonder how much value would need to be shared to attract sufficient numbers?