Sharktank Manifesto

Introduction

“First come, first served” job assignment systems, also known in the industry as shark tanks, have been used by language service providers (LSPs) for over a decade. This appealingly simple form of automation is considered useful by LSPs, as it enables them to distribute translation tasks quickly, with minimal manual intervention, reducing project management overhead, increasing the speed of job delivery, and improving overall scalability.

At the same time, these systems have caused significant, sustained damage both to the translation ecosystem and to linguists, negatively affecting their work-life balance, mental health, relationships, and income.

We believe that these issues can be alleviated with minimal investment from industry leaders committed to translation quality, ethical efficiency and human well-being. In this manifesto, we will first list the ways in which shark tanks have harmed our industry and translators’ health and livelihoods, and then we will propose a more efficient and humane alternative: EquiTASS – Equitable Task Assignment System.

How shark tanks hurt the translation ecosystem and linguists

They drive the best talent away

Top linguists often refuse to use shark tanks out of respect for their time and expertise, choosing instead to work under better conditions elsewhere. Because of that, they have no incentive to spend time repeatedly refreshing shark tank job pages in the hope of securing random assignments. For LSPs, the absence of such experienced translators in their talent pools results in poorer localisation outcomes, additional costs due to an increased need for linguist training, as well as a perceived shortage of qualified professionals in certain specialisations and language pairs.

They reduce translation quality
Shark tanks promote a “click now, think later” mentality, encouraging linguists to quickly grab assignments before fully reviewing project details, such as job type, work volume, deadlines, and other information. This can produce poor matches between linguists and projects, leading to missed deadlines and translation quality issues, which may strain client-vendor relationships. Furthermore, LSPs have to devote more resources to managing schedules, reassigning tasks, and adding quality control steps, all of which incur extra costs.

They diminish income
The shark tank system fosters a sense of artificial scarcity and desperation by encouraging often-oversized pools of linguists to race to grab jobs first. This dynamic primes translators to accept progressively lower rates out of fear that work opportunities may otherwise disappear. Some people do not find any substantial projects for months on end in the tanks despite repeatedly checking. This adds up to a considerable amount of unpaid time and labour, especially when freelancing for multiple vendors, which is standard in our field. The resulting drop in earnings often drives people to seek additional sources of income or to leave the profession, which means higher linguist turnover and therefore quality, capacity, and talent management challenges for LSPs.

They undermine work-life balance

In order to secure work through shark tank systems, translators have to be constantly tethered to their computers, at the expense of other aspects of their lives, such as outings with friends, hobbies, family time, exercise, self-development and holidays. This constant state of availability can lead to burnout, pressuring linguists into taking extended periods of time off work or quitting the field entirely, thereby reducing LSPs’ talent capacity.

They hurt mental health

1. By design, shark tanks induce a state of anxious hyper-vigilance, as missing a notification even by a few seconds or failing to accept a task quickly enough can mean losing a week’s worth of earnings.

2. Some systems do not send notifications for new jobs, requiring linguists to repeatedly open, check, and refresh task boards – again and again – leading to mental fatigue.

3. Because jobs may appear in the middle of the night, depending on the linguist’s time zone, many professionals feel compelled to wake up multiple times at night, or remain awake, disrupting their sleep schedule.

4. Shark tanks can induce addiction, as constantly clicking and only occasionally getting a reward (a job) reinforces behavioral patterns similar to those found in chronic gamblers.

They create a hostile atmosphere
Due to this toxic job distribution approach, often likened to the “Hunger Games” by linguists, freelancers are often turned into fierce competitors rather than fellow coworkers striving toward the shared goal of producing the best possible translations. As a result, their quality control work and other collaborative efforts can become strained, leading to friction and reduced willingness to support one another. This, in turn, creates quality and pool management challenges for the vendor.

They erode loyalty

Although some shark tank systems take the linguist’s skill level and experience into account, most do not. This means someone who has worked for an LSP for ten years with a flawless quality score is treated the same as a newly onboarded linguist or one with a poor track record. This reduces the incentive for the linguist to go the extra mile to ensure the highest possible quality or to build long-term relationships with clients.

They are susceptible to abuse

Shark tanks are vulnerable to cheating through the use of auto-clicker bots, browser extensions and custom scripts, leaving honest translators who rely on human reaction time with scraps. For LSPs, this creates a dangerous imbalance in job allocation, makes the other translators frustrated due to a lack of work, and lowers translation quality, as cheaters may sometimes outsource the jobs they get to other unvetted translators without informing the vendor.

What we want: an alternative solution

Even though we would prefer all translation jobs to be allocated manually by experienced project managers and indeed believe that a large portion of them should be assigned that way, we also recognize that a certain degree of automation is inevitable due to the sheer scale of localization operations today. Therefore, we would like to propose an alternative approach, as described below.

EquiTASS

To mitigate the negative effects listed above, the task allocation system needs to be more transparent, more supervised, and less dependent on immediate task pickup. Therefore, instead of the fully automatic “first taker gets the job” model, we propose a “human-in-the-loop” pooling method:

1. A new job gets posted in the system, visible to all eligible linguists (see next section);

2. The linguists express their interest by a certain deadline, which will depend on the job’s urgency (i.e. within X hours for standard tasks, within Y hours for rush tasks, and within Z hours for tasks that will become ready to work on in the future);

3. Once the deadline is reached, the system selects the five most suitable applicants;

4. A list of their names is then sent to the corresponding project manager for a manual decision on who to award the task.

To help the project manager make the decision, the names should be displayed together with the linguists’ suitability scores, in their descending order, along with a brief automated summary for the logic behind the score (e.g. “Tier 1 quality track record, 100% on-time delivery, Subject-matter match”).

• Linguists should be able to withdraw their interest at any time before the job acceptance deadline, without getting penalized.

• Should no one express interest in the task before the deadline, either the deadline can be extended or the linguist eligibility search scope can be expanded to include those with e.g. a different genre expertise or with their skill level one tier below.

• For a new season or installment of a title, the “legacy” linguist who worked on the previous one should be given a 100% suitability score by the system (i.e. a perfect match), as long as their translation was of good quality.

For transparency, once a job has been awarded in the system, the unsuccessful candidates should receive their auto-generated rejection notification, along with the total number of applicants and the rejection reason (e.g. “Position filled by a linguist with a higher suitability score”). Some general helpful guidance on improving one’s future selection prospects could also be provided.

This pooling approach would not only give linguists ample time to review the project details and determine if they are a good fit for the task, but it would also reduce the “hyper-vigilance” and gambling-like dynamics.

One source of inspiration for an EquiTASS platform could be the European Commission’s eXtra portal [1], which follows a somewhat similar approach.

Matching linguists to jobs
We envisage a system that takes into account each translator’s skills and other details when deciding on who will see the job offer once it appears in an EquiTASS platform. The following data should be considered:

1. Language pair(s) (including direct source-target pairs for pivot-language projects)

2. Skill tier/ranking in the specific job type

3. Specializations (e.g. subtitles for the deaf and hard of hearing [SDH], audio description [AD], dub adaptation, shortform videos)

4. Subject-matter expertise (medical, sports, children’s shows, etc.)

5. Sensitivity expertise (LGBTQIA+, disability, race, gender, etc.)

6. Availability calendar

7. Linguist-defined quiet hours (to protect sleep schedules while including night owls and those living abroad, e.g. an EN>JA translator living in the UK)

8. Linguist-defined minimum accepted rate for the job type (per word, per program minute, per subtitle, etc.)

Such tailored matching would improve the final quality of translations, boost work satisfaction and mental health for linguists, as well as extend their loyalty.

Suitability scores

Translators’ suitability scores should be calculated based on their skills, preferences, key performance indicators as well as some other parameters mentioned above. Different LSPs have their own scoring models, which take into account translation quality, on-time delivery, responsiveness and other metrics, and it is up to them to make those models as objective, fair and transparent as they can be, while ensuring the possibility of vertical mobility between skill tiers for linguists. Additionally, to avoid the “rich get richer” ranking trap – the dangerous feedback loop in which top-scoring specialists end up hoarding all the work, while newly onboarded and mid-tier ones get close to nothing and eventually leave out of frustration – both an allocation weight and a capacity cap must be established:

• A score penalty for applicants with more than X ongoing tasks currently assigned;

• A limit of Y ongoing tasks currently assigned

  • Shortform, longform and season-level tasks should have different limits/weights.

These limitations should contribute to a fairer, more equal distribution of work within the pool.

Work-life balance

An EquiTASS platform should have a user option to enable email notifications for new jobs as well as for rejections, so that linguists can be away from home and engage in life activities without losing access (via their phone) to task pickup or assignment verdicts.

Bot protection
With immediate task claiming removed, automated abuse becomes much less of an issue. Still, the new system should ideally come with anti-script protection where applicable. It can be based on behavioral analytics (time-to-click + click / reload frequency), but without intrusive keylogging. Easy-to-complete yet robust captchas can be surfaced for accounts with suspicious activity, with an explanation of why they were flagged as such, in combination with other smart protection mechanisms, like HTML honeypotting, proof-of-work, etc.

EU law

Regardless of their concrete implementations, EquiTASS platforms must abide by the EU Directive on Improving Working Conditions in Platform Work [2]. The deadline for the member states to implement this directive in their legal systems is 2 December 2026. By that date, all shark tanks used by freelancers based in the European Union will have to meet the strict requirements put forward by the respective countries, including provisions on platform transparency, algorithmic management, hidden ranking systems and appeal rights. Depending on the exact details, a failure to do so can lead to labour inspections, hefty fines, or a reclassification of the company’s freelancers into employees, with all the legal and financial consequences it entails.

Conclusion

We believe that the proposed system, implementable in practice, as well as other systems that improve upon it, would lead to a significant enhancement of the whole localisation landscape – not just for linguists, but also for LSPs, end clients, and audiences – and create a more sustainable future for everyone.

We hope this manifesto will inspire decision-makers in our field to discuss the current situation with shark tanks and to enact changes that will benefit all stakeholders.

References:

[1] Guide to the European Commission’s eXtra portal: https://webgate.ec.europa.eu/dgt/freelance/help/guide-en.pdf

[2] EU Directive on Improving Working Conditions in Platform Work: https://eur-lex.europa.eu/eli/dir/2024/2831/oj/eng

Accueil

Audiovisual Translators Europe

Contact

About

contact@avteurope.eu

© 2024. All rights reserved.