Adopt automated translation management for future-oriented LSPs

The language industry has changed tremendously over the past couple decades. In the early 2000s, a usual project would involve translating individual documents into one–two–three languages, tops. Today, we’re talking entire sets of documents, and the expected number of target languages has reached two digits — and counting. The world itself is getting faster, and no customer wants to wait for translated content for days — they expect to go live in all languages simultaneously.

This obviously increases the load on a translation project manager (TPM): It’s no longer about sending a few emails between yourself, the customer, and the translator. It’s dozens of communications a day, hundreds a week, thousands a month. And it’s not only communications — it’s statuses, progress, processes you have to keep in your head.

Project managers are at the center of everything that happens in the language industry. They are the connecting link between customers, translators, contractors, partners, and internal stakeholders. They are the first to be contacted when things go wrong. They are the ones to take the heat for missed deadlines and failed projects. No wonder project management is said to be the highest-churning profession in the industry.

At Smartcat, we believe that automation is the key to relieving TPMs of the stress they are put under. Just like airplane pilots progressed from pulling a bunch of levers to controlling heavily-technologized cockpits, project managers should embrace technology to achieve peace of mind and focus on the really important tasks.

The purpose of this whitepaper is to dissect the work of a translation project manager, and, piece by piece, think of how it could be automated. Not all sections provide answers, and those that do are not set in stone. Our main goal here is to start a discussion on a topic that the whole industry’s success relies upon.

Before we dive into detailed analysis of project management in the language industry let's briefly review the main steps in the process of managing translations (see diagram below):

Step 1: Handle customer request
Step 2: Prepare the project
Step 3: Find translators
Step 4: Execute the translation
Step 5: Assure quality
Step 6: Delivery
Step 7: Payments

Naturally, these tasks could be divided among a few people — it all depends on the structure of the LSP. The opposite is also true: one PM can — and usually does — manage several such workflows at the same time. This non-linear pattern increases the number of things that can go wrong.

Now, let's break this process down.

The goal of this step is to quickly pass from a state where the customer has a need to translate something to one where they have a concrete quotation and timeline commitment from the LSP.

Typical substeps are:

• Receive the request

• Calculate the wordcount

• Prepare a quote

• See if there are additional requirements such as

• Scanning a PDF

• Downloading website contents

• Removing sensitive data before translation

Without automation, both the customer and the vendor have no standard way of placing & accepting projects, using emails or instant messengers instead. This leads to human errors:

• You have to keep project-related requests together with all other mail and hence risk losing it.

• If an additional request comes in for an existing project, you have to remember which one that is and track it down.

• You have to re-attach all the required information and files in your project management system — if you have one — or, again, keep track of it in your mailbox.

• The customer has to wait for the vendor’s response to get a quote for even a simple translation request. And, if they send it out to several vendors, the first one to reply may get the job.

In simpler terms, neither the vendor nor the customer feel fully in control and aware of all the requests and projects.

Thus, the purpose of project management automation is to build a standard workflow that allows LSPs to:

1. Process customer requests faster
The customer can send a document — or upload it via a portal — and get an approximate quotation for it right away. Project managers no longer have to deal with emails or download/upload files. This is especially useful if the TPM works with many simple orders, where manual operations become a major time killer.

2. Enable continuous localization
Alternatively, they can use an API connector for major content management systems or repositories such as Wordpress, Drupal, GitHub, etc. Automated content exchange eliminates overhead related to updates, which can happen often with today’s agile-minded customers. The content required for translation is always there at the manager’s fingertips, ready to be worked on.

3. Avoid human errors
The files automatically appear in the same system where they will be handled & translated. This avoids costly manual actions and minimizes human error at the same time. All participants can be sure that contributors are working on the same exact documents & versions submitted by the customer.

As a result, the TPM has the “conveyor” constantly ready for new orders, and the customer can be sure it will be reliably handled.

After the order has been confirmed, the TPM’s goal is to make the project ready for translation as soon as possible.

Typical substeps are:

• Pre-process and analyze the files,

• Analyze accompanying information,

• Calculate internal statistics,

• Plan internal deadlines.

Without automation, this can become one of the hardest parts to manage:

• If a file is too large for one translator, you have to split it into several chunks to assign it to different people (this intersects with Step 3).

• You have to remember if you had similar projects in the past, which might or might not contain parts of the same text and use the same terminology.

• You have to check for repetitions, full or partial, within this new project, so that your translators don’t do the same work twice.

• You have to manually exclude text that does not need translation, e.g. source code snippets.

Thus, the purpose of automation is to let TPMs:

1. Follow a simpler process
The solution must allow the TPM to assign different parts of a document to different people without splitting it (see also Step 3). This reduces human error and speeds up the process: For example, if a translator goes unresponsive, you can just re-assign their part to another translator without having to spend time re-dividing the document.

2. Reuse past translations & terminology
Client-specific translation memories (TM) and glossaries help keep the customer’s voice coherent throughout projects. Customers care about style as much as they do about costs. Translation memories allow the vendor to provide discounts to their clients when repeated content is translated, while glossaries make sure that the same terminology is used throughout.

3. Calculate internal statistics
Internal statistics of TM matches, repetitions, and so on might not be the same as the one provided to the client. For example, the vendor might want to charge the customer for the whole word count, while paying less to translators for TM matches.

4. Provide context-specific comments
Screenshots and explanations for certain portions of the source text greatly increase the adequacy of translation, because translators always know exactly where the text appears and what it means — as not all strings are self-explanatory.

5. Better manage deadlines
The system should automatically suggest deadlines for each workflow stage based on historical productivity, or the number of translators needed if the deadline is predefined. Different language pairs and specializations have different average translation productivities. Having a system that provides data-driven suggestions will help TPMs make better decisions.

As a result, instead of a haphazard set of disconnected tools and resources you have a single dashboard where you can make the best decisions before proceeding to the next step of finding vendors.

In this step, the goal is to find translators who will do the job well and in time. As a TPM, you want to complete this step as quickly as possible, because today’s projects have very tight deadlines. And you want the translators to have relevant expertise and ideally positive feedback from previous customers.

Typical substeps are:

• Check in-house specialists, if you have any,

• Check known freelancers, based on budget & other expectations,

• Find new freelancers, based on these expectations,

• Optionally: Book specialists for a future job announced by the customer.

Without automation, the typical workflow consists in polling people for availability directly via instant messengers or email lists. This takes too much time and makes the process unmanageable:

• Some of your people might go on vacation or get sick.

• They might have too many tasks from other customers — or even your coworkers.

• You have to keep a spreadsheet or other means of tracking who’s available.

If none of them are available, you have to go to job boards or translation forums to hire new people. Here you have even more challenges, spending time:

• Writing an introductory message,

• Doing background checks of those who respond,

• Onboarding the chosen ones with your workflow.

Thus, the purpose of automation here is to:

1. Move the workflow to one place
When the whole workflow moves to a single platform, the TPM doesn’t have to waste time switching between apps/websites. Using a single UX to manage the vendor sourcing process not only saves effort, but also reinforces consistency and reliability: Whatever your preferred process is — in-house or freelancer-based — the sequence of actions is similar and familiar.

2. Make the selection process data-driven
The TPM has detailed — and verifiable — information about translators’ skills, experience, and feedback on their profiles, with the system automatically suggesting specialists based on the history of their translated texts and the source material in question.

Assigning the right people can be challenging, because many translators will say that they’re experienced in more subject matters than they actually are. Having a system with data-driven suggestions based on actual corpora will keep that problem at bay.

3. Enable availability and productivity tracking
The TPM can track each translator’s availability based on the current workload and historical productivity — both of which are easily calculated given that all work is done in one place. No two translators work at the same speed, and not all may be in the same time zone, especially in multilingual projects, so distributing the work among translators is a highly non-linear task. Having a machine do all the math is a great help.

As a result, the TPM can find & assign freelancers within minutes or even seconds and move on to other projects.

Finally, we come to the essence of the work, where the goal is to have the project completed in time and to a high standard (see also Step 5).

Typical substeps are:

• Monitor progress

• Handle questions from translators

• Avoid missed deadlines

Without automation, the whole process is a black box to the TPM:

• There’s no way for the TPM to see the project progress other than asking the translators,

• There is no way to communicate with translators but by sending emails or instant messages back and forth,

• There’s no way to be certain that the translators are actually doing what they say they are doing.

This keeps the project manager stressed throughout the project — because they will be responsible if something goes wrong.

Thus, the purpose of automation here is to help the TPM:

1. Track progress
The TPM needs to see real-time information about the progress of each and every document and subtask within the project. Keeping the finger on the pulse allows the manager to nip issues in the bud and ensure timely delivery. If a translator goes unresponsive or there is no progress for a long time, the TPM can re-assign the task to others.

2. Communicate within context
The TPM must be able to read and reply to translators’ specific questions within their context. Context-specific discussions save time on clarifying the context of every question and provide a single place where all comments can be handled. Besides, if it’s a multilingual project, the comment is visible for all translators so you don’t have to tackle the same questions again and again.

3. Stay alert
The system should send notifications when it detects possible issues. One example is the already mentioned lack of progress. Others include approaching deadlines, too slow — or too fast — a translation, lack of MT editing on a post-editor’s part, and so on.

As a result, the TPM will be able to see the whole process and respond to any alerting situations with a prompt and well-informed decision.

The goal of quality assurance is to make sure that mistakes are not just fixed, but prevented. In this regard, it is generally incorrect to refer to it as a “step”. It’s a continuous process that happens before, during and after any translation project. But for the sake of simplicity we’ll be describing it this way here.

Typical subtasks are:

• Reuse historical data before the process

• Control quality during the process

• Learn lessons after the process

• Handle customer complaints

Without automation, quality assurance is mostly reactive and intuitive. The TPM can assign an editor to review a completed translation, gather the feedback and, depending on it, either keep or refrain from working with the same translator in future. This approach is far from optimal:

• You don’t know if a translation is good before it’s done.

• You have to keep the data on “good” and “bad” translators somewhere.

• Some translators are “good” in some respects and “bad” in others — which you also have to remember.

As a result, many TPMs have to rely on their hunch alone when making quality-related choices.

Thus, the purpose of automation is to make quality assurance:

1. Data-driven
All historical data on quality is stored in a reliable and structured fashion. Having all quality-related data stored in a single place allows it to be reused by other project managers. Similarly, as a TPM you can reuse the data reported by other TPMs.

2. Proactive
The platform automatically suggests translators based on their experience and subject-matter expertise. As already mentioned in Step 3, assigning the best matching translator for a given job results in both higher productivity and better quality.

3. Collaborative
Editors can review translators’ work not after but during the process. If the TPM can assign an editor when the translation is just 10% done, they can promptly identify any major problems with style, terminology, or lack of skills on the translator’s part and fix them before they become too costly to fix.

As a result, quality assurance becomes omnipresent in the translation process, affecting every step and outcome.

The goal of this step is to hand over the translated documents to the customer.

Typical substeps are:

• Prepare completed files

• Deliver the result

Without automation, there can be a lot of manual work involved:

• “Gluing” together parts translated by different translators,

• Making sure formatting doesn’t go astray,

• For large files, uploading to file sharing networks.

This takes too much time for a step that involves nothing but technical tasks.

Thus, the purpose of automation is to remove almost all manual involvement at this step:

1. No manual “gluing” or formatting
These happen automatically, with each text element replaced with the right translation. With the machine doing most of the work, the TPM can now just do one quick check after downloading the file to make sure that everything is in place.

2. Single user experience
All the files remain in the same place, downloadable by the TPM or, in case of a portal solution, the customer. This saves time on sending emails or notifying the customer of completion and reduces human errors like sending the wrong files and the like.

3. External integration
Alternatively, files are automatically imported to the customer’s CMS or repository, making any manual actions unnecessary. This completes the localization process and relieves the TPM of the need to deal with third-party systems they have no experience using.

As a result, the TPM spends virtually no time delivering the translation.

Although this step, the goal of which is to have all the work paid by the customer and to the freelancers, is not technically a part of project management, it is usually the TPM who does it.

The substeps here are:

• Calculate volume-based payables

• Optionally: Calculate hourly payables

• Get paid by the customer

• Pay freelancers

Without automation, this can quickly turn into a nightmare:

• You have to find out each freelancer’s preferred payment method,

• You have to manually calculate all the payables, taking into account any TM discounts and the like,

• You have to manually send each payment to every freelancer,

• You have to fill out paperwork and comply with tax legislations in every country you pay to.

As a result, the TPM spends more time doing this than on their actual work — not to mention the indirect costs it involves.

Thus, the purpose of automation is to ensure:

1. Automatic payout calculation
All payables are calculated automatically based on work done, rates, TM matches, etc. With the integration between work & payments, there’s no longer a need to make any external calculations — all the data is available for the system to do the math.

2. No individual payments
The payment is made once and in bulk, with all further amounts distributed between freelancers automatically. This is the biggest time saver as there is no need for the TPM to handle each company-to-freelancer payment individually, which means hours of work saved every week.

3. Tax & legal compliance
All tax & paperwork issues are handled by the platform providing the payment technology, not the TPM. Dealing with the intricacies of handling payouts in every geographical location is a drag, so having a third party do this saves you not only from stress, but potential fines as well.

As a result, the TPM gets rid of one of the most boring parts of their job, while freelancers get payouts in their currencies and using their preferred methods.

Many translation project managers and LSPs are afraid that automation will take over their jobs: Why does a company need a translation agency if they can run the whole process in a few click themselves?

Nope. Companies will always need a helping hand when they go global. And the more crucial globalization becomes, the more they will need it. Granted, there will always be companies with their own localization teams, but this is nothing new.

What will likely change is what companies expect from LSPs and TPMs. In this regard, the focus will shift from running and managing separate jobs to organizing whole programs to enter new markets.

In either case, this is not something that is going to change overnight, or even in a couple of years. What we can say for sure, though, is that LSPs who don’t embrace automation will be left behind, stuck to solutions that no longer work.

Which side will you be on?