The Word Count

If you are in the market for localization services, this article is for you. I write about a fictitious business—Company X—and its managers and employees who have embarked on a buyer journey into the distant and unknown lands of localization.

I hope the following narrative helps you think through a scenario that you yourself might be facing if you need website and/or product localization services.  

An Industry of Words

Localization is a pricey investment. When done correctly, it adds value to your organization, increases global brand recognition, and boosts profits. It is an essential piece of the puzzle for companies looking to expand outside of their home market.

The cost of translation services is mainly determined by the word. For better or worse, this has been the principal mechanism for calculating costs, estimating turnaround times, and framing expectations for decades.

It's not just human translation that follows this model. Machine translation (MT) providers charge by the processed word. Translation management systems (TMS) charge by the number of hosted words. Some plugins charge by monitoring the number of words that traverse their API. The per-word paradigm is here to stay and it will be one of the first things that an LSP will ask you about.

This article is an introduction to the topic. It is predicated on Latin-based source words. Word counts can be examined from many different angles. Source content could, for instance, be 'source characters' just as well, i.e. based on double-byte Chinese or Japanese.

I have made assumptions to simplify the process and help offer concrete steps to move forward. The takeaway is that you will need to know your word count before contacting an LSP for a quote.

Hypothetical Use Case: Wordpress and Product Oh Yeah

Let's say that a fictitious business—Company X—decides it's time to localize its flagship product (Product Oh Yeah) and its website running on Wordpress. Oh Yeah has taken off in Company X's home market. The company wants to launch it overseas to spur growth. Company X is new to localization and it's VP of Content Strategy—Diane—has taken the following steps to get started.

1. Earmarked a $10,000 budget to cover the cost.
2. Appointed a Content Manager with no localization experience—Jill—as the localization manager (LM).
3. Prioritized the first languages to localize into: Spanish (US), German and Korean.
4. Set a deadline for one month from now.
5. Tasked Jill with finding an LSP who can produce quality translations at the right price.

If all three target languages are too expensive, then Jill is supposed to arrange for translation into Spanish (US) only to start. Diane's team will monitor the ROI and decide what to do later.

The LM's First Steps

With that, Jill Googles "language service providers (LSPs)" and see hundreds of hits. She asks around, does some research, and narrows the list to three LSPs. She starts emailing them.

Hello,

I work for Company X. We are looking for a translation agency to translate our website and a web application. We'd like to understand how much you would charge if we wanted to translate into:

- German
- Korean
- Spanish (US)

How much would this cost? We are hoping to find an agency who will grow with us and become a long-term partner.

Regards,
Jill

The recipients of that email on the LSP side will reply within minutes, sharing information about their company, touting their experience, and asking for a meeting. Importantly, they will not be able to provide a cost estimate yet because they need real data to do so.

Here comes a response to Jill's email from one of three LSPs she reached out to.

Dear Jill,

Thank you so much for contacting LSP #1 for translation services! We would love to help you achieve your translation goals.

To help us get started, could you let us know how many words need to be translated into those languages? And your timeline?

Please advise a time when we could schedule a Zoom call to go over your needs in greater detail.

Looking forward to talking soon.

Best regards,
LSP Enterprise Sales Director

The excitement Jill felt when she saw the reply appear in her inbox is tempered by the fact that she is tasked with work. She had wanted to show cost estimates to Diane soon, but Jill does not have any idea how many total words need to be translated.

The Temptation of Budgetary Exercises

Empty handed, Jill mulls over her options. She could hypothesize a number as a budgetary exercise, eg 10,000 words. Or she could plug her company's URL into any of a plethora of online content scraping tools to get a rough number.  

Both approaches warrant caution. It's understandable to provide a rough number to advance the discussion, but these approaches could be widely off the mark.

Scraping involves the use of a script written in a given programming language—Python, Javascript, what have you—to traverse a website, download public-facing pages, and calculate the number of words that it comes across.

It is a tricky technology because it can pull lots of unnecessary words ('noise') into its calculation and inflate the word count. Or there could be words on pages behind a firewall that should be in-scope but are inaccessible to the script.

In my experience, scraping results are more often than not imprecise. (If circumstances dictate that there is no other option, my recommendation would be to team up with a localization engineer to evaluate the results of the scrape.)

Let's say that Jill opts against doing a scrape, but copies/pastes a bunch of text from Company X's webpages into a Word doc. That would be more precise, albeit a carpal-tunnel-inducing exercise that is not scalable.

No bueno.

Another layer of complexity is the psychology of money.

Let's say that Jill goes ahead with one of the above approaches. No matter how carefully Jill caveats the budgetary cost estimate, the results will become firmly etched into her manager's mind, like scratching your name into semi-dry concrete that is hardening apace.

Diane will start drawing conclusions: "The all-in price for 10,000 words into three target languages is roughly $5,000. And if the estimate is inaccurate, it's probably only off by 10-20%. We just adjust up or down. We can afford this."

This is called anchoring bias and is very difficult to walk back.

An anchoring bias happens when you rely too much on the first piece of information you learn about something when making decisions or predictions. This bias comes into play with your finances especially when making investments or purchases. MoneyGeek.com

In my view, Jill has no choice but to figure out the word count. The LSP who wins Company X's business will need real data to do its job. Jill therefore has to take steps with her team to elicit a response such as below, from Beth, her colleague in the engineering department:

Jill,
We ran everything and the word counts you need are:

- Wordpress site: 30,856 words
- Oh Yeah: 17,888 words

Best,
Beth

Learning the Process

Even if Jill could share a precise word with the prospective LSPs, she would need to be patient. The LSPs will analyze the data, tell her the number of repetitions (covered in Part 3), and the total word count. But the LSPs will want to talk to her before sharing pricing information.  

Why? They will be concerned about commodification. The LSPs will—rightly—assume that Jill would add their prices to an Excel Sheet and compare them, paying closest attention to the lowest one.

The problem is that the Excel cost-comparison approach has a lot of gaps. For instance, the pricing does not inform a customer about the LSP's under-the-hood processes and services.

Good LSPs will help their customers navigate tools, services, costs, and quality concerns over variegated workflow steps. These activities become apparent downstream. Or in other words, certain essential information will not detectable from a number entered into a spreadsheet.  

So, Jill not only has to spend time obtaining word counts (internal challenge), but she also needs set aside time to give the LSPs a fair chance to explain their pricing structure and processes (external challenge).

Gather Information

I would suggest that Jill do a brainstorming exercise. She should write out her questions and share them with Diane and her colleagues to get answers.

Jill was disheartened about being unable to share pricing info with her manager, Diane, right away. Well, the next best thing to showing progress is asking salient questions. Doing so demonstrates planning and forethought.

Jill should set aside 30 minutes to develop questions designed to get at the heart of the matter: how many words need to be translated for Oh Yeah and the Wordpress website.

I've listed sample questions by category below get started.

General Internal Questions

1. Is the Wordpress site being managed internally or through a dev agency?
2. Who is the product manager for Oh Yeah?
3. Can I schedule a meeting with both teams?
4. Where is the source data for Oh Yeah and the Wordpress site hosted?
5. Which file formats can data be output to?
6. Are there elements of the website that should not be translated?
7. Are there validation fields set up on Oh Yeah and the Wordpress site? Can you send me an Excel or CSV list of them?
8. Is any of it confidential? I.e., does the LM need to execute NDAs with the prospective LSPs?

Takeaways

Onboarding an LSP for the first time is hard work. Even if the use case is straightforward. Jill is about to encounter questions from prospective LSPs that catch her off guard.

A key takeaway about working with an LSP is that it presents an internal / external dichotomy.  Jill has to assess the external LSPs pricing and processes. She also needs to manage relevant internal stakeholders on her side at Company X to help her get read.

Jill's next steps are to:

• Discover the total number of words that are in scope for translation by coordinating with internal stakeholders. This is a key step towards enabling LSPs to generate a cost estimate.
• Start to connect with the people on her side to help manage the data flow.
• Write out the tools, costs, and processes that will come into play to enable the LSP to provide its services.

Please see Part 2 of the Buyer Series, Workflow, where Jill is encouraged to consider the series of steps that need to be taken so that localization services can be provided smoothly. See you there!

Localization Roles

Company X: A business with a Wordpress website and a flagship product called Product Oh Yeah. Both need to be translated.

Jill: The localization manager (LM) working for Company X. Jill's title is 'Content Manager.' She was hired to create and manage content in Wordpress, and she has no previous experience with localization.

Diane: Content Strategy VP at Company X. Appointed to oversee the localization budget. Diane wants to translate the Wordpress website and Product Oh Yeah into French (France), German, and Spanish (US) if the costs fit the initial earmark. If not, then Spanish (US) only.

Beth: Company X Wordpress developer.