Lies, Damned Lies, and Statistics

I begin with a Mark Twain quote because the data for localization is confusing to outsiders and it really serves to insulate the industry from scrutiny. For LSPs and MLVs it is a great way say you pay one cost but we deliver more value than we charge you. For practitioners it is a way to create expertise, and for the businesses it often serves to justify investment into operational tasks that are required for international business. To be clear though I know there is a lot of value with working for LSPs and investing in localization so this is not an indictment of something that any party is doing wrong. It is a deep examination of the metrics in an industry. I know that is a lot coming from someone that loves the industry and languages. But let’s be honest and really hold a mirror up to ourselves.

There is a bit of a pyramid scheme going on here. I buy  a word for a dime, sell it for a quarter and throw in PMs to manage tasks. I’m oversimplifying but it is a wholesale market that sells branded and services to large clients in need. The industry has settled on a standard of 10% for PMs, but everything is negotiable (aka cost per word and PM costs are bargaining power to seal the deal). The economics of the industry are changing so LSPs are feeling pressure on their margins, but they are addressing it with less humans and more MT and post editing.The move is being supported by the clients to get content faster, cheaper, and for projects where full price and highest quality are not desired.  I don’t think there are too many agencies losing money. If there were we’d see less agencies every year, but we don’t. We are seeing consolidation to optimize and reduce costs to improve margins however.

So how does this relate to the data?  Let’s look closer.

Let’s say I’m doing 40 locales and 60M words a year. If you’re in the business, you know that the 60M words is translations into as little as 12-20 locales. So really we’re looking at as little as 1.5M source words across 12 locales that are then duplicated with new locale designations. This parent/child process is quite common in localization. On the vendor end you plan for 12 teams to translate 1.5 M words and have a few PMs managing the work. On the client end you can report out your team managed the translation of 60M words into 40 locales. Everyone wins but the metrics and economics of the process is muddy to outsiders.

But let’s further complicate our example.  Are the word counts actual word counts? Fuzzy word counts? Weighted word counts? or are they premium word counts? Each of these methods of counting words is common across localization vendors and they have full buy in to use them from their clients. Why you might ask? Because it makes costing and budgeting less volatile processes. And static budgets YOY from localization teams may be desirable to businesses that consider localization a cost with no benefit.  I’ve watched gross overpayments happen because managers were clueless to the workings of the processes and data from localization. Again this is not an indictment of the vendors or clients, it is just an observation.

Let me pause and define the types of content for you so that we are speaking the same language.

Actual words: This one can mean either the actual source words or the actual number of words created during the localization process. It mostly refers to source words.

Fuzzy words: Words that have some level of matching from translation memories, but they will usually still cost to translate or review.  The weighted words are derived from fuzzy words.

Weighted words: Weighted words are the number of words left when you multiply all the words times their percentage match. This weighted word process can be tricky because each vendor may use a slightly different method to calculate weighted words.

Premium words: The funny money or monopoly money of localization. Premium words are most frequently used by teams that want localization as a managed service with a monthly static cost. They usually build services like review, quality assessment, engineering, loc engineering, and PM costs into this flat fee cost.  They then track against premiums or actuals for their reporting purposes.  It is inefficient and often overly costly. The vendors love it because it locks in work and revenue, and the management loves it because it turns an industry with a volatile pricing scheme into a static cost. Unless of course management poorly plans growth, work changes, or required services change. Then the costs and pricing exceeds the static budgets and cost schemes.

But I digress. All of this is to say the metrics can either be very useful to help drive continuous improvements or it can be used to illuminate or obfuscate the localization cost and effort.

Localization Production Metrics

The standard metrics for localization are cost, time, and quality. Cost is either per word, per project, or per org/discipline. However they are measured lower is better. Costs are easily optimized if you change the methods and expectations, but it is difficult to drive too deeply into the vendors margins without necessitating a change in their methods.

Time. Time in localization is always tied to human effort. If less human effort is required then less time is required. You can cut the time required to translate by adding more translators, but this means more humans are necessary. Not only do you need translators and proofreaders, you also need  to add editors to keep consistency across all the translators working. So you may be able to cut the overall time of delivery, but you cannot cut the overall hours on the project. Translation takes time and though you can get more done quicker you are still using the same number of effort hours to get the work done.

Enterprise Business Metrics

Though the metrics we use for localization are important to optimizing localization production processes, they are completely useless to the company as a whole. The take away from localization is it is a cost of doing business. It takes too long and there are always concerns about the quality of the translations. So how do you change the narrative and get a seat at the table for the important conversations? I believe this is the most interesting part of localization and it is the space where localization becomes international business strategy.

So what is it that senior management cares about? That is the first question you need to answer before you proceed with creating the right metrics to discuss localization with the overall company. Usually it is money, mindshare, or eyeballs. It could be subscriptions, links, followers, fans, thumbs ups, purchases, pledges, etc. Whatever it is your job will be to figure it out and become an expert in that area. If you don’t know the right metrics or use them wrong you have zero chance of persuading the right people.

If you work at a public company start with your annual report. What are the metrics you measure? What do you notice about international market share and costs? What are the stated goals for the company? Do you have international expansion plans? What are you know for? Within all of this you will find the metrics you seek. If you speak the language and use the metrics of the company you will be able to influence the C-level.

What do you do if you’re at a private company. Look for internal documents about growth, expansion, vision, and key performance indicators.  These should help you design the right metrics for your discussions that will help influence your management.

Enterprise persuasion 101

Okay so now you have an idea of the metrics that matter to the C-level. How do you match localization with those metrics and make what you do relevant? This is where your data abilities or your creative ability to extrapolate from the data you have comes in handy. Anecdotes are great to help add color to your claims (e.g. “Thanks for providing documentation in my native language”) but in and of themselves they will not persuade the reticent or those not steeped in your business.

You need to show the millions you’ve spent in documentation is being used and is helping to support customers, grow the business, or increase customer satisfaction. How many unique visitors or downloads does the content get? How much more growth have you seen in a region since you added localized marketing documentation or marketing materials? What percentage of the company’s revenue is being driven by the particular market or region?

The key is to paint a picture with the data. Some will argue charts will do the trick. Hockey stick charts and massive YOY numbers help. However, you must control the narrative. Never let someone interpret the data for themselves. It will just ensure you reinforce what they already believe. That is the other beautiful and horrible thing about data we see what we want to. If I tell you “I spent $2M localizing content and we now have more sales in Mexico”, this does not show clear causality so lead them by the nose. Add more color. “We spent $2M had a traffic increase of 40% and a conversion increase of 10% from IPs in Mexico. This shows  that our spend on Spanish documentation and marketing materials is contributing to the brand, mindshare in Mexico, and our conversion rates in a new market.” These explicit statements help someone else see what you see. If the C-Level exec. knows they also added a sales team in Mexico they wull assume that is what led to the conversion. Without data to show an uptick in traffic and conversion data you will not convince the exec otherwise.

Conclusion

Here ends the first of many discussions on the differences between the data that matters to you and the data that you need to have to persuade others. Hopefully it was not too depressing.  Persuasion is a long and arduous process. If there is no money to change or no interest in linking localization to the broader business objectives it will take much longer and require much more effort to drive the discussions you need to have to impact the business as a whole.

Leave a comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: