It is ubiquitous tradition across industries to base company-to-company size comparisons on revenue. This approach is common practice in the translation industry, as well, in the form of both industry analytics/ratings and various RFPs. Try to find an example where you aren’t asked to provide your company’s revenue…
The concept is so ingrained in people’s minds that most of us don’t bother to ask the obvious: Is this a rational approach?
The goal of this article is to demonstrate that revenue-based size comparison is both conceptually inadequate and severely inaccurate, and can easily result in errors amounting to 100-200%. I also suggest a viable alternative methodology that allows us to evaluate the “real” size of the company’s translation business, based on concrete production indicators.
Comparison by Revenue: Apples to Oranges to Pomelos
Two most popular indicators of company size are gross revenue and headcount. Both are useful and relatively easy to measure, but, if applied alone, can result in an unfair comparison of apples to oranges.
There are several reasons for criticism:
- Price per word/hour varies significantly between regions, languages and companies
Companies A and B can translate the same number of words per year, but their most popular language pairs can cause their revenues to differ substantially. For instance, many Northern European companies may appear much bigger on paper due to traditionally high regional rates, while a Vietnamese company with a similar output, as measured by revenue alone, will seem smaller than it really is. Interestingly enough, if we compare by headcount the situation is reversed.
We can take the situation to the extreme and pose that both companies – we can call them A and B – are neighbors, and their production output is the same, but Company A is fortunate enough to charge its customers an average of 15% more. In this case, if compared by revenue alone, Company A will seem 15% bigger, which is obviously untrue.
Given that rates can vary dramatically – Northern European languages, for instance, are generally nearly twice as expensive as Central European ones, and just about four times as expensive as some Eastern European or Southeast Asian languages – revenue-based judgment about company size can easily distort the picture by as much as 100-200%. Less expensive vendors and/or vendors offering lower-priced languages can be severely underestimated.
- There are huge price variations for different localization tasks
Voice-overs (VOs) are, for example, considerably more expensive than translation. If a localization company does a significant amount of voice recording, its revenue will be much higher than that of a similar company that does not. The number of people involved and the hours spent on VOs is relatively modest, but studio rental or amortization and paying voice talents and sound engineers is expensive.
- Currency exchange rate variations introduce additional distortion
Developments in the last year have clearly demonstrated that currency exchange rate variations cannot be ignored, even when dealing with the most popular world currencies. The Euro has lost approximately 25% against the US Dollar over the last several years, let alone currencies like Russian Ruble or Ukrainian Hryvnia, which have, respectively, lost 50% and 66% of their value over the same period. Some of these markets are quite big. Combined, the European Union is one of the biggest economies in the world, and even the Russian translation and localization market is substantial (the language is among the top ten languages in the world by market priority).
Now let us get down to calculations and assume that Companies A, B and C all had similar revenues in 2013, and all grew by 5% in 2014. The only difference: Company A collects all revenue in USD, company B in Euros, and C in Rubles. It is easy to calculate, given today’s exchange rates, that Company A will seem bigger than B by approximately 25%, and C will appear to be half the size of A!
In the real world, everything is naturally more complicated. Most translation companies, even small ones, are global entities, and their revenues represent a mix of USD, EUR and multiple local currencies. But strong gross revenue dependence on exchange rate variations is still a fact, and comparing by revenue only yields a severely distorted picture. If we count in USD, Companies B and C will appear to have experienced a serious decline in 2014, with revenues shrinking by 15% and 45% respectively year on year. In actuality, both have grown by 5%!
- The mischievous gross revenues
In a world where everybody tries to impress both clients and the general public with gross revenues, companies that have several lines of business tend to combine revenues from all of them when publishing their figures. As far as all these businesses are a single entity (actually located under one roof), this is not an explicit misstatement, but it can result in incorrect comparisons.
Let us assume that Company A is involved in translation and localization only, and its revenue is $5M, while Company B’s gross revenue is $10M, but it performs a lot of software development or consulting that accounts for $7M/year. On paper, company B will seem twice the size of A (based on revenue alone), even though its translation business is considerably smaller.
- Global and local pricing trends
Sometimes the factors that lead prices to rise or drop in a given sector are not directly related to technology or productivity, but are the result of expectations or other sectors of the economy. For instance, the massive advent of Machine Translation (MT) in the last several years has resulted in productivity gains that are relatively moderate when averaged across the industry. However, it has significantly increased pricing pressure on companies in all areas, including ones where MT is not currently applied, or its effect is very limited. As a result, prices have generally gone down, often noticeably, and this has affected many key players in the industry. At the end of the day, a company can do more work, but experience flat or even slightly decreased revenue.
To conclude, both yearly revenue and headcount can be highly misleading when it comes to evaluating the company’s size. The obvious analogy that comes to mind is assessing the size of an iceberg based on its visible tip…
The Right Approach: Comparison by Production Volume
If we are to evaluate the company’s size objectively, the only sensible indicator is overall yearly production volume. For the translation industry, this includes the total number of words translated per year, the total number of hours spent editing or post-editing translations, etc.
If Company A employs two full-time translators, while Company B has four, but translators at Company A work twice as efficiently as their colleagues at Company B, their yearly output is going to be the same, and both companies should therefore be treated as equals in all comparisons, irrespective of their revenue. Fairly intuitive, isn’t it? Production volume is the only true indicator of the company’s real, production-based size.
While this approach causes few objections at the conceptual level, it is important to first answer a few questions:
- Is it viable?
In other words: Can we easily and reliably estimate the company’s production-based size given that each company offers a number of services, some of which are charged by word, some by the hour, and each with its own productivity level? How do we combine production volumes expressed in different units in a single formula?
- Can we create a metric that treats companies equally?
If somebody works slower, spends more time on the same task, and charges by the hour, would these inflated hours (reflecting lower performance) also inflate the company’s size? It is crucial to avoid this sort of loophole.
Fortunately, the answer to both questions is a resounding “yes”. Yes, a universal, flexible metric to assess the overall production volume of a translation company can be created, it isn’t too hard to do, and the required data is fairly easy to obtain.
The Universal Production Volume Metric: The Concept
The first step is choosing the ultimate unit of measurement.
Given that different types of translation and localization work are measured in words, lines, pages, pieces, etc., the only thing they have in common is that we either explicitly log, measure or count worktime spent, or can reverse-calculate it based on the volume done and average productivity. Hence work hours are the unit of choice, and company size is going to be expressed in yearly adjusted work hours.
The word “adjusted” here means that we need to make some calculations to do unit conversion, i.e. translated or edited words, lines, pages, etc. into work hours.
It is essential that we do not use actual hours spent for tasks originally measured in other units. Otherwise, calculations and comparisons will be unfair, because the result will inversely depend on productivity. That is, the slower you work, the more hours you spend on exactly the same volume, the bigger your company seems.
Instead, we will apply a set of reference productivities covering all tasks performed and applied universally. This way, similar production volumes will always result in similar numbers of adjusted production hours, irrespective of the actual number of hours spent.
One great thing about this approach is that we do not need to spend endless hours discussing how many new words a person can, on average, translate per hour. As long as the same reference productivity is applied to all data, comparisons will stay fair.
Absolute productivity values do not matter, they simply affect the scale. It is very important to apply correct ratios between different productivities, though, to fairly reflect all types of tasks. This relates to the ratio between the number of reference new words produced per hour and the number of reference fuzzy matches edited per hour, and other ratios like this.
In reality, this is not a big challenge. First of all, ratios between productivities are well known for the majority of popular tasks, and most companies and individuals more or less agree on them. These ratios are explicitly present in any table with TM or MT discounts. If somebody pays the company 10% of the new word rate for 100% matches, this explicitly says that we are expected to work on 100% matches ten times faster than translating from scratch. Even if original ratios we use are slightly incorrect, we can easily fix this later: It all comes to doing a simple recalculation.
The Universal Production Volume Metric: Details
As explained above, to create a universal production volume metric for evaluating company size we need to:
- Choose base categories like content (documentation, Web, etc.) translation, software localization, artwork, DTP, software engineering, software testing, etc. New base categories can be created as the need arises: for instance, categories for voice-overs and audio production, etc.
- Set a reference hourly productivity Pbase for each base category, such as the number of new words translated or number of screenshots made per hour. Productivities can always be adjusted later if anything changes.
- Create subcategories within each base category. These will cover various types of work within the same base category. For translation and localization, for example, they would include all types of TM matches or MT post-editing, for DTP they would cover various levels of complexity and/or page sizes, etc.
- Assign relative productivities to all subcategories. Each relative productivity is a ratio of the productivities for the subcategory and base category. I.e., a value of 10 for 100% TM matches means that productivity for this subcategory is 10 times higher than for translating new words and a value of 0.5 for complex DTP means that this task is two times slower compared to regular DTP, etc. Productivities can always be recalculated if anything changes.
- Place all tasks within a base category that are charged hourly in a single subcategory (Hourly). For translation and localization, this subcategory will include tiny translations charged at minimal fee, etc. This category has a productivity value of (1/Pbase). This allows us to recalculate work hours into adjusted base units later.
- Calculate the final production volume through a simple linear calculation.
- For each subcategory, divide the total number of units Vi produced by the company within a year (100% or High Fuzzy matches, complex DTP pages, etc.) by the established relative productivity Pi for this category.
This recalculates production volume for this subcategory in adjusted base units Ai = Vi/Pi. For instance, 10 translated 100% matches will be equivalent to a single new word if relative productivity for that category is set to 10. Likewise, page-setting 5 medium-complexity pages will be equivalent to doing 10 regular pages if relative productivity for that category is set to 0.5.
- Add adjusted base units for each subcategory, including jobs charged on an hourly basis, to the volume specified for the base category. This produces the total number of adjusted units for each base category At = Vbase + ∑i (Ai).
In case of translation this is the familiar adjusted total number of words with a slight twist: We have also taken into account the adjusted number of words the company could have translated using the time billed on an hourly basis.
For other base categories this is the total adjusted production volume for the category expressed in base units, such as standard pages for DTP, etc.
- To calculate the total volume of work per category in adjusted hours we need to divide production volume At expressed in adjusted units for each category by reference productivity Pbase for that category: Ht = (At/Pbase).
- Use the Ht numbers separately to compare companies by category (i.e. translation output or DTP output) or sum them up to compare companies by the total volume of production in adjusted hours.
A few important notes:
- Tiny translation tasks typical for agile localization and charged hourly (at minimal fee) need to be counted as hourly tasks, exactly as they appear in work specifications. Otherwise, serious distortions can occur in the data.
- Separate base categories and subcategories need to be created for translation into BiDi or from hieroglyphic languages, because productivities for these languages are different.
The resulting metric covers everything under translation and localization, which can get quite big, but we can always separate things like multimedia, software testing etc. This way we get a number of simple, standalone metrics that allow us to compare companies using multiple categories (i.e. various lines of business. For instance, there can be a translation and software localization category, a software engineering and testing category, a multimedia category, etc.
This distinction makes a lot of sense: there are companies specializing in DTP or testing, and their figures in these particular areas are impressive, but their translation volumes may be modest, and there are companies where translation dominates over everything else. Depending on their needs, clients can focus their attention on categories specifically important for them. For overall “total business size” comparisons, it is sufficient to simply add numbers across all categories.
The Universal Production Volume Metric: Example
A sample metric with a reduced number of base categories and subcategories is presented on the figure below to illustrate the whole process. For the sake of simplicity, it only contains two base categories – translation and DTP (Column A). Most translation-related subcategories and some DTP subcategories were added to Column B. Column D (Productivity) is the key variable – it contains actual average productivities for translation (300 new words/hour) and DTP (7 standard pages/hour).
Fig. 1. Sample Adjusted Production Volume metric – calculating totals. Click to enlarge.
You can open the live sample spreadsheet presented on the figure above using the following link.
For all subcategories, productivities are relative to the productivity for the base category. For instance, the number 10 for translating 100% matches means that productivity for this category is expected to be 10 times higher than for the translation of new words. Custom productivities are added for BiDi and hieroglyphic languages.
Yearly production volumes for various categories are entered into column G (Volume), while column H (Adjusted Base Units) contains adjusted word-counts (measured in words) for each subcategory in translation and adjusted page-counts (measured in standard A4/Letter pages) for DTP. All values in column H are calculated as respective volumes divided by productivity (Ai = Vi/Pi). Figures I used are for illustration purposes only and are not related to any actual data.
Columns I (Adjusted Base Units TOTAL) and J (Adjusted Hours Total) contain adjusted totals for each base category. These are exactly the figures reflecting actual volume of work performed by the company throughout the year, namely the total number of adjusted translated words and adjusted standard pages done (totals for respective base categories, including hours recalculated into base units), as well as their equivalents in adjusted work hours.
When estimating the total production-based size of a company that offers services in multiple base categories, the total can only be calculated through adjusted work hours, because we cannot combine words and pages in any other way. The total production-based size of the company in the sample spreadsheet slightly exceeds 130,000 work hours. Approximately 78% of this number accounts for the company’s translation business and the remaining 22% reflects its DTP business.
The calculation is reminiscent of a work specification with line items, which it actually is. There is also a difference: the numbers are adjusted totals for the whole company, and we are counting adjusted translated words or pages and adjusted work hours rather than money.
This begs the follow-up question: Is it feasible to obtain these integral data at a company level without excessive effort? My answer is affirmative:
- Any serious company bills customers based on specifications built around exactly the same task categories and subcategories.
- Any serious company has an ERP system where it stores specifications created for each completed job so that these jobs can later be billed. Any TM or MT system produces logs with statistics that are either imported into this ERP system, or are available automatically in case TMS and ERP systems are already integrated.
- It all comes to a relatively straightforward query summarizing data for each category for the whole year that produces a relatively simple table with totals.
Whether companies are ready to share considerably more detailed data is a different question. In reality, dozens of figures really aren’t necessary. All preparatory calculations can be done in-house, and for each base category, only two numbers are required: Adjusted Base Units Total (words, pages, etc.) for all work billed by the unit, plus the Total Volume of work billed by the hour. The number of adjusted hours is easily calculated using reference productivities.
Fig. 2. Sample Adjusted Production Volume metric – simplified view for publishing data. Click to enlarge.
You can open the sample spreadsheet presented on the figure above using the following link.
For the most popular Translation/Localization base category, many companies already have statistics for what I’m referring to as “adjusted translated words”, and weights/productivities applied are typically very similar across various clients, vendors and tools. In case these data are available, or you want to save time on running queries and analyzing results – it is sufficient to provide just two numbers!
Regrettably, this approach does not work so well for other localization areas like DTP or multimedia, because there are no universal notions of “adjusted DTP pages” or “adjusted voiceover hours”; more in-house calculations are required to normalize the data.
Minor complications can only be caused by slight differences in category/subcategory structures each company uses, which can be something like a different fuzzy match grid. On a metric level, this can be easily resolved with adding redundant, partially overlapping subcategories to each base category. It is sufficient to assign a correct relative productivity to each such subcategory.
For instance, if one company uses a comprehensive grid, differentiating between 75-90% TM matches, 90-95% TM matches and 95-99% TM matches, while the other uses only two subcategories, i.e. 75-95% TM matches and 95-99% TM matches, we can add all five subcategories to the metric, assigning each one a correct relative weight. This way, any company can enter its data into relevant fields and leave the others empty. The totals will still be correct and can be properly compared to one another, and the overall number of subcategories within each base category is limited.
The suggested universal, flexible metric allows us to compare translation and localization companies by actual production volume (measured in adjusted work hours). This metric is viable and makes accurate and fair comparisons possible, for both gross production volume comparisons at a company level and more thorough comparisons within each category, like translation, DTP or testing.
In order to build the final metric, it is sufficient to conduct a one-time, thorough assessment, revealing all valid base categories and subcategories and assign correct relative productivities to them. As time goes by, minor corrections can be made to the metric, new categories/subcategories can be added, and obsolete ones can be removed.
The metric is considerably more accurate compared to the traditional revenue-based approach: it does not suffer from deficiencies like strong dependency on the company’s rate level, productivity, core language pairs and/or key lines of business, currency exchange rates, etc.