Discourse on quality is well diffused in all sectors, including that of localization, thanks in particular to the concept of Total Quality Management (TQM) and its components and more recently ISO certifications. Its real impact is more varied. Here are some elements of reflection gleaned from day to day localization project management.

Most companies are preoccupied with customer satisfaction. Its central role is recognized and accepted (proof of the spread of the « Quality discourse ») and the taking into account of competitive pressures. Furthermore, the quality approach has long presented a strong business argument for the conquest of new markets. Associated with a contained pricing policy, it makes it possible to preserve all the competitiveness necessary to the economic survival of the company. Management technique then consists in correctly controlling quality over the long term. Too many companies, in fact, redouble quality-related efforts to obtain markets, then let these efforts subside once the relationship is established with the customer.

Evolution of quality

In order to sustain this approach and, at the same time, honor the relationship of confidence established with the customer, it is necessary to jointly define the level of quality which is expected at the proposed price. Indeed, it is the customer who is most apt of defining the quality of a service being the one paying for it.
However, it is necessary to guard against excess: the attention paid to the satisfaction of the customer is sometimes extreme, even counterproductive, because expectations are ascribed to the customer that no analysis confirms. In these cases, there is a major risk of focusing on actions which the customer may not even be aware of, leaving real expectations unsatisfied.

How many times have we worked long hours on the internal consistency of a product while letting some misprints and unfortunate phrases slip by, even though the customer seemed more attached to the quality of the language. In all these cases, our approach was unadapted, because the service corresponded more to our own definition of quality than to the real expectations of our customer. It was, thus, necessary for us to emphasize the improvement of linguistic quality whilst pointing out to the customer the importance of the internal consistency of their product for the end-user. The lesson that we have learned from this problem is that one cannot allow oneself to focus on one qualitative aspect to the detriment of another. We thus note a swing from the old culture of « unacceptable quality » to that of « super-satisfied customer ».

Does unacceptable quality result from the statistical methods of the MIL STD type, which « authorize » a level of non-quality (Acceptance Quality Level or AQL), that is to say defects being allowed to remain which ruin a batch, despite being within a proportion considered acceptable. This approach represented a compromise between the level of quality, the quantity delivered or the negotiated price, which was acceptable in an economy of excess demand. Today, with supply exceeding demand, no defect should ever be allowed to reach the customer.

Quality should be controlled to the point of « zero defect » and even if the majority of companies are not at that level of control, they do perceive this as an objective. We have now to define at which moment quality can be best controlled.

Control at the source or through filtering?

The « control of quality » can be envisaged as of old, that is to say by controlling the translated documents in order to prevent any which are not in conformity from reaching the customer, or by controlling the operations and process « at the source » with prevention being given the priority.

The quality discourse advocates the passage from one to the other; control at the source replacing filtering efforts using a progressive approach. The practical impact of the discourse is known, but the levels of its development vary considerably.
To begin with, the preventative approach and the control of processes suppose means and competences which are generally lacking in the more modest companies.

This pragmatic approach is inadequate. It should be regarded as a makeshift  solution (or stopgap) pending a more efficient response and improvement at the source. In fact, filtering risks forgetting that the origin of quality is not in the control and sorting, but, rather, in the operations themselves. The cost of a « control » of quality by filtering can be prohibitive, and in the face of more efficient competition, the risk over the long term is real. The cost implied by a lack of quality and the need of a solution to contain this, weigh up to 25% of production costs.

No customer is willing to pay for the part of non-quality in a delivered product; competitors enforce pressure on prices; the customers make the market and cannot easily be controlled by their suppliers. Thus, the costs of a lack of quality cause a reduction of the margin of the company by virtue of the equation:

Margin = price of sales – costs of a lack of quality
When a procedure is complex and difficult to control, the filtering step tends to become entrenched over the long term. Even on a temporary basis, one ends up getting accustomed to this manner of organization, integrating it in daily management practices.

An interruption in the filtering or any deterioration in carrying it out quickly demonstrates the effectiveness (apparent) of its existence, with the quality delivered deteriorating rapidly.Appearances, thus, give filtering legitimacy, masking the fact that one continues to tackle the consequences whilst ignoring the causes.

Embarrassed and irritated, we can declare that we know the tools to resolve the problem, the approach to make good progress and certain good practices, but Urgency commands us to interest ourselves in finding effective short-term solutions.

This urgency has existed for months, even years. The parable of the woodcutter illustrates perfectly our obstination to answer this issue using the filtering method:
A person out for a stroll sees a woodcutter weary from trying to cut wood with a blunt axe. The stroller asks:
« Why don't you sharpen your axe, your work would be so much easier?
    - I do not have the time, answers the woodcutter, sweat dripping from his brow, I must cut wood. »

Redoubling his efforts to tackle the consequences rather than daring to pause, analyze the situation and tackle the causes... a situation which is frequently encountered.

Recently, one of our customers asked us whether we would carry out an update on a product which we had been in charge of for years. This project management had always been difficult: large unplanned volumes, unflexible deadlines, generally poor organization (documentation often being managed before the resource files). The problems of quality, owing to emergency management, did nothing but worsen. This update concerned nearly 80,000 new words to be translated within seven days. In deciding whether to take or leave the project, we were faced with the following dilemma: either accept the offer by distributing the volume among several translators, being aware of the impact of the time limit, the number of translators implied and the impossibility of editing on the already compromised quality of the product, or we refused, thus, creating an opportunity dreamed of by a competitor to criticize the work provided and the risk of losing our customer. Our choice, although difficult, was to treat the update as an occasion to inform our customer of the dangers of such conditions. Not wanting to accept this type of situation anymore, we decided to reread the entire documentation (more than 500,000 words).
All this was done at our expense, because in sum total, the customer pays for an edited and finalized project. However, the extent of the reworking, necessitated by these difficult working conditions (not of our making), could have been lessened if we had tackled the causes with the assistance and the comprehension of our customer. In the final analysis, nobody would benefit by proceeding in this way: this situation induced too much stress, entailed supplementary expense and permitted the delivery of versions which were not as good as they should have been.

Quality control: a question of priorities

We understand very well, that actual demand is based on three requirements: a « zero defect »localization service, in a limited time and at the best possible price. Current trends place these three requirements on the same level:
Quality = Time = Minimum price

In plain terms, it is what one would call wanting the butter, its money’s worth and the smile of the dairymaid too.

This equation is solvable as long as the fragile balance between these three elements is preserved. The three requirements being strongly dependent on one another, the importance granted to one is at the expense of the others, and this is what generally occurs. For example, if one restricts oneself to the aim of providing a minimal price whilst respecting impossible delivery times, quality will inevitably suffer at one time or another.
Quality = Time = Minimum price
(-) = (+) = (+)

Afterwards, if one wishes to improve on the quality of the documents, these would need to be edited which undoubtedly involves additional costs and delay.

Quality = Time = Minimum price
(+) = (-) = (-)
In the final analysis, none of the criteria will have been fully satisfied.
Quality = Time = Minimum price
(-) = (-) = (-)
Nobody will contradict the following statement: quality must be considered a paramount element and not relegated to third place. We work for the constant improvement of quality while trying to benefit from the lessons provided by the various problems encountered.

Quality: a shared effort

Among these are obviously the problems of spelling, grammar, formatting, syntax and style. It goes without saying that these quality problems lie within the scope of responsibility of the supplier of linguistic services. Parallel to these questions of a linguistic nature, there are problems, known and yet recurring, that simple precautions would make it possible to avoid.

Localizability

Localizability is the process consisting in preparing an application for its localization in several languages. An application prepared under such conditions comprises two conceptual blocks: a block of data and a block of code, the data block containing all the resources of the user interface and the code block containing only the applicable code for all the local environments and all the languages. The code block must, thus, remain identical for all languages. Only the code block is adaptable according to the selected languages. This phase of preparation, if it is important for the adaptation of the software in the desired languages, is not always correctly carried out in practice. Indeed, we find ourselves regularly confronted with problems directly related to a lack of preparation of the files. Here are some examples, resulting from the practice:

Problems of formats: Address, date, hour, measuring units

Standard formats are not available. In consequence, all the input fields and routine treatment of addresses should be able to manage a whole panel of address formats. There exist, for example, the input fields “ State» and “ Province» in many English applications. If that refers to an existing concept for the Anglophone users, it means absolutely nothing to the majority of other users. In the same way, systems of postal codes are not identical from one country to another, some comprising only figures, others comprising of figures and letters.
The same applies to the formats of date, time and measuring units. Example:

Country	Date format
United States	10/12/54
Spain	12/10/54
Japan	54/10/12

Problems of concatenation of chains

The concatenation of chains, if it makes it possible to reduce the size of a chain, can represent a real problem as regards localization. Let us take the following example:

English	French
Chain 1: one after the other	Chain 1: un après l'autre
Chain 2: The directories will be removed	Chain 2: Les répertoires seront supprimés
Chain 3: The tables will be removed	Chain 3: Les tables seront supprimées

Taken separately, these chains are completely correct. In English, the concatenation of chains 1, 2 and 3 gives a satisfactory result. But the result is less obvious once these chains are localized in French: for grammatical reasons, the word ‘tables’ being female, the concatenation of chains 3 and 1 is not correct, because chain 1 should post  “une après l'autre“  and not “un après l'autre.”  The same applies to many other languages. This problem can be easily circumvented by avoiding the concatenation of chains in resource files.

Problem of translation

It also frequently occurs that identical English words are not translated in the same way in another language.

English	French
Chain 1: Create Wizard	Chain 1: Assistant de création
Chain 2: 'Chain 1' did not complete successfully.	Chain 2: Chain 1 a échoué.

In the English version, chain 1, title of a dialog box, is also used in an error message (chain 2). In French, chain 1 functions perfectly as an autonomous chain, but not in the error message, because the definite article is necessary (« L’assistant de création »).

Use of variables

The use of variables can also lead to errors and harm the quality of the document. If in English, the system of variables can represent a savings, it can prove to be a great source of doubt and error for a translator. In the following sentence, the variable {0} can be substituted alternatively by “classification category» and “access group»:
« Optional: Type a description of a {0} »
Knowing that these two entities are of a different gender in French, the translation of « a » is impossible in French …

Resource files

The development of an application, if it is done in keeping with the spirit of possibility of its localization, makes it possible to save time, and, by the same token, money because the application will not have to be revised later on in the process of development. Even if that represents more work for the people in charge of the technical drafting, the later work is greatly facilitated.

Messages and dialog boxes

The English chains are generally shorter than the chains translated in other languages. The rate of expansion must be taken into account at the time of the design of the user interface so that the messages in other languages can be posted without having to revise the text.

Bitmaps

The use of bitmaps containing text is not recommended because of the lengthy treatment and, furthermore, the obstacle which can be represented by the size of the text.

Preparation for help

The technical drafting must be simple and coherent to permit easy recycling of the already translated segments. Particular attention must also be paid to the drafting of the software options in help. A simple modification of an option involves long research for the translator.

Timing

The place of the localization and the translation in the editorial chain implies heavy pressure in terms of planning. Indeed, the schedule established by the software publishers sets a time for the development and a time for the multilingual localization of the product. Any problems encountered in the phase of development also delay the provision of files to the supplier of linguistic services. Since the phase of localization intervening at the end of the process and at the date of marketing are defined in a fixed manner, the deadlines set are often reduced to their minimum and tend to make it difficult to provide impeccable work.
Moreover, since the time for the translation is often inflexible, we are often forced to sacrifice the time devoted to editing which, once again, harms the quality.
A better planning taking into account possible slippage related to the phase of development, would allow a better management of the localization projects, and this, on all levels.

In the end…

All these problems can, on the whole, be avoided or circumvented by keeping in mind localization at the time of product development. If such is not the case, the cost of the localization will be appreciably increased and the problems of quality will only be more difficult to identify and correct.
We can presume that everyone is aware of the impossible equation between quality, ever faster delivery and constant price reduction. But due to the growing complexity of international competition at all levels, priority seems currently be put on price and speed. Some clients may acknowledge the issue, while others simply pass the "hot potato" on to the external vendor, letting them sort it out.
While time and cost have a simple relation, quality is antagonistic to both. Linguistic quality is necessarily based on a human process, which means it takes time and costs money. If you cut back on both, quality will suffer as well.

Philippe Pailhé, Project Manager at WH&P.