The Devil is in the details

I often get asked this question: "How does Bureau Works compare to SDL, Phrase, Smartling, Crowdin, Transifex, etc.?" I understand why people want to frame things comparatively. Comparison matrices are a gateway to logical decision-making. You list all the possibilities in one column and then add columns for all the criteria: cost, ease of use, features A, B, etc.

It looks great and is helpful when illustrating a rationale. However, in my opinion, this kind of thought process is disconnected from reality. Reality is a lot closer to experience than it is to an abstract list of characteristics.

From an abstract perspective, most good translation management systems are identical. All of them have a file parsing mechanism that transforms files into translatable assets viewed in a translation editor that offers the translator suggestions from knowledge bases such as a translation memory or glossary. This abstract answer leaves a decision-maker clueless and they have to dig deeper into the features in order to differentiate the software:

1. Does it have an in-context feature?
2. Does it have workflow flexibility and automation?
3. Does it support multiple translation memories with different settings?

And the list could go on. These examples paint the picture that if we go specific enough, we can make correct, rationally founded decisions.

But in our experience, the answer to the question of which system is best for me lies far from a comparison matrix and resides in the experience that your users and stakeholders will have with the platform you decide to implement.

This typically has much more to do with fine, borderline imperceptible details and subjective things such as culture and customer obsession. I have seen great systems poorly implemented or simply an unmatched use-case fit deliver horrible results, and I have seen so-so systems do the trick when implemented wisely.

Software clearly has to do with experience, implementation, and use-case compatibility, just as much as it has to do with hard features. The features, as difficult as they may seem, contain an entire universe within them.

For example, let's consider spell-check. When analyzing features, most people would take a check-the-box approach. However, a closer look will raise questions such as:

1. What dictionaries does the spell-check use?
2. How frequently are they updated?
3. How does your glossary handle slang or English expressions in a particular language?
4. How long does it take for the spell-check to run?
5. Does it require a separate prompt or does it work seamlessly in runtime?
6. How easy is it to implement the suggestions?
7. How good are the suggestions?

And that's just the beginning. From a managerial point of view, a spell-checker is just a simple feature that you either have or don't. But from a translator's perspective, the spell-check's efficiency and effectiveness can represent more productivity, peace of mind, and better translation quality. It can be the sole divider between a frustrated, stressed-out translator and a satisfied one. 

It's tempting to stay shallow and focus on breadth rather than depth. The greater the breadth, the more capable any given tool may seem, and too little breadth can easily make you come across as limited. Stacking features can help you remain competitive. A check-the-box approach takes software far when it comes to search engine discoverability, instant value perception, and even valuation. But when it comes to true value, nothing beats experience.

Many software companies, just like us, are focused on exploiting the most out of breakthrough Large Language Models (LLMs) such as ChatGPT 3.5. And for so many, there's an illusion that connecting to such a powerful engine will be enough to deliver an amazing experience to users. But, in my opinion, that is just the very tip of the tip of the starting point.

Connecting to ChatGPT is great, but there is much more to consider:

1. The quality of the responses is a factor of the information that it is fed and the prompts it is fed. This, in itself, is an entire microverse. In simple terms, any given answer is only as good as its originating question, and the answer is also a function of the information any given entity has access to. 
2. The way people interact with these feeds and how these feeds interact with the translation experience is just as important, if not more so, than the quality or usefulness of the feeds themselves. If it's easy, seamless, and intuitive, it tends to be widely adopted and praised. If it's hard to use in the slightest of ways, it will be rejected even before a conscious decision has been made by the user.

The key insight for us at Bureau Works is that the details are where the magic truly happens. Details are painful because they require so much research and work to produce little tangible benefit. A completely refactored spell-check, for instance, may change the user experience but offer little material for a marketing manager to leverage as far as growth goes.

But, in my opinion, as tempting as breadth and scale are, and as painful as it is to put so much thought into apparently little things, it's the only way to truly deliver an amazing experience for our users.

In conclusion, when it comes to selecting translation management software or any software, in general, it's important to look beyond the surface-level features and consider the details and subjective factors that contribute to a great user experience. This requires research, effort, and attention to detail, but it is the only way to truly differentiate and provide a competitive advantage. The devil is truly in the details, and it's up to us as software providers to embrace this idea and iterate on the best possible experience to our users.

‍