I shouldn't really have been surprised. On opening the box containing my latest mobile phone, the only documentation to be found was an almost wordless instruction sheet showing the basic parts of the phone. That was all. Not even a Quick Start Guide. Those of us who have worked in the language services industry for a while can only think back nostalgically to the days when products came with user manuals. It was a time when "delivering on time" for product launches comprised an initial round of translation on pre-release content, followed by a second round on the updated, final content, then the insertion of images, final DTP and delivery of the quality-assured multilingual manual directly to the printers to "save time". The old way and the new way But nowadays, projects have gone agile. There isn't the time or need to produce a 40-plus-page multilingual user manual. New products and their variants are launched frequently, and user assistance is built into the product or provided online. Today, the two-week-sprint cycle of agile development results in "continuous localization" with a relatively constant stream and high volume of small chunks of new and updated text for translation. The distinct deliverable of “the manual” is no more. This isn't completely new, however. The automotive industry has been doing continuous localization of service manuals and technical service bulletins since the mid-1990s. The more advanced language service providers (LSPs) have been working on their technical stacks to support this kind of workflow since then, too. Over the last 30 years, continuous localization has spread across market verticals and content types, going fully digital and, in cases such as my mobile phone, ditching the printed manual altogether. This has been made possible because original source content and translations are increasingly managed in content repositories: for websites, marketing communications, documentation and software. New and updated chunks of content are pulled from the repositories for translation and pushed back after translation. And for some content types, the requirement is for a non-stop, 24/7 translation conveyor belt. The need for automation In the continuous localization scenario, the process needs to be automated: exchanging content to and from translation, preparing the files pre- and post-translation, assigning translators and carrying out quality assurance checks. As a result, LSPs have implemented workflow technology to automate as much of the process as possible and use approaches like Lean Six Sigma to find and eliminate waste and improve business processes. Indeed, today, most LSPs support continuous localization as a standard part of their service delivery. Focusing on the “happy path” is not the whole story Given the length of time continuous localization has been around, you'd think that source content should be able to whizz through a workflow with hardly any friction at all. Going by the 80/20 rule, get 80% of the process as efficient as possible, reduce the cost of delivery, increase margins, improve time-to-market and achieve better customer satisfaction. It should all be done and dusted. Right? Perhaps not. This nirvana-like vision assumes that translation is a factory-like process—that source words are fed in at one end and are turned into translated words at the other as if on a manufacturing line. To a large extent, continuous localization programs need to think like this in order to function. Businesses design processes so that 80% of the content contributing to 80% of the revenue flows along the route of "the happy path"—the standard set of automated steps. But if "the happy path" gets all the attention for process efficiency, the 20% of content that doesn't flow through the happy path can end up contributing to 80% of the cost. Unfortunately, the world of workflow automation doesn’t pay too much attention to the 20%. Why is that? Most theories about workflow and the workflow tools on the market consider the 20% as exceptions: “unusual” or “abnormal” events to be minimized, or variations to be coped with within the tolerances of the standard process. This suggests they are unique cases in relation to the workflow, hard to predict and difficult to manage. But what if the exceptions and variations are a natural part of the process? When exceptions are the rule While flying, for example, we can think of the smooth cruising between take-off and landing as the consistent "happy path" that can be automated using autopilot. The take-off and landing, although one-offs, are not exceptions or variations; they're essential parts of the flight that generally require the intervention of the pilot because of the variability of conditions that might be faced during those manoeuvres. In every continuous localization program, we see events that, from a happy path perspective, might be considered exceptions and variations, but are a natural consequence of the inherent change in the kind of work that we do. (Think of adding or removing languages, changing content and cancelling submissions.) And these events need to be explicitly managed rather than minimized or added to the tolerance of the workflow. What happens when these events are not managed well is akin to when, in a stream of traffic on a multi-lane highway, a slow car or truck changes lane, a vehicle breaks down or an emergency services vehicle tries to get through, sirens blazing. There's a disruption to the regular flow of traffic. If the vehicles behind aren't redirected in some way, the event can end up producing a massive traffic jam. Just like an aircraft has a pilot and crew, and smart motorways have introduced active traffic management via cameras and overhead gantry information, translation services have a project manager with a team to step in and sort things out. However, exception or variation management is not something that's tackled much in the world of workflow studies. And it's dealt with even less by off-the-shelf workflow tools. So how does a project team effectively manage the 20% of change that occurs naturally in a professional services environment when the process has only been honed for the 80% happy path? The essentials of exceptional exception management There are three elements to supporting effective management of the often neglected “unhappy path”: Good taxonomy: a way to classify the exceptions and variations to help triage them quickly and set up workflows designed to deal with each type in the most effective way possible. While it takes some effort to establish this framework, doing so helps inform the strategic approach to managing the different exception types in the most cost-effective way. A classification framework for handling exceptions is one of the outcomes of the Workflow Patterns Initiative. Good technology and data: a workflow system that provides a way to visualize what is going on in the process, enable the project management team to step in quickly and focus on what needs to be done and prevent a traffic jam with more issues to deal with. As commercial workflow systems don’t handle exception management well enough for continuous localization, language service providers tend to develop their own proprietary solutions or customized solutions that wrap around commercial systems. Even better if it’s informed by the framework for exception handling. Good project management: an experienced team trained in how to work with continuous localization programs, set up automated workflows for the 80% that travels along the happy path and manage the different methods to deal effectively with the 20% that does not. The team also needs to be knowledgeable in the exception handling framework, which comes from experience gained over the years with some of the most demanding clients and projects. Planning for it all Managing continuous localization programs requires a high level of localization maturity from both client and provider, time to set up and reach a business-as-usual state and a holistic approach to managing the flow of work—the regular and predictable as well as the exceptions. Naturally, we want the 80% majority of content to flow as efficiently as possible. But, the remaining 20% needs attention too. And it's a risk to ignore it. To be successful, continuous localization programs require more than automation of the happy path. In the fast-paced professional service of continuous localization, the so-called exceptions in the 20% are as mainstream as the 80%. Think about that the next time you fly in a plane, cruise along the highway in a stream of traffic or open the next user manual-free box of your latest IT purchase.   Many thanks to Solutions Architect Stuart Sklair for this insightful post!