The Multilingual Challenge
Picture a company based in the UK and operating in 20 different territories. This company has successfully built, deployed, and enhanced its own bot for its local market and the results have been brilliant.
It’s very likely that this company would like to replicate its success by expanding its bot solution across the different territories in which it operates. But of course, customers in other markets speak different languages, have different needs and all interact differently with this company. Also, not all the markets in which this company operates have the same strategic relevance (i.e. amount of customers and contact volumes).
How can this company expand its solution to all the territories when each one has its own strategic relevance? How can they adapt to customers’ demands and optimize resources to guarantee a good return on their investment?
This article explores different strategies and considerations that every company should keep in mind to successfully roll out chatbots across different markets in a global economy.
How to approach the Multilingual challenge?
Before making any changes to their bots, companies should consider three factors:
- MARKET RELEVANCE: Are some markets more relevant than others? Which markets are the most strategic? Which markets have a high level of customer interaction and which do not?
- CUSTOMER EXPERIENCE: How business critical is Customer Experience to each market? Do customers require the same level of attention to detail? How sensitive are customers to a non-localized response? (i.e would UK customers be disappointed if the VA tells them to use the elevator instead of the lift?)
- CONSTRAINTS: Are there any constraints on the bot developing team? What resources are available for developing and localizing? How much time and budget are available?
Once these questions have been answered companies can choose, for each market, one of the following development strategies for their project: Localization, Assisted Machine Translation o Machine Translation.
This model is based on using the main solution (UK solution in our example) as the “master” solution for the different languages and/or regions, which we will call “local” solutions. Of course, each dialogue will then have to be localized and adapted to each language and market. It calls for resources with local language skills and time dedicated for development and maintenance but allows 100% of accuracy and control guaranteeing the highest degree of customization for bots allowing them to truly “speak like a local”.
How does this work? Teneo allows the possibility of designing and developing a main solution (Master) and, once it’s ready, to share it to as many other countries as needed using the Teneo’s native Localization feature. The local countries will inherit all the flows and logic designed in the Master solution, which will then need to be localized to the specific language/market.
The developers will, at any time, have the freedom to create and add additional local content that might be necessary to support that market. The final solution will then be made of content inherited and synched with the Master, plus local content to cover country specificness (A local solution contains about 80% of the Master’s content, which is common to all the different countries). This speeds up development and saves a considerable amount of time (60%+ savings in development time for standard solutions) and any company using Teneo can quickly localize bots in more than 86 languages.
Going back to the UK company example, Localization would be a good alternative for those markets that are strategic and with quality demanding customers and it can be done as easily as selecting which flows they want to Localize from their English solution.
Assisted Machine Translation
This strategy introduces Machine Translation services as an intermediary. In this case, user’s input is translated via these services while answers are handled via a localized solution. The benefit of this approach is saving significant time in developing a local solution while reutilizing all the content already developed in the main solution, a low cost of ownership and the possibility of controlling the outputs to each territory, hence controlling the tone and message given back to the user.
This is an intermediate approach that would allow the UK company to roll out to those markets for which outputs demand a certain degree of control to what their Chatbot is replying.
How it works?
This can be also implemented in Teneo using the Master-Local feature in a different way than Localization. They can use a Machine Translation service for inputs and handle outputs by creating a local branch in the same language as the Master solution with localized outputs. This allows for intent recognition reutilization while having the flexibility to update and fully control outputs.
This approach is the quickest and cheapest to implement and involves building a bot in a single language and having a machine translation service directly translating inputs and outputs from customers. Although it can reuse 100% of existing content, the main risk in choosing this approach is lower intent recognition accuracy from requests, and a lower control and response quality since inputs and outputs are directly translated. Since there is no control over the content being translated, user experience may not be as good as in the other cases.
When and why would the UK company apply this? Machine Translation would be useful to stablish a presence in countries with low amounts of customers and interactions where the cost benefit usually doesn’t justify setting up a localized solution or if the company lacks the resources for properly localizing a solution.
How does it work? By integrating any Machine Translation service of choice with Teneo Inputs and outputs are sent to be instantly translated by adding integrations or scripts in flows.
The table below shows a comparison between the different models explained before.
All the considerations exposed in this article might be overwhelming and discouraging at first. But there is good news to this: there is no best or worse model since they can be combined and quickly aligned to meet each market strategy.
Let’s go back one last time to the UK company: After a strategic analysis of their markets they determine that 20% of those territories are highly strategic due to its elevated amount of customers and volume, 40% are markets which are slowly growing and the remaining 40% are territories with a low amount of customers in which they still decide to be present. How can this company then cater for all those markets in an efficient way?
Localizing their projects would be the best strategy and resource investment for those highly strategic territories. For those growing markets, assisted machine translation could be applied to have some degree of control of the output quality. Finally, Machine Translation could be used for quickly establish a presence in markets with lesser amount of contacts and customers.
Thanks to the flexibility provided by Teneo, these three approaches can coexist simultaneously in the same setup and this company would be able to easily switch between strategies as their market presence increases or decreases without having to start developing from scratch or making significant efforts in escalating their projects.
Tackling the multilanguage challenge may seem an impossible task at first. However, this can be simplified merely by aligning the company’s strategy for each market with a multilingual development strategy.
By taking advantage of the flexibility provided by Teneo, companies can combine different multilingual strategies within the same project and therefore establish a presencemphasised texte on each market in a resource, time and cost-efficient way.
A Teneo Tuesday article. #TeneoTuesday