Steps to creating a world-class speech solution
In general terms world-class tends to equate to a desire to achieve what is best for the customer, to build something that is a model for other organisations, or to achieve a high level of performance as defined by benchmarking and use of best practices. However there is not a sufficient definition with reference to a speech solution, and an additional complexity is that world-class may depend upon whether the solution is being seen from a business or caller viewpoint. From a business perspective a World-Class solution may be one that reduces operating costs, maximises existing technology investments, or increases caller satisfaction, whereas from a caller perspective world-class may be a solution where the caller doesn’t have to wait to do something simple, isn’t having to jump through hoops to get the information they wanted, and where the caller feels in control of the dialogue from start to end. Vicorp’s viewpoint is that a world-class solution that satisfies both business and caller needs shouldn’t start out with the goal of delighting the caller but should instead focus on satisfying usability requirements. Our goal is for callers to say that they did what they wanted to do in the way they wanted to do it. In this paper Vicorp sets out a series of steps that will help ensure that your solution is considered world-class.
Get the downloadBelow is an excerpt of "Steps to creating a world-class speech solution". To get your free download, and unlimited access to the whole of bizibl.com, simply log in or join free. |
Introduction
In general terms world-class tends to equate to a desire to achieve what is best for the customer, to build something that is a model for other organisations, or to achieve a high level of performance as defined by benchmarking and use of best practices. However there is not a sufficient definition with reference to a speech solution, and an additional complexity is that world-class may depend upon whether the solution is being seen from a business or caller viewpoint. From a business perspective a World-Class solution may be one that reduces operating costs, maximises existing technology investments, or increases caller satisfaction, whereas from a caller perspective world-class may be a solution where the caller doesn’t have to wait to do something simple, isn’t having to jump through hoops to get the information they wanted, and where the caller feels in control of the dialogue from start to end.
Vicorp’s viewpoint is that a world-class solution that satisfies both business and caller needs shouldn’t start out with the goal of delighting the caller but should instead focus on satisfying usability requirements. Our goal is for callers to say that they did what they wanted to do in the way they wanted to do it. In this paper Vicorp sets out a series of steps that will help ensure that your solution is considered world-class.
Understand the domain
The business:
It is vital to understand the business – their reasons for looking at automation, their positioning in the competitive marketplace, their contact channel strategy, and the typical call handling costs.
The calls:
In order to understand the types of calls that your agents handle Vicorp recommend that you spend time listening to calls coming in to your various lines of business and note the reason for the call, the transaction process, and data needed to complete the call. Combining this with an understanding of the volumes and durations of each call type will enable a business to decide which calls should be automated and what the associated business case might be.
The callers:
To deal with callers in the best manner possible, businesses "really need to understand the intentions of their customers as well as the way they want to be treated," Robert Wollan, senior executive for customer contact transformation at Accenture. The solution should be designed with the users as the primary focus. In order to do this one must first understand the typical caller demographics, how regularly will they use the service, their motivation for using it, and their expectations regarding identification and associated security.
Design for the users
The advent of reliable speech technology brings with it high expectations from users. As speech interaction is such a pervasive component of our daily communications, each of us have strong intuitions on how automated speech interactions should work. A successful design will have the users’ goals as the primary goals as the primary focus and should consider the following:
- Aim for an intuitive dialogue structure – callers will often come to a voice application with a mental model regarding the conversation that will take place to complete a transaction. If the dialogue radically deviates from this model then there will be an increased number of errors resulting in caller dissatisfaction. Incorporate comprehensive grammars – the full range of valid caller responses should be successfully anticipated, and can be validated through usability testing (see next section). Recognition failures result in elongated call duration and increased forced transfers.
- Ensure progressive error handling – the repetitive re-prompting commonly found in systems of the past has proven to have the effect of infuriating the caller rather than improving the situation. Error recovery messages should be designed in such a way that they mirror the human conversation model, offering a positive contribution to the interaction that is context specific, progressive and sensitive to the needs of the caller.
- Provide appropriate feedback – at each stage in the dialog the caller needs to trust that the system has heard and understood correctly. Insufficient feedback leads to confusion in dialogue traversal, reduced confidence in user interaction and greater hang-up rates.
- Minimize call duration and facilitate frequent users – if callers are used to carrying out a task quickly and efficiently, they’ll expect the same from the voice application. Call length needs careful consideration in the design of dialogue structure, prompt wording, prompt recordings, and the incorporation of non-speech audio.
- Choose an appropriate persona - A persona is the ‘personality’ the caller assigns to the speech system based on how it sounds or seems to behave. Careful consideration should be given to details such as voice type – sex, age, accent, language – speech style such as level of formality, speed of delivery, and more subtle qualities such as perceived friendliness, authority or patience. Once this is defined then a consistent dialogue script can be created. To be successful, a speech solution must be quick, easy and comfortable to use (see Figure 1: Virtuous Circle of Design). If there is a choice of service, callers will always prefer - and use - the quicker, more comfortable system.
Integrate your channels
The Alignment Index for speech Self-Service produced by Dimension Data and Cisco in 2008 showed that the area that most irritates or annoys callers about using an automated speech application is when they are transferred to an agent without any context (37% of consumers). By integrating a speech solution with CTI the contact channel maximises the effectiveness of automation and considerably increases the Return on Investment. Deploying a speech solution as an integrated customer contact channel will enable businesses to enhance customer service through segmentation, tracking caller behaviour through the different channels, and providing a consistent service across every touch point.
Test, test, test
A speech solution will invariably fail to deliver on its promises if it is not tested prior to full release. Informed enhancements to prompts or dialogues at this stage can make the critical difference that ensures total business and caller satisfaction. Vicorp recommend at least the following three stages of testing and tuning:
Design Stage - During an iterative design stage both informal testing and prototyping can be conducted to evaluate design decisions, validate prompt wording etc before expensive development work starts in earnest. Informal testing methods include:
- Evaluation of a paper-based representation of the dialogue, which may involve checking every path through the dialogue to ensure that the call flow makes sense and that prompts follow on from one another
- Role playing, with one participant reading out the prompts and another acting as the caller ‘Walkthroughs’ with representative users and key stakeholders, domain experts, other interface designers to obtain ‘expert’ feedback
Prototypes can range from extremely simple, paper-based sketches to full systems that contain nearly all the voice user interface functionality of the final system.
- Wizard of Oz prototyping creates a service with which a caller can interact naturally whilst being unaware that he or she is not using a fully automated system. This technique allows you to test out dialogue theory before implementing it; it may also be used to test and refine recognition grammars to ensure that they support the vocabulary ‘callers’ actually use.
- Rapid prototyping uses Vicorp’s xMP toolset for Agile development of a proof of concept that can illustrate the proposed functionality, and be used for usability testing at early project stages. Additional benefits are that the business stakeholders and the design team can work together to build the formative designs, phone in to hear the call flow, then iteratively adapt the design and requirements as needs dictate.
- Functional prototyping employs a more or less accurate version of the finished system. From a user testing point of view, a prototype should be close enough to the final product to allow you to examine usability questions in detail and make strong conclusions about the use of the final product.
Post-development - Typically the testing that takes place at this stage is off-site and then client-site System Functional testing, client site Performance and Integration testing, User Acceptance testing, and at least one round of Controlled Usability testing
Post-deployment – At least one tuning phase should be carried out to assess system performance and if necessary make improvements based upon real-user data. The key is to capture authentic data from callers using the live system, and then use the data to refine the grammars, parameters, and other elements of the dialogue. The reason this must be done after deployment is that changes should be made following analysis of authentic call data that is truly representative of the target population.
Realise value with a planned roll-out
In order to fully support the investment in a Speech solution Vicorp recommends that the business plans a rollout strategy that will consider how agents, and callers should be made aware of the solution capabilities. Internal training material can be created covering topics such as:
- What is the service?
- What does it mean for me?
- What is the caller experience?
- When would a call drop to an agent?
- What the caller experience has been up to this point
- How the agent should react
- Hints and tips for helping a caller successfully use the service
- Accompanying sound clips
- Frequently Asked Questions
This will enable agents and call centre staff to understand the solution benefits and assist them in endorsing the service to both peers and callers.
example: regular updates of how the service is performing such as ‘thanks to the new speech service agents have had a reduction in X type calls and have increased their sales conversion rates by Y’, ‘this week the service took Z calls, of which A were 100% fully automated’. This will counteract negative perceptions caused by system transfers to agents.
Target continuous improvement
Once a solution has gone live an organisation may make changes relating to company strategy, product offering, system hardware, business logic, or self-service applications; in addition customer needs and behaviour alter over time. This will all have an impact on solution performance.
Vicorp recommend that the business should plan for an ongoing continuous service improvement programme that will cover both objective and subjective evaluation. Objective evaluation methods typically analyse system log files, and monitor, record, transcribe and analyse calls to determine how well the dialogue performs in interaction with the caller and what events took place during each call. Subjective evaluation methods typically analyse feedback from callers in order to quantify satisfaction. This can be done in a number of ways depending upon resource availability: Callers could be transferred to an agent after an interaction with the solution, an automated questionnaire could be placed at the end of the service and callers could be routed into it if they are willing to take part in a survey, customers can be directed to the Internet through the speech solution, or prompted by agents or direct mail campaigns to complete a research questionnaire, if caller details are captured by the service, agents can conduct call-backs to obtain feedback. The results will show the solution performance and enable you to assess the potential for improvements from both a business and caller perspective.
Summary
A world-class solution can be achieved if the following steps are followed:
- Worked with key stakeholders to gather business requirements
- Qualified processes to be automated
- Understood customer demographics
- Created a user-centric design
- Included innovative functionality (that doesn’t impact usability)
- Maximised usage of channels
- Communicated internally
- Thoroughly tested
- Monitored over time
When your customers say that they did what they wanted to do, in the way they wanted to do it then you have created a World-Class solution
Want more like this?
Want more like this?
Insight delivered to your inbox
Keep up to date with our free email. Hand picked whitepapers and posts from our blog, as well as exclusive videos and webinar invitations keep our Users one step ahead.
By clicking 'SIGN UP', you agree to our Terms of Use and Privacy Policy
By clicking 'SIGN UP', you agree to our Terms of Use and Privacy Policy