The use of customer feedback metrics in call centres
Before you start, the objectives for your post-call IVR survey need to be clear. Ask yourself what do I really want to know? And what does Opinion-8 allow me to find out? It is worth getting to grips with what Opinion-8 can do for you, before you decide how you can make best use of it. These are some issues to consider:
Enhancing agent performance
Post-call IVR surveys are often used to identify agents' training needs and enhance performance. If this is your primary objective you need unbiased feedback, and so you will want to avoid agents ‘cherry picking’ customers to refer to the survey. This can achieved in one of three ways: - Positively encouraging the agent to feel they have something to gain by referring callers, that it will help improve the service they are able to offer and the systems within which they work. Incentives are also useful, such as a weekly prize for the most callers referred etc.
- Enforcing the rule that all callers need to be offered the survey and all those who agree to take part are referred.
- Automated call transfer. By automatically transferring calls at end of contact with Agent you avoid the agent influencing which callers are transferred. This is Opinion-8 'stealth mode'.
Improving the service overall & increasing staff morale
Opinion-8 is a valuable tool in helping you to identify key ways in which you can improve customer service. 90% of customers’ frustrations with contact centres are to do with the way the system works, the lack of access Agents have to information etc, rather than how individual agents handled the call. Focusing on individual agent performance therefore risks missing the main point. Agents want to be able to offer a good quality service, and are usually all too familiar with the typical frustrations of the customers, as these are often frustrations they will have with their own working conditions. If Opinion-8 is presented as a tool that offers the company the management information necessary to identify how to improve the service overall it is likely to improve staff morale. A performance monitoring approach of the ‘stick’ or ‘stealth’ variety will not have this positive impact on morale.Absolutes v. Trends
Absolutes tend to be used for PR purposes, and are therefore what most of us are used to being presented with: e.g. “95% of our customers are happy!” However, from a management perspective, it is likely to be more useful to establish trends. Knowing how customer satisfaction trends vary according to changes made in a call centre means you have information on which to act, as a manager. Trend reporting is based on weekly samples.Key indicators & question groups
As well as an overall customer satisfaction measure, you might want to get feedback on separate issues, such as (a) how well the call was handled (soft skills of agents), and (b) how well the enquiry was concluded (the ‘hard’ service).Comparing feedback across these two may be particularly useful. For instance, where there is a large discrepancy across these measures, eg. if the soft skills are very good but the hard service is poor, it may mean that your agents simply do not have access to enough information to provide a good service.
If you want to measure separate indicators, you need to develop groups of questions that relate to each: Opinion-8 will then report on performance across this group of questions, giving an overall measure for that indicator.
Thresholds
You can set ‘thresholds’ in a survey. Eg. You might want to measure how many customers are dissatisfied overall. If so you need to set a threshold for this measure. For instance, you could decide that a customer should be classed as dissatisfied if overall score is below 6 on 3 questions in a group, ie. average score is less than 2. You can then measure performance against this threshold on a weekly and monthly basis.Thresholds can be set to trigger action. For instance, if a customer rates that they are dissatisfied eg. on call handling, this can elicit follow up questions, or trigger a call back to follow up the issues. Performance overall against thresholds can also trigger alerts to managers through SMS or email.
Sample Size
It is important to get enough customers to participate in the survey in order to be able to draw valid conclusions about what your customers in general think of your service. The size of the overall customer group determines the size of the required sample. With a smaller number of customers, you will need a high percentage to complete surveys in order to get meaningful information. Standard sampling practice is to include all members of a particular group if the number in the group is 100 or under. If the group size is 400-600, about 50% need to be sampled. For larger groups, 20% of the total number of the group is an appropriate size. With 1,500 or more in the group, a sample size of 340 is generally considered adequate for a 95% confidence level. However, the following points are worth bearing in mind in the context of customer satisfaction surveys:In-built Bias
There is likely to be a built-in bias in any customer satisfaction survey, in that the customers who agree to participate in the survey may have extreme views, ie. be particularly happy or dissatisfied, meaning that the sample will not be representative. Because of this, we would recommend aiming for a higher sample than the minimum required for statistical significance.Comparisons
In order to make valid comparisons between groups, you need a valid sample size for each group e.g. if the organisation overall takes 900 calls, a sample size of 20%, i.e. 180 would be adequate for an overall satisfaction measure. But to compare results for 3 teams within your organisation, one of which takes 100 calls and the others 400 each, you will need to sample all of the first team’s calls, and 50% of the other teams’ calls, ie. 100 + 200 + 200 = 500 in total.Remember, the larger your volume of calls, the more likely you are to want to compare results e.g. for call centres based in different locations, or results for different types of calls. This requires a larger sample size.