Watch the Webinar replay

The first in our new webinar series took place last week.  This event proved to be a great success with 176 people joining us live on the day.

Click on this link to view webinar replay . This is the same link as the original email invitation. You will need to add in your email address to view the webinar.

Read on to also view the slides and all of the questions and answers…

Agenda and slides

• Top Tips for Quality Monitoring – Jonty Pearce, Editor, Call Centre Helper – View slides
• Quality in the Industry Today – Duncan White, Managing Director, Horizon 2 – View slides
• Using Quality Monitoring to meet Corporate Goals – Jonathan Wax, VP EMEA Nexidia View slides
• Live product demonstration – Jonathan Wax, VP EMEA Nexidia
• Interactive Q&A

The experts

Jonty Pearce
Editor Call Centre Helper

Jonty Pearce is Editor of Call Centre Helper, the UK’s most popular online call centre magazine.

Jonty has been working in the call centre industry for over 20 years in a combination of consultancy and marketing roles. During his career Jonty has visited some of the most innovative contact centres in the UK.

Jonathan Wax
VP and GM Nexidia EMEA

As VP and General Manager of Nexidia EMEA, Jonathan Wax is responsible for driving revenue growth and managing the day-to-day operations for Nexidia in Europe, the Middle East and Africa.

Jonathan delivers solutions that enable companies to manage and optimise the people, processes and technology that they deploy within the contact centre to improve their operational effectiveness and the total customer experience.

Duncan White
Managing Director Horizon 2 (Speech Analytics Consultancy)
Duncan has extensive commercial experience at board level in a number of industries, most recently being Director of Analytics (EMEA) at Verint Consulting, a top 30 US Enterprise company.

Based on his first hand knowledge, Duncan has a unique ability to interpret and apply the output of an analytics programme to a customer’s situation.

In association with Nexidia. Call Centre Helper’s webinars provide good quality information in an easily digestible format.

Click here to watch the webinar replay

Questions & Answers:

Q: How long does it take to set up a system?

A: The answer depends on a number of factors, but we can do an initial on-site demo with your own data in a matter of hours. In the case of a typical Quick Start engagement then we are normally up and running, processing audio and giving the client live access to their results in 2-4 weeks.

Q: Can the system pick up on sighs, tone, pace of speech and does the system pick up key words from the agent as well as the customer?

A: The system can differentiate between speech and non-speech, but not sighs, or tone/pitch. And yes, the system picks up on anything in the audio file. As most recorders record onto a single, mono recording, we pick up on both agent and customer.

Q: Do you have customers that would be willing to demo the application that we might be able to benchmark?

A: Yes, we do have referenceable customers. At the same time, our initial step in the sales process is a Proof of Concept where we take your audio and run it through the system and show you the results.

Q: Will this software work with CIC or Avaya?

A: Yes.

Q: Can you use the Nexidia system with NICE Call Recording?

A: Yes.

Q: Is there a way we could view a sample/ dummy version of Nexidia?

A: Yes, we can arrange for either a face-to-face demonstration or a Webex one.

Q: Would this system replace NICE or work alongside with it?

A: The system is not a call recorder, so it would not replace Nice, but work alongside it. That said, the system does have the capability to directly integrate at the switch level to enable you to perform analytics on a real-time basis. In some instances this may replace the need for a recorder, on others there will still be a business need for a recorder.

Q: Can you adjust the filters from day to day or is a set of filters initially set up on day one?

A: All of the filters with Nexidia ESI which include dates, time , time range, call length, agent, site, direction, non-talk, etc. are user configurable and can be changed dynamically by individual users.

Q: How do you maintain call quality in case of any major tech issues at exchange and huge call flow? We being an ISP, I often find agents feel a bit rushed to make the customers understand the problem, and this in turn affects call quality?

A: I am not sure that there is a technical solution to this. We have got examples of clients who have deployed analytics and you can see the impact of when a specific system is down. You can learn from this and develop a contingency plan that includes a plan to coach/train your agents.

Q: We listen to calls as a Pass or Fail instead of as a %, the result depends on if the mandatory points from the scripts have been said. The agents also listen back to calls with a QC person and team leader, this seems to work well so they can give their own feedback?

A: If I’ve understood you correctly, you appear to be scoring a call solely on level of script adherence which can be a good place to start if you know that your script delivers the quality outcomes that you want.  However, I would suggest that there may be additional areas that you might consider including in your definition of quality.

Q: Is there an industry standard for difference in scores between evaluator and calibrator and can analytics highlight and show the calls where a phrase isn’t present?

A: There isn’t an industry standard that I’m aware of.  Yes, speech analytics can do invert searches, so we can show the absence of a given phrase.

Q: How many agent calls should the quality assessor evaluate daily if you’ve got a minimum 150 agents with a minimum of four per agent in a month?

A: I would say all of them!

Q: In Jonty’s Top Tips at the start of the webinar he said to avoid scoring calls as a percentage: how should we score quality then?

A: You can continue to score as a percentage but present/report back in another format. As per the previous answer, there are other more persuasive methods you can use.

Q: If your voice analytics are programmed to find certain phrases, how does it differentiate between, for instance, sales opportunities or complaints about that product, especially where the customer is a calm complainer?”

A: Phrase Differentiation: The answer here is that the search term should be a contextually relevant phrase. So in this instance if we searched on a product name we would get both sales call and complaint calls. However, if we had 3 queries, one that was looking for complaint phrases, one looking for the sales process phrases and then one looking for a specific product, we would perform the analysis by saying show me calls with complaint and product name, or show me calls with a sale and the product. We could even look for product, sales process and complaint! And again you could do this for multiple products.

Q: I feel speech analytics is targeted at the response of the customer and not the work that the agent puts in, soft skills like  rapport building, empathy.  How can these skills be picked up?

A: The speech analytics engine will fundamentally process an audio file without any bias to one party or the other. If you have stereo recording, then you can start to analyse individual parties. In practical terms, most analytics is driven by the agent side of the conversation, because these are typically more structured, so that there are more consistent responses from the agent. In terms of the softer skills, these can be harder to measure, if for no other reason than it is hard to get people to agree a consistent measure of an individual’s calls. Quality programmes that support soft skills typically have to run an ongoing calibration programme. Simply put, if something is voiced/spoken, then it should be discoverable. If it is either implied, or covered by tone, then it won’t be.

Q: Even if we make our sample size bigger there is still a probability that we might miss a lot of bad calls. How can we decide what sample size to use?

A: This is a good question. There are tools to assist you in working out the sample size that you would need to move to. However, the advantage of the Nexidia approach is that you can analyse all of your calls, therefore you remove any issue with sampling.

Q: Most organisations have a customer survey at the end of calls. Have there been any comparisons done with the use of speech analytics customer experience view compared to real-time feedback from customers?

A: Post-call customer feedback is invaluable outcome data that should work with the conversational analysis to understand what the drivers of customer satisfaction are.  Improving and measuring those drivers should generate an improved CSAT score so they are collaborative rather than competitive.

Q: Please, in an organisation where a call centre agent is seen as “nobody”, not allowed to embark on personal development programme and is under pressure to deliver performance for his appraisal,where do you think the agent can derive his motivation? Don’t you think this kind of attitude from the organisation can change the agent’s idea of building a career in the job?

A: Agent behaviour will be driven by the way their managers behave and the organisational culture that sits around them.  Trying to motivate agents in isolation without that culture change will be difficult.

Q: My centre handles 16,000 calls; does it evaluate all within a day?

Q: Is it also able to recognise other languages?

Q: Witness system?

A: Yes

Q: Asian…?

A: Yes

Q: Are we saying that from the flags raised on Negative exp calls, then we only view and score those calls?

A: Scalability: Assuming that you have an Average Talk Time of 600 seconds, then this would be 2,666 hours of audio a day. Yes we can cope with this. Languages: Please see the earlier answers. Call Recorders: We have integrated to all leading Call recorders including Nice, Verint, Witness, Eyretel, Autonomy eTalk, Racal, Red Box, CyberTech, Mercom, TISL and customer developed recorders. Call Flags: Yes, when you click on a flag, say Negative experience, the system will present you with only the calls that met the criteria for that category.

Q: The filters are adjustable day to day, week to week & month to month, etc?

A: Yes, all the filters are adjustable by the users in real time.

Q: If percentage QA skews the results, what is the best way to truly measure effective QA when using the normal process of QA of taking samples of agent calls?

A: I’m not sure I understand this question?

Q: What happens when a client yells at you and you try to apologise but the client goes on and on and doesn’t give you the platform to apologise?

A: Coping skills are an important part of agent training and essential in situations where a solution isn’t feasible for whatever reason.  Teach agents how to defuse and ‘manage’ a customer through choice architecture and conversational redesign.

Q: Could we use the speech tags from the speech analytics tool into the Quality portal or are they limited to the pre-defined categories?

A: You can use any of the queries, or speech tags, from the main ESI analytics application within a Query Initiative in the Query Portal. You may also create unique queries to support a new Query Initiative.

Q: Is this type of analytics scalable to small call centres?

A: Technically the system can scale to small call centres. However, it may not always be economically feasible to do so. We do also have an alternative solution to address the needs of smaller contact centres.

Q: Does the cost of the software depend on the size of the company?

A: The cost of the solution is determined by a number of factors, the main factors are the volumes of audio being processed per day, whether the engagement is an initial 4-month hosted deployment or a full licence purchase. The Nexidia approach is an evidence-based one. We will initially perform a short Proof of Concept, that will validate both the technology and the business case. Typically we then move onto a 4-month Quick Start Hosted engagement. We would be delighted to discuss this in more detail as a follow-up to your attendance on the Webinar.

Q: You mentioned that % scoring is not the best way to score calls. We are currently trying to design a new form and scoring procedure and wondered what are your ideas and advice around this? Thanks.

A: From a behavioural economics perspective, there are a number of better ways to drive behaviour change rather than just presenting  a ‘naked’ score.  Red, Amber, Green can be a simple way of replacing scores, symbols comparing against peers, the norm, is better still in promoting improvement.

Q: How does someone conduct repeatability and reproducability in a calibration?

A: Calibration requires a sample of calls to be assessed across a number of variables to ensure consistency of the scores being derived.  These include all raters scoring the same calls to check inter-rater variability, the same rater re-scoring the same call to check intra-rater variability and then calls from some time back being re-rated to check temporal variability.

Q: How do you tackle FCR and CSAT in your QA form?

A: QA forms should be looking for the drivers of CSAT and FCR rather than the outcomes themselves: those drivers can be determined through analysis.

Q: How do you take care of the accent neutralisation in the software?

A: Accent Neutralisation: Nexidia develops individual Language Packs for each accent range. We have used both the US English and the UK English models with clients who have operated off-shore contact centres and we have not seen any major impact on results.

Q: What’s the technology to measure customer reactions on 100% calls?

A: 100% analysis – This is achieved using the core Nexidia ESI application. It has the capability to process in excess of 20,000 hours of audio per day on a single indexing server. This effectively means that we can scale out to support any contact centre operation.

Q: This is all good and well, but is the software only grading English calls? Does it include different dialect of Arabic? My problem is that Kuwaiti Arabic dialect is different to Egyptian, etc?

A: Nexidia currently supports 35 different languages, so the support is not limited to English. In fact, we have 3 major English variants, US English, UK English and Australian English.   Arabic – yes.

Q: You have talked today about not scoring by percentages – do you have any samples of alternative forms I could view and ideas on how I can link them into behaviours?

A: See previous answers.

Q: What sort of cost are you looking at for this application?

A:  Pricing is dependent on a number of issues, as mentioned previously.