Peter Wilson, Verint: Tips for speech analytics buyers

In the latest in our series where IT leaders provide their take on the purchasing process, Peter Wilson, strategic consultant, customer analytics EMEA at Verint, shares his tips for those looking to buy a speech analytics solution.

MyC. What do practitioners need to consider, before they start looking for speech analytics tools, to help determine their requirements?

PW. There's a whole series of things that really need to happen before you implement speech analytics. One of the first things you need to understand is what is it you’re trying to achieve? What's the goal? What's the strategy that the organisation is after? Sometimes there is a very esoteric statement of the customer and the Voice of the Customer. I think what you need to do is define what that means for your organisation. Are you looking at "Why are customers calling us and what are the root causes of those calls?" The other thing off the back of that, and the biggest problem I think, is what are you going to do with that information?

One of the reasons I see speech deployments ultimately failing is because organisations don't actually do anything with the data. They get the trends, they get the information, but actually the culture within the business at senior level is not to implement those changes because they see it as too difficult. One of the things that organisations need to understand and build into that planning for deploying speech is having the governance model in place that has buy-in at the senior levels, so that when we present data that backs up an hypothesis that we may have about why customers are calling, why customers are complaining, they can engender change within the business. It's having that strategic plan of where deploying speech technology is going to take you and how are you going to get there.

MyC. What kind of questions should they ask themselves?

PW. "Do we really know why our customers are calling?" I think that's one of the primary drivers. Often organisations have an idea of the primary reasons for calls, but actually there are usually supplementary reasons as well. Another question to ask is: "Are we trying to validate a hypothesis of why we think people are calling? Do we understand what our customers are telling us?" The answer generally speaking is no, they don't. Particularly when organisations have got digital transformation strategies, they're trying to push simplistic calls out to self-service, and yet organisations are saying, "Why have my call volumes not gone down? Andy why has my average handle time gone up?" They don't know the rational for it.

What speech analytics will help them do, if they ask those questions, is understand that actually the calls that you're now getting into your contact centre are more complex, so therefore your agents are having to spend longer with the customer, and it will really shine for them if they've got any issues in their digital strategy.

MyC. How can buyers convince the CFO that investment in this kind of technology is a wise decision? How can you get buy-in?

PW. That is always the biggest challenge. Mainly because the CFO's look at a balance sheet and say "Well, where does this technology fit on the balance sheet? How is it going to affect my top line or bottom line?" They don't necessarily see the connection between the Voice of the Customer, customer experience and customer satisfaction, and how that translates into a monetised value that sits on the balance sheet.

Part of what practitioners can do is translate those messages into something that means a monetary value to the CFO. One of the things I try and get organisations to do is look at the brand and the data that they hold about their customers and see those as assets, and put them on the balance sheet, and then you can use speech analytics to say, "What effect is dissatisfaction and customer complaints having on the brand? We can monetise that.” "What effect is our data and loss of our customers, what effect is that having on us? We can monetise that.”

For practitioners to get buy-in at the c-level, and particularly the CFO, if that's the sign off, it really is about monetising the value of the data that comes out of the back of it.

MyC. Are there any particular challenges in the speech analytics solutions market that buyers need to be aware of?

Often there is a confusion. Organisations see speech analytics as voice recognition, and so we need to change that mindset. It's not like speech recognition, it's not like the speaking to the IVR and recognising your account number, which is more of a one to one transactional level. Speech analytics is about large volumes of data being analysed very, very quickly, and identifying trends in that. I think one of the issues is around that understanding of what speech technology is.

Another issue relates to understanding how quickly you can get insight. What's the lag between the calls being recorded and the data being presented, and then being able to do something about that. Organisations need to really understand that timing around presentation of data, and some speech technologies are faster than others, while some are inherently slower but more accurate. There's always that trade-off between speed and accuracy.

Organisations need to really understand that timing around presentation of data, and some speech technologies are faster than others, while some are inherently slower but more accurate. There's always that trade-off between speed and accuracy.

One of the things that we get asked a lot is around real-time speech analytics compared to near real-time or historical speech analytics. Whilst near real-time is great for certain things, it's a lot more difficult to prove the business case for real-time speech analytics, because it has somewhat limited value in terms of next best action type scenarios, but other than that it's very difficult to get a business case around that. When you're looking at your large volumes of data and really getting an understanding of what's happening within the business, that's when speech technology's much better.

Realistically, when you look at the types of technologies that are available for speech analytics there are only two - there's phonetic indexing and there is transcription, or as some people will understand it, speech to text. One of the advantages of transcription is the ability to do ad-hoc searches very quickly, because once you've transcribed the calls you can then do very, very quick searches on that transcription. So organisations are either going to choose phonetic indexing or transcription, or there are certain vendors that use both of those technologies together. So it’s choosing between those two, and then really understanding what your vendor is going to do in terms of presentation of the data, how quickly that data will be presented. If you're happy to get data tomorrow, and then do something about it then, that's fine. But if want to do something a little bit more quickly then you need to look at the vendors that can provide you insights faster.

MyC. Once practitioners are at the solution selection stage, what advice can you share to help buyers find the most appropriate vendor for their needs?

PW. Once you've decided on the type of speech analytics technology, whichever path you go down, then you need to step back and look at who's providing that technology and what sits around it, because it's the services and the consultancy and the advice that sits around that. Explore if there is a consulting division that will help you build those categories, and that helps you run and then get to know that technology and how to best implement it.

That helps you understand the best practices, and how to build the right words and  phrases, and how to boost certain phrases above certain others. It's understanding what are the wraparounds that sit with the technology. Or if the vendor is just going to throw the technology in your face and say “Thank you”.