What is speech analytics software

Speech analyst let the genie out of the bottle. Just imagine: 100% of the conversations between clients and managers are available for you to study. Not everyone is ready to apply this knowledge, but they are the future: if the data is there, the business will use it. In this article, I'll explore some of the myths and misconceptions that prevent marketers and salespeople from successfully using new customer and deal data.

Misconception 1.

Without verbatim decoding, speech analytics is useless? Speech analytics software is a machine learning-based technology that translates human speech into text format. In recent years, for a number of technologies, the error rate does not exceed 4–5%. And although decoding of Russian speech is complicated by a large number of phrases in a sentence, the recognition accuracy reaches 80%. That is, 8 out of 10 words will be recognized and available for searching for the desired dialogue, tagging, training the neural network taking into account the context.

We have heard from clients more than once: speech analytics does not provide a verbatim transcript, in sentences not all words agree, how to read it? But, attention, the question is: does the employee who listens to the calls translate them into text? Not! It tags, marks, finds pivot words. If you make him transcribe what he heard and type the text, you will have to pay a whole company of specialists 24/7.

If we take the entire array of records as a unit, we get a potential ""coverage"" of data for a person - 9.9%, for a machine - 80%. The difference is almost 10 times!

No need to wait for perfect transcripts. It is enough to take into work an array of data in the broken language of artificial intelligence and in a few clicks find calls in it that need to be listened to by a specialist. And if before connecting speech analytics the employee analyzed random 5% of calls, then after that he can perform the same amount of work, but only for those calls that require special attention. For example, find calls in which the agent and / or customer used a specific word.

Misconception 2.

Every conversation is unique - the machine algorithm will not understand the essence Marketers and sales managers should jointly determine the parameters that characterize a high-quality, targeted call. And this is not a whim, but a vital necessity, taking into account the trend for work throughout the sales funnel. This exercise is also useful for those who are not yet thinking about implementing speech analytics.

Our experience shows that about 40% of companies do not have such criteria, or the opinion on this matter is very different from marketers and sales managers. Hence this eternal confrontation stretches.

Meanwhile, a telephone conversation, which most likely leads to a sale, is always distinguished by a number of strong points. For example, it can be questions about delivery for a furniture store, mentioning the name of a specific doctor for a clinic. The more there are, the higher the probability of a sale. To find these triggers, you need to listen to at least 10–20 conversations, including separately those that led to the sale and those that ended in nothing.

What is important to a person, for example, when ordering rolls? Waiting time, cost and delivery time, discounts, quantity of ginger and sticks. Or take a call to a real estate agency: price, area, mortgage, finishing, contract, guarantees, deadline. It doesn't matter if a young freelance mom or a gray-haired professor calls. In everyday life, they may speak differently, but when ordering goods and services, they will use the same clear and specific language.

To determine the required set of phrases, you need to start from the tasks: studying the needs of customers, the work of sales managers, identifying certain categories of calls. Below is a fragment of the transcript and highlighted words that can be such markers.

Misconception 3.

It is enough to listen to 10% of calls to find out the average temperature in the hospital From the point of view of statistics, a sample of 10% of calls excludes ""congestion"" when listening to the most successful or, conversely, only failed ones. For the average temperature in the hospital, this is really enough. You will understand how, in general, employees are transmitting benefits, and you will selectively recognize the reaction to certain offers. If someone messes up, you can talk to him, and he will correct himself.

What if you listen to all similar unsuccessful conversations and find the problem? Perhaps the training is out of date, the motivation is not working. Or find not one or two, but all the conversations where clients praise the manager - can you find something in common in them, adjust the business process? What exactly leads to this result? The fact that when analyzing 10% of calls is an error, when listening to the entire volume it is a growth point.