In this article, we are going to go through the most common call center QA terminologies that are used. Some of them you will be familiar with, others maybe not so much. However, I will separate them all into headings so that you can skip to the section you would like to know a little more about.
We will start the article off by telling you the typical call center QA terminology. I will put them in alphabetical order, so feel free to jump about the page looking at what you need to.
An active calibration is a call that is currently going through the process of calibration. That can be from either a Call Criteria analyst (Internal) or a team member of the company who owns the call center (client). It is the process of running through the human or automated scoring of a call and checking that it is correct.
Agent ranking is a statistic that shows how well or poorly a specific agent is performing in comparison to a group of other agents. That metric can either be as a percentage of correct points or by number ranking.
Agents are the people who are designated to make or receive calls on behalf of the call center. They can be one of three types:
No matter which of the three types an agent is, they are all called “agents” in the QA process, as the type is usually insignificant.
Artificial Intelligence is a machine that is programmed to do something for a human. That could be anything, but in the call center realm, it often understands human speech to the extent that takes certain aspects away from the agent. For example, AI can retrieve specific details, such as name, social security number, etc., to save time for the agent.
An Analyst is someone who analyses the data from any given field, such as the keywords of speech analytics, to the scores of agents based on scorecard results. In a call center, the analyst is responsible for the delivery of services from the call center to the customer and ensures that there is a consistent output. In call center QA, the analyst will ensure that the scorecards, systems, and processes that provide the call center with their statistics are correct.
A bad call is any call that is not scoreable because of one or more reasons. Those reasons can range from a weak signal, meaning that the agent cannot hear the consumer, to a bad recording, meaning that the human or AI QA team cannot score it. Also, see non-scoreable.
All calls are scored based on scorecards requirements, and a certain percentage of those calls are calibrated to ensure that the initial scoring was correct. If 10% of all calls are calibrated and accurate, then we can assume that the majority of the further 90% is right, too.
Some call center QA providers will charge by the amount of time a QA team needs to score a specific call. However, Call Criteria charges by the call minute, which is how many minutes a call lasts.
Different sections of call centers will have different aims, such as sales, customer service, lead generation, etc. Each of those sections can be called a campaign, and you can run various campaigns within each of the sections based on different scorecards, scripts, etc.
Coaching is the term used for training an agent on specific sections of a scorecard. If an agent is performing poorly, and regularly not completing a scorecard question, recorded line statement, for example, then a supervisor will carry out coaching to help the agent with those points.
The Call Criteria dashboard is where you will see all of the information of your calls and agents via a web browser.
Deep learning is a multi-layer computer algorithm that takes outputs of one or more computer systems and generates a meaningful output.
All calls go through the scorecard process. Some of the points (see the scorecard section), if missed, will create an instant fail. The calls with one of those points missed will be a failed call.
Speech analytics will look for specific words in a call to automatically score it. The words that it looks for are programmed in via a keyword syntax. For example, speech analytics will search for terms such as recorded calls.
Machine learning is like Artificial Intelligence, but it learns from its surroundings and multiple inputs, instead of a single track of finding words in a transcript, for example.
Missed items are parts of a scorecard that the agent did not say. If there is a requirement for the agent to say where they are calling from, and they do not, that would be a missed point.
Not interested is a term used to show that the prospective consumer told the agent that they are not interested in the service. Some call centers require that to be a statistic to gauge how many calls they have where people are interested or not.
Natural Language Processing is the process of turning the language we as humans use into computer-readable information. That can be in the form of text or voice and multiple languages.
A non-scorable call is similar to a bad call. However, the things that make a call bad are if the call gets disconnected, if there are audio issues, it was the wrong agent, etc. A bad call and a non-scoreable call are sometimes interchangeable.
There are many times where an agent, QA analyst, supervisor, or manager needs to create notes on the system to inform others of issues or rectifications to concerns previously raised. These are called notifications and will show in the notification queue.
Some call center companies choose to have their QA team as an internal section to their own business. However, others see the benefit of using a different company for their QA, and that company is an outsourced QA. Call Criteria is one of the best examples of an outsourced QA.
A predictive model is a way of using previous statistics to predict future outcomes.
Quality Assurance (QA) is a set of well designed, highly organized activities that are set out in an easy to follow manner to produce the highest quality possible from the beginning, right through to the end of a process is as efficient and effective as possible given a set of defined parameters.
Quality control (QC) is a procedure or set of processes intended to ensure that a manufactured product or performed service adheres to a defined set of quality criteria or meets the requirements of the client or customer.
A rebuttal is a word that means disputing a prospective consumer’s argument as to why they do not want to complete a purchase.
Also, see calibration. When a call is calibrated, that can be by a client user, or Call Criteria agent. However, sometimes, some of those calls are recalibrated by another user to check that the calibration check was correct. That is a recalibration.
If someone listens to a call that has already been through the scoring process, and they believe the score to be incorrect, they will ask for a regrade. That is when they run the call through the scoring process again.
Reviewed is a status that indicated that a call has gone through the QA process and scored accordingly.
A score is a total score that an agent receives throughout either a single call or a group of calls over some time. If the agent gets all of the scorecards points correct, they will have a 100% score, but if they miss some points, a percentage of that score will be deducted from the 100% to provide an overall rating.
A scorecard is a list of items that an agent needs to say during a call for it to have a 100% score. Those points are varied from call to call, and also on each scorecard. Scorecards can include anything that your business requires, such as basic introductions and recorded call statements, through to more specific items such as asking precise questions based on your requirements.
Speech analytics is much like NLP in the fact that a computer uses algorithms to decipher which words have been said, and in which order. Speech analytics is often used to score a call without human intervention. However, there are more common examples of speech analytics, such as Alexa, Bixby, and Siri.
SQL or Structured Query Language is a programming language that is used for managing data in a localized database.
Telephone Consumer Protection Act 1991 (TCPA). That is a set of guidelines that the call centers, agents, and QA teams have to abide by to keep sensitive data, such as the social security numbers, driving license numbers, etc. from being recorded and maintained.
A transcript is a product of when speech analytics turns the voice of someone into a text format.