In part 1 of this article, we talked about the general Key Performance Indicators hidden in your call recordings. In this segment, we’ll deal with the gold standard of KPI, the Net Promoter Score.
Net Promoter Score
Your Net Promoter Score (NPS) is generally considered the most important Key Performance Indicator (KPI) that your company measures. It’s a consumer perception measure based on the following question:
How likely are you to recommend [your company or service here] to a friend or colleague?
But obtaining that measurement can be deceptively tricky because collection methods and NPS itself suffers from some flaws. In this article, we’ll discuss how your call recordings can help you fine-tune the number you’re getting and, more importantly, focus on the specific reasons for that number. Knowing the cause of a low-performing NPS will tell you where you need to make changes to start seeing better KPI all over.
Calculating Net Promoter Score
In acquiring data to build an NPS, companies use surveys to ask, “How likely are you to recommend us?”. Whether the survey is a live phone call, web form, text message or email, customers who are contacted will be asked to rate their answer from 0 (nope, never, no way) and 10 (yes, absolutely yes). Their rating will put each respondent into one of three categories from which the final NPS is calculated:
Promoters
A score of 9 or 10 rates the respondent as a promoter. These individuals are considered enthusiastic customers.
Passives
A score of 7 or 8 rates the respondent as a passive. These individuals are considered satisfied customers who aren’t happy enough to be promoters.
Detractor
A score of 0 to 6 rates the respondent as a detractor. Not only is this customer dissatisfied with your company or service, but they could also discourage others from turning to your company.
NPS Challenges: Are You Getting The Whole Picture?
In a word, no. NPS holds sway over customer service managers worldwide, but it has some serious flaws that need to be understood to get a working understanding of your customer base. These flaws stem from looking too heavily at the final NPS while failing to include other data that paint a 3D picture of your customer sentiment.
Intent to recommend is a one-dimensional measure of customer satisfaction:
NPS seeks intent to recommend but neither reveals nor measures other outcomes. A client may not recommend you because they covet your service’s exclusivity and don’t wish to expose you to others. You now have a completely satisfied but non-promoting customer. Your NPS has a flawed single angle not detected in that single NPS number. You’ll need more data points from your customer conversations to assemble an accurate picture of this customer group.
Promoters and Detractors need more data to identify them properly:
The simple fact is that a customer who scores your company a 6 might still be loyal and even basically satisfied but would like specific improvements in the product or service. But a large enough sample of these “detractors” could mislead your customer service efforts. Should you launch a whole new customer service initiative or just pay attention to support requests? If you’re going strictly by your NPS, you don’t have the data to make an informed choice.
But why? NPS doesn’t tell you about your promoters, passives and detractors
Repeating the survey on the same respondent shows low reliability in scores after only a small number of days. It’s essential to recognize that NPS is entirely an outcome KPI. Managers must use analytics tools that can derive understanding from rich data sources like customer interactions.
That’s not to say that NPS isn’t predictive, but searching through interactions unlocks customer motivation and sentiment. Keywords and phrases can be searched in a company’s call volume to go a step farther and find individual customer feelings and trends across your entire customer population. Deep analytics can help you to find out how many customers are mentioning your competitors, how many are dissatisfied with your website, how many are having a support issue, and so forth.
All things being not equal: sample size
Calculating an NPS from a customer sampling lower than 100 increases the risk of an inaccurate result. For example, in a sampling of 50 customers, each customer accounts for 2% of your NPS. Larger samples are better both in your NPS sampling and in your analytical survey of your call data.
Overcoming Information Gaps in NPS
We’ve talked about the 3rd dimension of data that comes from your call analytics. Here are some practical applications for your call data that AI-driven analytics and speech-to-text transcription offer that will open your ears to the customer’s voice.
Contact Quality Management
That 3rd dimension of knowledge culled from your call data is crucial to making sense of all your KPI, especially Net Promoter. Knowing you have a dissatisfied customer is not the same is knowing why you have one. That’s why it’s vital to train your call agents to probe for the right information. Aside from simply making for a more informed agent on the call, the data your agents push for will elicit a larger body of useful knowledge to analyze. Build your approach to obtaining rich customer data around improved contact quality between customer and agent.
Call scripts: find the needle in the haystack
Call scripts can be employed to mine crucial data from your customers that your analytics engine is hungry for. Even if your customers don’t say the exact key phrase you’re searching for, you can pinpoint where in a call they spoke about the keyword’s topic. One way is to have your agents follow a procedural call script, which allows your analytics engine to target when they brought up the topic. That will help you classify calls that don’t have the requisite customer-spoken keywords but have critical sentiment or information.
For example, perhaps you’re searching for customer complaints about subscription fees. To do this you’re using the search terms subscription, fee, cost, and too much. Now say you have a customer who simply says I don’t want to pay that. In this instance, you’d miss that customer complaint. However, if your call script includes topical responses for your agent, then you might have your agent respond with “I understand you want to discuss your subscription fee, is that correct”. This action inserts the necessary keywords into the call audio without the customer needing to say it.
Call Metadata: What Your Customers Can’t Tell You
Too often, companies ignore a data set that speaks directly to your call center efficiency, most especially in its automated procedures. This metadata contains statistics about each call that can point to inefficiencies that are negatively impacting your NPS and other KPI. Metadata can start by answering some basic call center questions through statistical analysis. Questions like:
How many callers never make it through your outbound auto-dialer?
What’s the average call length of successful sales staff?
What’s the average call length of lower-performing sales staff?
How many calls hang up while the caller is still navigating your IVR system?
Let’s take that last question and run with it a bit. The metadata can tell you not only how many people hang up, but at what point in the call they did so. This can be very useful data. Consider this scenario:
In your daily call volume of 1,000 inbound calls, you notice a hangup rate of about 150.
30 of those hang up in the first 10 seconds; those are unavoidable wrong numbers or accidental disconnections.
The other 120 hang on the call long enough to make a language selection. However, 60 of those calls hang up around 25-30 seconds. Examining your IVR menu prompts, you learn that it’s about 25-30 seconds until your customer picks a department to be routed to, but after that your IVR forces more product questions out of them, and that’s where they hang up.
How do you unpack this pattern to understand it? Well, maybe a few were disconnected, or distracted by another call, but the likelihood is that almost 40% of your hangups are frustrated customers. They chose a language, a department, and then decided to bail when your automated system wouldn’t put them in a queue or, better yet, on the phone with an agent.
Here’s a place where your customer isn’t communicating with you but still might wind up being a detractor. However, with the understanding gleaned from your metadata, you make the executive decision to cancel that secondary department selection in your IVR menu, and instead route the call to a queue. The customer will be more than happy to speak to a human even if they have to be routed one last time.
Call Recordings Are The Key to High KPIs
Without a doubt, your call data represents an enormous collection of unfiltered and honest customer feedback. Through AI-drive speech analytics, you are privy to your customer’s feelings and sentiments about your service or product. Grading your customer’s level of satisfaction can be difficult if you rely solely on solicited feedback. Reach out to us today to talk about turning your call data into a full-time, on-demand customer consultant that can show you how your customer really feel about you.
Brian is a freelance technology writer and media editor based out of Central New Jersey. He’s logged 20 years of experience in the Telecom industry and side-hustles in the record industry. Brian started his career in technology at a company that made analog modems. He migrated to a marketing career in the call recording industry where he learned exactly how and why calls are monitored for quality assurance. These days Brian fuses his skills together to deliver his researched observations about telephony and compliance laws in polished articles and videos. He’s also composed the music for a long list of big Hollywood trailers. He does not miss the sound of analog modems but he is endlessly fascinated with phones.