“Sorry about the Meerkats, but your break down was our breakthrough,” Jerry said.
Jerry’s been a client since The Dark Ages. He called immediately after the board meeting ended to offer his apologies and his usual half-Lombardi-half-Belichick muster.
“Next time, just send popcorn instead of the damn greenbar,” I replied. Yes, this call happened in 2022, and yes, this multi-billion-dollar company still sends PAPER reports in advance of meetings. Said reports? Printed on greenbar paper. WHO EVEN MAKES GREENBAR ANYMORE?!? I need to immediately track down their off-the-grid bunkers and have a word with them. But I digress…
The Meerkats are his Marketing and Operations officers and directors. There are about two dozen of them now, and oil and water are a better mix. Jerry calls them The Meerkats because meerkats are the most murderous animal species on earth. One out of 5 meets their death at the hands teeth of another meerkat. (The ever-charming meerkats are also cannibals.)
Earlier, the VP of Customer Service, the VP of Marketing, AND the Sales Director got into it about who is in charge of Voice data, analytics, and reporting. If you’re not yet into Voice, (a) you should be, and, as a warning, (b) it’s complicated. Voice itself is pretty easy, but where it resides and who’s in charge of it? Well, that’s where things get messy.
My “break down” was not an emotional outburst but an on-the-fly version of who should be responsible for what, based on seeing similar battles in other companies.
Clearly, I’m not a call center/customer service person. (You need to like people to get those jobs.) But I am a marketer, and even though many of us are complete control-freaks, I’ve learned that a lot of the Voice stuff belongs with the people who can do something about it. For example, Customer Service/Care/Love/whatever. Plus, things like measuring sentiment are very different from a marketing point of view versus a customer service perspective, and it’s essential that both teams get actionable insights.
How many things do you measure in your eCommerce efforts? Take that number and quintuple it. That’s just the start of how many valuable things you can track in Voice. It’s overwhelming. That’s why I’ve listed my standard go-tos – the things I recommend that folks new to Voice start measuring below. This is by no means an exhaustive list. As I’ve gotten more acclimated to reading – and acting upon – Voice data, I’ve started to value new things. (This loosely translates to: I figured out what will give me the biggest bang for my buck and changed some of the things I focus on.)
First, it’s important to clarify that Voice Analytics means different things to different people. Pretty much everyone agrees that it’s the use of technology to measure and analyze Voice, but some folks believe that’s just about tracking emotional content in phone conversations. In other words, figuring out the tone, pitch, cadence, and so on to predict what the customer believes/feels so that you can change your approach in real-time to improve your overall outcomes.
In this article and throughout this site, Voice is spoken about as a medium, so when we talk about Voice Analytics, we include anything that happens with Voice on any Voice application and Voice device/assistant. Voice Analytics are not just about sentiment and emotional analysis (voice, tone, pitch, cadence, emotional state of the user, etc.) but also about Voice paths/journeys, Voice conversion, Voice optimization, Voice hyperpersonalization, conversational analysis, response effectiveness, and so on.
A solid Voice Analytics suite should help you better understand things like your user(s) behavior; engagement and retention; comparison metrics; audience demographics, infographics, and psychographics. It should help you find opportunities for operational improvements (training, compliance, performance, and efficiency) and marketing improvements (increase sales, optimize user paths, and better customer satisfaction.)
In my experience, marketers, especially those newer to Voice, focus on things somewhat outside their control (for example average contact/call length.) While those things are indeed interesting, they may be better left to your Customer Care team. This will allow you to focus on items that help you better understand your customers and prospects and how to market to them most effectively.
One of the best things about Voice is that it shows you how, where, when, and why your visitors struggle. You get to see and hear precisely what folks ask, what they understand and don’t, where they trip up, what information they need, and so on. Acting on the findings enables you to streamline your sales process for the most return and the best customer satisfaction. With Voice, you’ll also get the best VOC (voice of customer) and general sentiment information you’ve ever had.
Voice is constantly changing – sometimes things get better and other times they get worse, so it’s essential to look at your Voice Analytics over time so you can establish the proper goals and standards. Keep updating your benchmarks so, they’re current and match the real-time information you’re receiving. Most companies have Customer Service or Legal monitor compliance and quality assurance, and it’s often helpful if Merchandising and/or Marketing track the items that impact them as well. Pricing, for example.
Please note: If you’re leading a Call Center or are a Voice of Customer/Brand person, many of the things below will apply to you. With that said, the analytics packages you use may focus more on your users’ overall interaction and emotional state. This article is heavily skewed to what marketers need.
VOICE METRICS FOR MARKETERS
It’s useful to start with an overall view of how your Voice efforts are performing. It helps to break this down by type (device, assistant, chatbot, etc.), but if you can’t do that quite yet, just look at Voice overall and realize that some things may not apply to you right now. Items to look at include accuracy; sessions and session time; overall response time (how many seconds it takes for you to respond to the user); CPT (cost per transaction); the average number of utterances; the average number of queries; and overall Understanding.
If you’ve got all of those down pat, you’ll want to look at the frequency of queries; breaking down your overall Understanding category into smaller chunks (by user, by language, by accent, etc.); types of queries; a more in-depth look at overall response time by answer type; how you fare when there’s a lot of background noise; and so on. If you are using humans and bots in your efforts, it’s best to look at them separately. You can certainly look at them together as well (I do!), but, as you know, it takes different approaches to change things with people than it does with algorithms.
Other things of interest:
One of the biggest reasons you’ll love Voice overall is that the conversion is JUICY. Once your customers have their information stored and at the ready, placing orders is a breeze! With that said, like tracking your abandoned carts online, it’s very important to follow your abandoned point on Voice. Also referred to as the drop-off point, this is the place where people stop the discussion.
There are many ways to measure this, but when you’re first starting out, one of the easiest is to look at where they bailed and whose side the conversation ended. Some vendors will tell you specifically not to do this – typically because they think it’s biased – but if you use it wisely, it’s one of the most helpful things you can track at the beginning. Sort your list into two buckets – and then review the outcomes individually. You’ll easily be able to spot the themes and the areas where people struggle. Please remember that the flow of Voice conversations typically revolves around the user but is powered by the agent/bot and their scripts/flows/journeys.
Some vendors refer to the abandonment point within the context of “frustration” level. It’s essential to separate your operational frustration (things like speed and performance) from accuracy/path/journey issues. If you combine all the frustration into one bucket, make sure you understand your levers/bias.
And in case you’re wondering, trigger emails, SMS, and push notifications for Voice are the best thing I’ve seen performance-wise in the past decade. You’re going to love them!
How many visitors out of 100 are coming back and using your Voice channels again? If you can easily track multi-channel Voice usage, that’s fantastic. If you can’t, don’t let it stop you from measuring the channels individually. For best results, break things down into buckets by entry point (Voice application, Voice assistant/device, info chatbot, live chat, call center, etc.), so you can see where you’re strong and where you need improvement.
After you’ve figured out your overall repeat rate, you’ll want to determine how many days, on average, it takes for a user to come back to “talk” to you again. This number is especially helpful when looking at your Voice efforts’ success. It’s easy to get people to use you once. It can be harder to get them to come back the second time if the first conversation/connection was less than optimal and/or they came from an app. Incidentally, this is why it’s important that your first experiences are successful and positively remembered by your users.
MOST COMMON WORDS
Voice Analytics software has all sorts of information about your users’ vocabulary, how they cluster words, what they say and don’t say (the latter can be very telling), and so on. When properly organized, this information can be beneficial for your traffic (paid, organics, social) teams and your content creators.
I recommend using ALL the data, not just a sampling. You can track specific words that are of interest to you. You can look at what the data shows you. Or you can do a little of both. (My recommendation.) Also, look at these vis-à-vis the words people use in your internal text search function and your overall navigation. Voice excels at pointing out the words and phrases your customers and prospects are using that you don’t use and/or handle well in/on your sites, applications, devices, and assistants.
Common themes are also helpful. Theme reporting often comes from Customer Service, but wherever you get it, it’s great to sort into significant categories so, it’s not just one big, ole WEB bucket. When you review your Voice Analytics by/about the theme(s), you’ll want to clearly see who is going to buy and who isn’t. (If you’re not selling anything, switch out the word “buying” for “inquiring,” “quoting,” or whatever it is that your goal is.) You’ll want to know the most common themes so you can develop a plan to fix/improve the issues that arise. Plus, you should look at how people use your Voice application, assistant, etc. Many marketers spend a lot of time tracking what people are looking for (definitely helpful!) but not enough time looking at the usability/results of their Voice products and services. This information will be incredibly useful for the next couple of years, so don’t forget to track it!
As an aside, when looking at your Voice Analytics, please remember that you’re looking at unstructured data turned into insights. It’s not perfect but debating the merits of it is somewhat moot at this point. Use your insight and testing powers to see what works and what doesn’t. Also, if you use confidence scores from a vendor algorithm, ensure you know the rough formula behind it, especially how long the open window is. Many open windows are too short and thus give you very biased information. (This is especially true if you have a robust offline business in addition to your online business.)
INTENT AND DRIVER ANALYSIS
Voice Analytics are a good way to determine user intent – overall and individually.
Why are people using Voice, and what are they using it for? What themes are coming up most? What do your visitors want from your Voice channel now? What are they indicating that they’ll want in the future? How much traffic are you pushing? Who’s coming organically, and what are they looking for? What are they asking about? Product? Pricing? Competition? After-sale items? Additional products? Ordering or return information? Customer Service? Product help? When starting out, it’s easiest to break this down into big buckets. As you get better at reading and reacting to your results, you’ll want to divide the buckets into smaller areas of concentration so you can dive deeply into what matters to whom.
We’re still in the early days of Voice, so even the highest propensity intent is often unrealized. The user may have never purchased/inquired by Voice. They might not know to use their device well enough yet to figure out a logical progression. They might be distracted and/or want to come back later. Often, they use two devices simultaneously (a Voice assistant and a tablet or phone, for example.) Regardless, even if you feel there are lots of open loops – and nothing is completed – you’ll still get good information about what people want if you track it. Even better, you’ll get great ideas of what you need to do to help keep moving people down your pipeline and how to fix the areas in your pipeline that are running as smoothly as they should be.
When you look at intent, it’s essential to look at Product Requests as a separate entity. Anything that people say about your products, the uses, competitive products, pricing, and so on should be put into one big bucket to look at. This information is a veritable gold mine for your Content, Merchandising, Triggers (Email and SMS), and New Product Development teams. It’s also an easy way to gain quick, real-time data about what’s tripping your customers up so you can fix/optimize their experience. Voice-wise, many companies find that their biggest questions about products are regarding availability. (Read: if the item is in stock, how many you have and how quickly they can get it.)
USER RANKING AND PRIORITIZATION
Depending on how closely you measure intent, user ranking and prioritization may already be covered.
If not, one of the things that is useful to do with Voice – especially if you have follow-up remarketing and retargeting strategies — is measuring user’s propensity to purchase. Marketers typically use a 3-5 point scale, with #1 being the most likely to buy in x (typically 30) days and #5 being least likely (will never purchase.) Knowing where your users are in your sales funnel/process allows you to target them effectively and efficiently. (As an aside, the 3-5 or 3-7 ranking system is legacy, but even the companies who use a ton of Artificial Intelligence tend to keep their rankings simple.)
Keep in mind that Voice today is like the early days of the internet in many ways. So, calculating some version of who is coming for what and how likely they will give you something of value (an order, request for quote, inquiry, etc.) can be extremely valuable. Not to mention, Voice customers are often more responsive to follow-up marketing than other types. Prioritizing people into intention-based groups makes it easier and more economical to remarket to them with the right offers and messaging.
If you’re not interested in ranking people based on propensity to buy, that’s cool. Just do it on things that will help you market to them better in the future. (For example: behavioral, attitudinal, demographical, or something else.) The key is to break things down into more manageable buckets.
One of the best-kept secrets of user ranking and prioritization is how much it will help you if/when you use Artificial Intelligence to develop event models and predict future outcomes. And yes, AI can do the ranking for you. When the models are properly trained, it excels at it.
Knowing which users – and channels — are most likely to adopt AND be successful with your Voice efforts is critical.
Again, you can look at this in big buckets/chunks (email, paid, organics, etc.), or you can look at it by individual source. When you’re starting out, doing it in big chunks is perfectly fine, so don’t beat yourself up if that’s all that’s available to you. You’re not looking for perfection here; you’re looking for solid ideas that will help you improve your overall Voice efforts and sales.
And if you’re thinking this looks pretty elementary, tracking Voice is different – not impossible, not hard, just different. For some marketers, the legacy systems don’t yet accommodate Voice tracking all that well; that’s why it’s important to note that something is better than nothing, and you will definitely get actionable information from it either way. Plus, it’s beneficial for future spending.
You’re also likely to want to break things down in different ways. Voice data is higher quality and often far more comprehensive, so you’ll need to figure out what things you’ll use and which you won’t. It’s critical to focus on actionable information.
If you are a traditional direct marketer (two-step marketer, cataloger, sales organization, etc.), you may want to consider separating your direct channel altogether. Channel comparison will help you determine that.
PATTERN FLOWS AND USER PATHS
When it comes to the patterns of Voice commands, I leave a lot of that to Customer Service. So much depends on conversation flow; it’s a self-fulfilling prophecy. From a marketing perspective, it’s good to look at where they start, end, and where there are giant hiccups/blocks. This is very similar to looking at user paths in your website analytics. It differs because it’s typically much easier to change/improve the outcome in Voice. As an aside, if you want to purchase an outsource package, the ones that provide you with turn-by-turn (aka step-by-step) user path visualizations can be very useful.
There are many reasons to look at user pathing. Still, when comes to Voice, your biggest priority should be to figure out what your bottlenecks are and then redesign your conversation flow and overall user experience for better outcomes. Having a difficult time with this? Measure one pre-defined event at a time. Voice flows can be wonky to look at – especially when there are multiple strings – so pick one event (reorder request, for example) and look at that in depth. Having trouble looking at a series of events? Connect them in one journey/funnel and look at them by step and at the final destination.
The goal here is to effectively control your paths based on what the user wants/needs.
Identifying high-risk customers is one of Voice’s superpowers. Whether you’re getting the data from your Voice assistants, Voice applications, chatbots, live chat, call center data, or somewhere else, the conversations between you and your users are chock full of signals and sentiment. I like to look at customers who are predicted to churn so we can stop them BEFORE we need to reactivate them. Many companies, especially sales organizations, look at churn after it happens. Doing it that way is still valuable, and it’s different to market to a customer who has dumped you versus one who is kinda-sorta-maybe thinking about it.
This is one of the areas that both Customer Service and Sales/Marketing should look at. Customer Service can create detection programs and at-risk/attrition models for their agents. Sales/Marketing can create contact plans to save customers from leaving and find ways to get them back if they’ve already left.
Have a loyalty club or a subscribe-and-save program? Potential churn reporting is beneficial here. And for those bound to ask, yes, churn reporting works with or without Voice. Meaning if you aren’t yet engaged with Voice marketing, you should still do churn reporting. If you are doing Voice marketing, your churn reporting will be enhanced by Voice metrics as you’ll pick up a lot of additional information/anecdotes that you won’t usually get in traditional commerce.
Because Voice is still in its infancy, knowing “whom you are getting where, why, when, and how” is very useful. Look at the platform the user joined on, the location, the time, and what they said. It’s essential to include recognized and unrecognized utterances as well as errors. It’s also helpful to measure who is a new user and who is a repeat user. In fact, tracking repeat visitors can be so fruitful that many companies count the number of user conversations by individual number (1, 2, 3, etc.), not grouping (1-2, 3-6, 7-10, etc.)
One of the things that Artificial Intelligence excels at is upselling and just, overall, getting people to buy more stuff and more often. Upsell Probability scores your users based on how likely they are to accept upsell offers. Like Churn reporting, gauging upsell rates should be done regardless of whether Voice is part of your marketing arsenal. Having Voice and the associated Voice Analytics in your mix is a power-up. AI and predictive modeling not only tell you which product(s) best suit which people, but they can also help build bespoke scripts for every user so you can target them with just the right offer and messaging.
Tracking and analyzing your upsell probability will help you get more bang for your buck out of your Voice efforts. Many companies also successfully use it coupled with product predictions to figure out what to showcase in their ads.
Voice Analytics is great for tracking how often your competitors come up in conversations. It’s also good at tracking pricing questions, which are possible competitive issues in many cases. You can sort how people talk about your competitors by problem (pricing, service, product, etc.), product, delivery, and so on. Plus, you can match up EXACTLY what they say about the competition.
In the past, I’ve found that a lot of competitive research is way too water-is-wet for my tastes. (It’s not real intel if we already know it.) Competitive Summaries from Voice efforts changed my mind. Why? Mainly because you can see/hear what the users said about the competitors and then look at what responses/solutions/answers worked best after that. This allows you to beef up your scripts and overcome future customer objections before they occur.
WHAT ELSE? WHAT ELSE DO MARKETERS LOOK AT WHEN IT COMES TO VOICE ANALYTICS?
In addition to helping improve their sales and marketing efforts, many companies use Voice Analytics to help them identify new products, services, and features. Some use them solely for customer care efforts and quantifying issues, while others use them for lead generation and competitive research.
Whether you’re using Voice to save time, make money, or just have plain fun, whatever the reason is should dictate a lot of what you look at. Nowadays, some marketers look at acoustic features and sentiment/emotion detection. Sentiment/emotion detection looks at the feeling(s) behind the words. Most analytics packages put them into positive, negative, and neutral buckets for an overall picture of your users’ feelings. If you’re inclined, you can drill deeper into the specifics by product or category or by the emotion (happy, sad, anxious, annoyed, etc.) There’s no doubt that this data can be exciting, and if it’s not read correctly, it can cause a lot of turmoil. (Or, at the very least, be wildly misleading.)
Additionally, having a one-page dashboard that tells you all the little minutiae that happened is helpful. This is good to look at YOY (year-over-year) and possibly MOM (month-over-month) or WOW (week-over-week), depending on usage. How many discussions were started? How many were first-timers? How many did X? (X being placed an order, signed up for a newsletter, asked for a quote, or whatever else is important to you.) How long did they last by total dialog time vs. active microphone time? When were they received? What device were they from? Did they use a supporting device? (This takes effort to track but is uber beneficial to know.) What was their geolocation? What channel(s) did they use? What percentage were complaints and/or negative? What was the overall sentiment – what percentage were positive, negative, or neutral? What’s the average number of utterances per user, and how many can we match? In many companies, Customer Service provides a lot of this to Marketing as an abbreviated/summary version of their reporting. Again, track what you’ll use.
I mentioned it above, but please be careful about sampling data. Sometimes it can’t be avoided, but if you do, know the biases and contextual recognition levels. I’ve seen too many companies go on wild goose chases for a “trend” they saw in their Voice Analytics. If you see something big happening, it’s often beneficial to find the root cause of it before you “fix” it. Voice traffic is still wonky, so you’ll sometimes see wackadoodle blips in your stats.
SHOULD I INCLUDE MY CALL CENTER DATA IN MY VOICE ANALYTICS?
Lastly, one of the most common questions I get around Voice Analytics is, “should I combine my call center data with my other Voice Analytics?” Some companies do. Some don’t. The key is to be consistent. If I’m looking at things like sentiment or other voice-of-customer information, I like all the information in one bucket. However, I like to look at Voice Analytics separately from assistants, apps, etc.
Voice is still new, and we’re still finding different ways to make it better every day – that’s true for calls, chats, queries, and so on. What’s different, though, is how the process works – different devices and ways of communicating, different agents, and different intelligences. We have some clients who are whole hog on everything being “united” (yes, I charge them extra for my pain and suffering). Still, I’ve got to tell you that besides the inherent drama and pointing fingers, the actionable information doesn’t come as quickly. Also, please remember that many Voice applications/devices/assistants/etc. are powered/enabled/assisted by artificial intelligence. AI is one of the main reasons Voice is skyrocketing, and it tends to change the way we look at things. To keep on top of it, having clear delineations in your metrics is good.
What are you measuring in your Voice efforts? Have any questions about Voice analytics? Tweet @amyafrica or write firstname.lastname@example.org.
Please note: it’s critical to gather Voice data ethically and use it responsibly. This includes following all the rules outlined in COPPA (the Children’s Online Privacy Protection Act.)