In part 2 of his blog series, Tim Harbers, CTO at Contexta360 looks at term trends.  

Welcome to the second blog on how to find and get value out of trends in speech analytics. Part one considered trends that involve metrics you can follow over time, such as call volume and agent performance.

This blog looks at terms that are popular, which is one of the most common ways to show an overall trend. See, for example, the way that Twitter displays trends:

Wikipedia defines a trend as “a collective behaviour that lasts a short time”. So a trend is always connected to a certain time period, usually the one that is most recent. “Short time” is always a relative term, and it can range from looking at the past few years (macro trends) or much shorter trends such as the past 15 minutes (micro trends).

To find trending terms, the system evaluates the occurrence of each term and compares it with different periods of time. For example, the past hour is compared with the whole of last month. If the increase in occurrence of the term in the recent past is significantly higher than on an average day, the term is considered to be trending. Making a comparison is important in deciding if a term is part of a trend or not. Just using the most frequent terms from recent conversations will not work, because these terms may always occur frequently. That doesn’t make them trends.

There is also an important difference between relative increase in percentage and absolute increase. Consider the following example. The word outage is used on average in 1 per cent of each conversation. However, last week this percentage went up to 4 per cent. On the other hand, the word payment is normally used in 30 per cent of the conversations and this increased to 35 per cent last week. Which of the two words count as a trend? Even though payment has increased more in absolute difference (5 per cent vs 3 per cent), the increase in the word outage is much more significant in relative terms (4 times vs 1.17 times), assuming there are enough conversations to make even small percentage differences statistically significant.

On the other hand, you don’t want to see too many trending terms that rarely occur, even if the increase is interesting. This is especially true when your dataset size is relatively small. So the best algorithms for finding trending topics are always a trade-off between frequency and difference in frequency. One example is to calculate the z-score (or standard score) for each. term:

z-score = ([current trend] – [average historic trends]) / [standard deviation of historic trends]

When a z-score is used, the higher or lower the z-score the more abnormal the trend. When the z-score is highly positive then the trend is abnormally rising, while if it is highly negative it is abnormally falling. The z-score is also used in the calculation of TF-IDF, which is typically used by recommender systems.

Trends can be visualised as a list of terms, like in Twitter, or can be displayed as a word cloud. The latter approach gives the user an instant visual indication of the frequency of the term, as the font sizes reflect this.

In addition, words could have different colours, for example, to indicate their type.

Word clouds can be used in several analyses, depending on the use case. The use cases can include:

• Recent trending terms
• Agent group trending terms
• Agent trending terms
• Topic trending terms
• Metric trending terms.
• Part of conversation trending terms

Recent trending terms

Recent trending terms are most relevant for call centre managers. A live dashboard for team managers can show word clouds from the past 15 minutes, the past hour and the past day. This means quick action can be taken whenever something happens.

Agent trending terms

Rather than comparing trend frequencies over time, it can be an interesting exercise to compare term frequencies of a single agent with the term frequencies of their entire team. This can offer insights into the agent’s performance and their typical way of speaking.

By knowing which words are used more frequently than the average, QM managers and agent coaches can get better insights and give more personalised coaching advice. Not every agent-specific trending term will be very interesting, for example the name of the agent will almost certainly be trending this way. Other words, however, can give indications on where the agent can improve, for example words connected to empathy and words that can indicate that the agent is not so sure of themselves.

Topic trending terms

Trending terms can also be used in relation to topics. For each topic, term frequency is compared with the average. This means a word cloud can be generated with the most correlating terms. Topic trending terms can be put into a few categories:

Terms that are (synonyms of) the topic itself

These are the most straightforward ones and also the least interesting.

Terms that describe causes and effects of the topic

These terms have a clear correlation with the topic. For example, if the topic is “internet connection problems”, these would be words such as Wi-Fi, modem, cable, and so on. Usually the words are self-explanatory, although it may give the user more insights into the main causes and side effects of a certain problem.

Words that describe the sentiment of a certain topic

Some topics are more sensitive than others and sometimes a topic almost always leads to an angry customer. If that is the case, then the topic word cloud will include emotion-related terms (for example: terrible, disappointed, great or happy). This gives the business analyst a quick indication of how the user feels about each individual topic.

Other terms

Sometimes you will discover topic-trending terms that you didn’t expect to see. Still, these terms can correlate with the topic in a significant way. Being able to zoom in and drill down on these terms helps you to get more context and to explain why they are correlating. Hopefully you have uncovered new and interesting subtopics!

Metric trending terms

Agent scorecards have multiple metrics on which to score the agent. Comparing terms used in high-performing interactions with those underperforming interactions can also sometimes generate interesting terms.

One example is to check for interesting terms in conversations with high silence. Those trending terms could potentially give insights into the causes of these silences. Finding terms related to conversations with low sentiment might give you insights into the potential reasons why customers have low sentiment. And viewing word clouds related to First Call Resolution can give you valuable insights into which business processes still need to be improved.

Filtering on parts of the conversation

It is not necessary only to analyse trending terms in the conversation as a whole. You can also focus on the parts of the conversation that interest you the most. The two most common conversation filters are speaker-based filtering and timing.

Conclusion

Finding interesting trends does not have to be limited to looking at line charts the whole time. By detecting trending terms from a relevant dataset (topic, metric, agents) and a reference set you can unlock a lot of value for all areas of your business.

Tim Harbers, CTO at Contexta360 looks at how to detect trends in the call centre and act on them to avoid future issues.  

No day in your call center is ever the same. Changes happen all the time and on all fronts: new agents join your team, other agents retire. You get new customers with new wishes and demands. Your company offers new products, launches new campaigns and sometimes new problems will occur that sometimes go unnoticed. And just as easy, external events like a pandemic can happen out of nowhere and impact every division of your business in a major way.

Catching new trends in your callcenter very early on and acting on them immediately can avoid lots of problems later on and can beat the competition. This post contains several examples on how to detect and investigate interesting trends in your call center.

Uncovering trends help you to:

• Plan ahead, spot new opportunities before anyone else.
• Find broken processes before they become a major problem.
• Respond quickly to outside events.
• Give your agents the right information when they need it.

This blogpost is divided into two parts. Part 1 is about the most common type: trends over time.

Trends over time

For trends over time, line charts are usually used as the visualization. In one blink of an eye, it is usually clear which time trends are happening, when they started exactly, how strong the trends are and whether they are still occurring or not.

Abrupt trends

Abrupt trends are easily detected by sudden spikes (or dip) in the line chart. They are usually caused by a single high impact event.

For abrupt trends, it is very important to discover the cause and respond as swiftly as possible. Ideally, each call center manager has easy access to a live speech analytics monitor, always looking for spikes and ready to respond.

Abrupt trends in call volume data

When your call volume suddenly spikes, the first thing you need to ensure, before mobilising everyone, is if the spike fits your usual call volume pattern. For example, in some businesses it’s completely normal that a volume spike occurs just before lunch break or just before the call center closes. Or always on the last day of the month, when you are in the tax business. But if the call volume spike does not fit the usual pattern, you can be pretty sure that something big is happening.

The next step is to identify what type of calls are mostly contributing to the increase (or decrease) in volume. By splitting the overall call volume into specific topics, using topic detection technology, it should become clear very soon what topic is mostly responsible. Is it a certain problem with your product or service? Or did something happen outside that you are unaware of at the moment? The more specific you define your topics, the faster you can find the root cause of volume spikes.

Abrupt trends in agent performance

It is more uncommon that agent performance changes abruptly. Agents can improve (and they can sometimes become worse), but tf this occurs, usually it is a more gradual change.

There are however some exceptions.
Abrupt spike in silence duration: this is usually caused by a sudden problem in one of the system’s the agents are using, which prevents them from helping their customers. Or maybe new agents have joined the team who don’t know yet how to answer certain questions.

Abrupt change in average call handling time: usually this correlates with changes in call volume. For instance, if there is an outage in your business’s service, call volume of people reporting this will spike. But since these types of calls are usually very short, the average call handling time suddenly dips.

Abrupt dip in customer satisfaction / sentiment: This is usually caused by either a sudden problem in the company’s service or product, or a sudden problem in the customer call center. When call volume spikes and customers are put on hold much longer than usual, this can also cause sentiment dips.

A good speech analytics or automated quality monitoring solution shows you what sudden changes in agent performance are relevant for you at each point in time.

Slow moving (gradual) trends

Gradual trends are trends that happen more slowly over time compared to abrupt trends. But the impact of these trends can be just as big. Usually dealing with gradual trends are more strategic than tactical and require more long term planning.

Call volume

When your call volume gradually increases (or even decreases) over time, you have to plan ahead as a call center manager. Sometimes the volume change can simply be explained by the growth of your number of customers. In that case, be prepared to scale up in time with the rest of your company.

When the number of clients in your business is stable but call volume is still increasing, something else is going on. Possible explanations of this are:

• Something is unclear about a new version of your business’s product or service, which leads to increasingly more questions.
• An external event triggers this (e.g. competitor actions, changes in customer behavior or the weather)
• The increase can also be due to more or less repeat calls, if the first time resolution changes over time (see Agent Performance).

Using topic detection, identify the topics that contribute the most to this trend. Word cloud analysis can help you get a list of terms from this topic that have grown significantly.

Agent performance

Slow moving trends not only happen in call volume but also in agent performance. By tracking many agent performance metrics, using automated QM, it is possible to quickly identify the metrics that contain these trends. Once a trend has been identified, it is important to narrow down where in your callcenter these trends occur. First, look for the agent teams that contribute most to these trends, especially the ones where performance drops. Once the right agent team is found, look for the individual agents to give them tailor-made coaching.

The trends to look for include:

• Increasing Average Call Handling Time: what parts of the conversation are longer than necessary for this agent?
• More negative sentiment: find negative sentiment examples
• Drop in First Call Resolution:
• Growth in Silence: look for the words surrounding the silence. Is there any trend that can explain the increase in silence?

Interested to learn more about other trend types? Stay tuned for Part 2!

Contexta360’s CTO, Tim Harbers, imagines how contact centres will evolve in years to come.

It’s a sunny day when I arrive at the office. After passing the usual security and health scan, I get to my desk and log in with a single voice command. So far so good, my unique voice signature is instantly encrypted, verified and approved. Meanwhile, one of our mobile coffee robots passes my desk to serve me my morning cappuccino just the way I like it.

The first thing I do is to open the automatic daily team report. Our team did really well yesterday, the estimated NPS was two points higher than average on a regular Tuesday. Last week’s team meeting and coaching session has paid off. The first-call-resolution metric, in particular, has improved dramatically. I’m interested in how we did this and decide to drill down into the details. The company has finally fixed the knowledge base bug apparently and, also, we have learnt to better work together with the automatic call assistant. Plus, the assistant’s AI engine has once again been upgraded to help with even more questions.

I look over my screen and see the large, live-monitoring dashboard of my team hanging on the wall. The trending call topic of today is obviously an issue with our company’s latest add-on that was released a week ago. Some 27 per cent of the calls have been about this subject lately and it is increasing every hour. To refresh my own memory, I click on the percentage to see the necessary details. A guideline and sample script are provided to help me learn what to say and how to resolve the issue over the phone.

Time to make the first call. I insert my earplugs and press the ready button on my system. Immediately the first customer profile and her question pops up on my display. Even though this customer has subscribed very recently, the IVR and attached customer prediction model have labeled her as high potential, so it makes sense to have a human handle this call. I greet the customer, while not worrying about staying close to my microphone. Current voice optimisation technology is perfectly capable of making the sound of my voice as clear as possible and cancelling out any distracting background noise. This gives me much more freedom.

The customer wants to subscribe to a new service and is responding to one of our latest campaign ads. She is, however, unable to find and subscribe to this service in her customer portal. Immediately my digital assistant, eavesdropping on the call, is trying to get my attention. Within a few seconds, after checking the customer profile, a suggestion in my terminal pops up. Apparently, the customer is still on a trial subscription and company policy forbids changing subscriptions during a trial period. The assistant lets me know that I am authorised to make her a custom offer to start a new paid subscription, with 10 per cent discount for this customer type. The customer hesitates and wants to know a few more details about the subscription. By checking the knowledge base and extracting the right information automatically, my assistant can also help me with that part.

Everything is now clear to the customer. When I am about to wish her a wonderful day and end the call, a warning message pops up on my screen. It is my automatic compliance checker reminding me I forgot to inform the customer about the changed payment terms of the new subscription. I quickly correct this mistake and the compliance checker is satisfied. After ending the call, I can leave the afterwork and the actual subscription to the system itself. The system generates an automatic summary and updates the customer’s CRM record with the new subscription information.

It is 11am and I am invited to a virtual coaching session. After logging into the virtual coaching environment, my personal report from last week shows me my improvements as well as my weaknesses. First the positive: my overall predicted NPS score has gone up for the third week in a row, this time by 8 per cent. My professionalism and patience are especially appreciated apparently. The report provides me with a couple of fragments from last month’s calls for me to listen to, each has a comment attached about how I could improve further. Most of them I still remember, like the one from last week where I almost lost my temper. But I’m surprised about a few of them. One call that I thought I handled very well didn’t actually solve the problem and the customer had to call again the next day. Some indicators of this that I missed are pointed out to me by my virtual coach.

Directly after my lunch break there is another call waiting for me. This time, the caller has specifically requested to talk to a human. I’m not looking forward to this as usually these are the difficult type of customers who are typically frustrated with our service and want to blow off some steam. However, because my performance on these types of calls is higher than average, the system prefers to give these calls to me in general (but not too many in a row as this doesn’t help with motivation).

The caller is requesting to change the login credentials for his subscription, but he is not able to pass the online checks. I go over the standard identity verification process. Immediately, the first alarm bell goes off. The automatic voice scanner is warning me that the caller’s voice fingerprint does not match what is on record for this subscription. Moreover, the acoustic emotion detection also indicates nervousness and talk speed faster than can be expected for this particular query. I am on my guard right away and want to make sure I am doing the verification process thoroughly and correctly. Fortunately, the automatic compliance assistant is there to keep an eye on me. Because the caller has neither any login credentials, nor access to the phone that we have on record, I can’t help him any further and politely invite him to one of our physical stores for a full identification.

Finally, the lines are closed, and my day ends with our weekly team briefing. It is a mostly informal meeting, where the conversational analysts share with us the latest trends and newest best practices that they have identified.

This might look like futuristc science fiction, but much of the technology that supports the agent is actually achievable today with Contexta360. Learn more here…