From massive data into actionable insights. That is the focus on these 3 example company cases I briefly describe. When you have tens of thousands of customer calls per month, how can you get your weekly top issues automatically fixed? How to know the biggest value demand topics from 100 000 chat discussions automatically? Or when you get thousands of open text feedbacks weekly, how AI could help you to see the major issues, so you can fix the issues and create revenue based on the insights? If you are having similar issues, read the 3 cases below to learn what could help your daily life and improve your business.
Insights from the content with AI/ML
First I will tell briefly about the technology and why it’s in the main role when trying to get insights from massive open text/voice customer data content.
With the latest pre-teaching and learning natural language libraries and models, the data amounts are no longer limiting to get real-time insights and understanding. Aiwo has been able to create models that will learn the topics from your customer’s talk and provide summaries of the top issues for you to fix. The raw data can be tens of thousands or millions of messages or discussions.
The latest technologies eliminate the language barriers, and support multilanguage data, and are still providing insights in a language you prefer. AI-run qualitative analysis provides not only the theme-level understanding of major issues but also the sentiment analysis allowing you to find the main issues that bother your customers or makes them happy. All this is possible to reach without any manual tagging or work from your side, only the data is needed!
In summary, AI (artificial intelligence) /ML (machine learning) techniques remove the limits of the data amounts. With Aiwo CX, that’s based on these techniques, you can:
- Listen to 10 000 calls a week and find the top issues why the customers are calling,
- Read every support email (in 10 different languages) and tell the main pain points that need to be fixed,
- Read every chat discussion (hundreds of thousands) and tell in real-time what customers are asking you today, this week, or this month
- Check every open feedback or NPS open text answer and find the areas to improve the experience of the 8 or less scored answers. This way you can improve the NSP with correct data-driven actions!
Let’s dig into 3 real-life cases covering the topic.
Example 1: Find and eliminate failure demand from the voice data
Customer Content (DATA)
- Phone calls (100’s per day)
Example Insight Found?
- Customers didn’t understand their bills
Outcome
- Less phone calls (Lower cost to serve)
- Better CES and CX for users
The first case concerns the problem of getting insights from support call center data. It’s impossible to listen and manually analyze thousands of calls every day. Doing the manual tagging from the discussed topics is manual work and a very unreliable way to react to the new arising issues. One employee might not realize the arising failure-demand issue, as it might only appear in few calls per day. But if the same issue occurs in calls received by 10 or 100 other colleagues then that might be already a critical issue.
Thus it’s extremely beneficial to get real-time insights from all of the calls, not only to eliminate failure-demand topics to lessen the cost of service (Consider what the cost of one call is to your support center?) but also to make happier customers when promoting your services based on what they wish from them.
Read more about failure-demand from our excellent article: Failure demand arises because we measure production efficiency and structure organizations based on industrial production logic, says the leading subject-matter expert in failure demand Hermanni Hyytiälä
In this case example, a large electricity company utilizes AI/ML technique analytics to find and eliminate failure demand based on the customer support call center data. They find out from the analyzed customer insight that there is a hidden but growing topic in the calls related to the change of the invoice. The customers are calling the call center to get a better understanding of the bills.
How this was done:
- Rest-API-interface to the call center
- Aiwo CX transcripts the speech to text with our partner’s Inscripta’s solution
- Aiwo CX qualitative analysis shows the top negative issues why customers are calling to all support lines, as well as the “silent arising topics”. These are usually the final moments to react to the issues before they become fatal!
- Top issues, biggest changes, and trends are seen based on the customers’ own narrative. Possibility to see these from real-time dashboards or automated weekly reports.
Finland’s national public broadcasting company’s Yle News wrote an article (in Finnish) on this specific case. Read the article from here: “Tekoäly nuuskii asiakaspalautetta ja kertoo siitä johtajille – paljasti sähköyhtiölle, että asiakkaat eivät ymmärrä laskujaan”
Example 2: Insights from massive NPS open text data
Customer Content (DATA)
- NPS feedback
Example Insight Found?
- What users actually gives the scores
- Website survey feedback includes the App comments
Outcome
- Multiple fixes for Digital services
NPS and feedback channels can be misleading. The second example shows that customers were giving low NPS scores, but not to the service where the feedback was asked, but due to the fact that other services did not have feedback forms. The qualitative analysis of the open text forms revealed this.
When you’re getting thousands of text replies, it’s too time-consuming and quickly too biased to try to manually find the issues related to given scores. Aiwo’s automatic theme analysis shows clearly answers to the question “WHY?” by analyzing the open feedbacks behind the numeric scores and giving actionable insights from the massive amount of data without any manual work.
Read more about NPS and its challenges from here: 5 Ways How AI Is Revolutionizing NPS Feedback Analysis
How this was done:
- All the NPS query data and feedback data is sent to Aiwo API
- Real-time analysis of the feedback by Aiwo CX
- Possibility to see directly any themes arising and how the NPS score changes based on the themes
- Themes showing the real issues and root causes for lower scores, pointing to correct places for improvements.
From customers own words:
We had a brand and website reform which brought us feedback from our customers through our website. At first, we thought that all the feedback is targeted to our new website. From Aiwo we were able to find out that there are lots of feedback on the website that is targeted towards our other channels. Through the themes and categorization in Aiwo, we can simultaneously see what works on the website and understand the customer’s voice as to which service or channel the feedback is directed to. – Elina Rasa, Customer Experience Manager, Fennia
Read more about this example reference case from this article: Fennia finds the root causes of customer feedback behind NPS numbers and open feedback
Example 3: From a deep open feedback analysis to a 6-figure profit
Customer Content (DATA)
- Open feedback from website
- Thousands per day!
Example Insight Found?
- Users hated “autoplay”
Outcome
- After A/B testing -> jump on usage
- Higher CX score
- Profit increase
- Less churn
Previously many B2C companies “hid” the open feedback channels from their digital services because simply it was impossible to manually read and react to all feedback. But the wisdom to improve your profit and customer satisfaction and experience lays exactly on those open text feedback. The customers will write about their experience, about issues you don’t even realize to ask or didn’t know existed. At the same time, they are also revealing in more detail the reasons for the NPS number they have given.
In this example case, our customer MTV utilizes the insights from massive feedback data after it has been analyzed by the Aiwo CX model. They have been done A/B testing after finding out an issue from the data that Aiwo had visualized for them. After A/B testing and the remedial actions, they could clearly see the effect of the change in their CX Score and more importantly, in the revenue and profit.
How this was done with Aiwo CX:
- All website feedback is transferred to Aiwo API
- Aiwo CX makes a real-time analysis of the feedback
- Constant actions are made based on the most impactful (to CX score) themes.
Description from this one example findings by MTV‘s Service Design lead Jan Rosnell:
We found out that the videos that started with autoplay seemed to irritate our customers for many reasons. We ended up testing it with AB testing and it provided us value. So, we implemented the changes based on the finding from Aiwo CX and it was an enormous success! Our CX Score increased tremendously, and we got a six-figure positive increase in profit! We are a mass service so even the smallest changes can turn into a big money.
Watch more about this from the video interview (Aiwo Studio) of our customer: Aiwo CX helps MTV achieve 6 figure increase in profit
Conclusion
It’s important to understand that you already have all the data to make CX-related improvements. You just need to change the data into insights. The data amounts are not anymore limiting the work, but actually making it possible to do the right actions based on the volumes and the sentiments of the feedback.
The wisdom for better CX-related actions is in every encounter with your customer! Meaning every chat message, phone call, email, and social media message from your customers. New technologies enable simple and customer-focused surveys, where one single numerical question (like NPS) gives you the score and open text answers include the WHYs behind the scores.
Summary
ML (machine learning) provides data-mining of the insights in real-time and your time is used to improve the experience based on the findings.
- No data limits when using AI/ML
- Strategic insights lay on the VoC (voice of the customer) content
- Finding and understanding failure demand from e.g. service phone calls and email channels
- Focusing on CX gives serious ROI with big customer masses