Following on from our recent blog on 5 questions to consider before purchasing a text analytics solution, we’re covering another topic that we are often asked about - what makes an effective text analytics solution?
Organizations typically first encounter text analytics as a part of an end-to-end Customer Experience (CX) platform. CX platforms incorporate basic text analytics software as a part of a much broader solution that covers customer engagement, surveying and communication.
But these end-to-end CX solutions come with one major disadvantage. Organizations are confined to using the text analytics component purely for analyzing customer experience data. If an organization is looking to apply text analytics to data from other parts of the business (for example, employee experience, L&D and operations), it is impossible to do so. A specialized text analytics solution must be sourced.
With that in mind, we have covered a number of considerations to help you choose the right comprehensive text analytics solution for your organization’s needs.
It is important to note that the most critical requirement, beyond the below, is clarity around targets and objectives of using text analytics. Knowing the amount and type of data you will analyze, how much you have to spend on the solution and how much time you want to put into setup and maintenance (that could be spent digging deeper into the ‘why’ behind your customer feedback!) will ultimately impact which solution you decide upon.
Level of accuracy and insights
Open-ended customer feedback could be some of the most valuable data held by your organization. Feedback on how your customers feel about your product or service provides valuable insight into what is working and what isn’t. Which, when analyzed properly gives you the exact information you need to improve CX and subsequently build loyalty, increase average customer LTV and grow referrals to new business.
Unfortunately, many analytics solutions currently on the market reduce this valuable data to a set of codes and analyze it as if it were any other type of structured data, causing organizations to miss valuable information and the opportunity to empathize with their customers.
The problem here is search vs discovery. Constrained by historical methods, analysts are searching for things that they already know they need to look for and are missing out on the emergent issues that their customers are providing feedback on. This puts organizations on the back foot, unable to accurately see and fix the problems customers are facing until they become a much bigger issue.
Recent advancements in text analytics solutions remove this issue by allowing analysts to discover which topics or issues are being written about and identify emerging trends in their data as they arise for customers.
Kapiche improves on this further by identifying influential concepts and then giving the much-needed context around these concepts to paint a complete picture of the customer's feedback. Kapiche does this via an AI algorithm that builds a unique language model, without input from the analyst. Without the need for human input, assumptions about what is important to customers and inherent human bias are removed, ensuring a much more accurate picture of the data.
Time to results
Before considering a text analytics solution, most organizations will try the manual approach. That is, manually reading each verbatim and hand categorizing it. Whether it’s done internally or outsourced to a 3rd party, this approach has a number of significant limitations. Firstly, the time to results is often measured in days or weeks, and a solution like this isn’t scalable.
Worse still, some solutions try to replicate this process but replace the manual reading with machine reading. Often, these systems still require an arduous setup process (from days to months) that has to be repeated for each different type of data analyzed, significantly impacting scalability and time to result. It almost defeats the purpose of using a technology solution in the first place.
Luckily, recent AI-powered text analytics address this problem using unsupervised machine learning. These types of solutions dramatically decrease the time to results. Organisations are able to analyze 1,000s of verbatim in seconds with 0 setup, as well as being able to go back and run analyses on historical datasets to track progress over time.
No matter which solution you choose, be sure to consider both how long will it take to start seeing results, as well as what effort over what period of time will be required to use this solution on a different data domain like internal employee surveys rather than customer surveys.
Easily consumed output
Text analytics software is often used by several people within the business - both directly and indirectly. The first, and most obvious, is the analyst who will use the tool to deep dive on the customer feedback data, pull out and understand key themes from in-depth analysis and subsequent visualizations to tell a story for the rest of the business.
However, having an analyst identify actionable insights from the text analytics software is, by itself, insufficient to inspire change throughout the rest of the business. Analysts must be able to share insights out in a way that is easy for the relevant people to digest and subsequently make the necessary changes to improve customer satisfaction.
Executives want to see a high-level overview of key issues facing customers, factors that are positively and negatively impacting customer opinion or NPS and potential changes that need to be made in order to improve customer satisfaction. As such, it is important that these insights can be shared with the executives and other key stakeholders across the business - including those who aren’t analytically minded - without having to dig through the data themselves.
The added benefit here is that the ability to share insights across the business builds cross-business buy-in - which we feel is one of the biggest contributing factors to whether an organization’s customer experience program will succeed.
Trial using your own data
The above requirements are vital to choosing the right solution for your business, but how can you know what each solution is all about without testing it first? Considering the sizable investment that some solutions require, having access to trial the tool prior to purchasing ensures the software is fit for purpose and will deliver ROI.
Current text analytics solutions almost always have a demo - and understandably so. They typically require large amounts of customization or set-up, so a demo with a sample dataset allows you to visualize how the program works and how actionable insights are discovered without having to wait for setup to take place to suit your requirements.
Some solutions will provide you with generic datasets and analyses predefined for a particular industry. Identifying the actionable insights specific to an industry dataset can be challenging when you don’t understand the industry, so look for opportunities to use your own data.
Ideally, to really give you an idea of the software’s capabilities, you should be able to trial the tool yourself. Kapiche chose to provide a full access trial as analyzing their own data and working their way around the tool is the best way to fully understand how it will (or won’t!) aid their business. Having access to a trial in this way should allow you to validate what you already know or suspected about your data, whilst also presenting you with new, actionable insights that you weren’t previously aware of.
Before you purchase any text analytics solutions, take the time to make sure it will achieve the goals your organization has set. There are many different platforms and software packages available that all have their pros and cons. Ending up with the wrong solution could mean additional money and time delays, preventing you from taking fast action to improve your organization’s customer experience.