test

  Bellomy's Text Analytics Tool The Competition
Approach We use an ensemble analytics approach that combines rules with NLP statistical matching and a powerful AI that we built in-house. Most competitors use either a purely rules-based or purely AI-based text analytics approach. For those that do use an ensemble approach, tools can be costly.  
Capabilities In addition to the essential features, we’ve developed additional capabilities and tools. Our tool can automatically augment client data with location-based data and weather data which enable unique analytics. We include a few tools not found in other similar offerings: word usage tool, date & time reporting, and geography reporting. Most of the tools did a decent job of one or two aspects of text analytics but either completely lacked or had minimal capabilities for other vital elements. For example, all did sentiment, but few did actionability; most offered emotion models of some sort, but few provided direct-impact calculations for NPS, CSAT, etc. Few tools integrated with the non-textual data beyond simple filtering.
Trained to analyze across multiple platforms Our tool streamlines the process of analyzing open-ended data with close-ended data all in one tool. Most systems are trained on social media, movie, store, and product reviews but not customer experience and survey responses.
Understanding sentiment Our approach to sentiment is human-based, not numeric-based. Instead of assigning a number, we apply positive, negative, mixed, and neutral sentiment assignments. We believe that emotion classification is preferable for understanding sentiment intensity and the nuances that consumers express. Results are explainable, and users can usually decipher what drove the system to classify text using the tool. Many models are numeric-based, assigning a sentiment score that is a number that can be hard to interpret.
Industry-specific models  We started by creating a utility-specific model, and now that industry-specific model is our starting point for others. These are included as part of our offering. With our pre-built models as starting points, the amount of work needed to begin using and extracting value improves.  Nobody was providing utility-specific models. Some competitors offer industry models but as an add-on or upgrade.
Accuracy We can tailor the tool for your needs and industry — grocery, utilities, financial services, and more. We manually update these different pipelines each day by reading through verbatims and refining the tool’s capabilities and accuracy. While some competitors claimed 95% accuracy, tests and real-world examples proved that to be very misleading. For example, one claimed 90%+ accuracy, which it may have offered, but it only classified 40-50% of the data we sent it, most of which it could not classify at all.)
Ease of use With simple training on Bellomy’s tool, users are empowered to take action. If they see a wrong classification, they can fix it. If they want to create new topics or adjust existing ones, they can do that in real-time. Lack of training, understanding, or capabilities within the tool may require users of other tools to reach out for assistance frequently. More capable tools had a hefty price tag and required hours of consulting work or professional services to leverage thoroughly.
Case management features Our tool is integrated with our case management solution to make it easy for users to make discoveries and take action. Significant findings can be acted upon immediately and assigned to others for work.   Only one system we viewed (at that time) integrated text analytics with case management. 
Reporting tools We included a showreel capability in our tool. Showreels are a unique feature intended to help researchers and others keep track of interesting, meaningful, or important comments and streamline the process of building report decks with them. None provided reporting generation tools like showreels. Most provided some sort of dashboarding; however, the dashboards were exported as images, which is an uncommon method for reporting on this type of data.