Nitrio: ML-powered Intent Extraction Platform
Nitrio uses its ML-powered intent extraction platform to analyze rep-to-lead messages and deliver advanced sales strategies
50% increase in daily throughput
5x reduction in manual operations
20% reduction in operational cost
Enhanced product performance
Nitrio required a modern and robust, ML-powered intent extraction platform, since:
- NLP platform used manual rules and heuristics-based models, which caused bottlenecks and inefficiencies, both from an infrastructure and team’s performance standpoint, and failed to accommodate the company’s growth
- The existing platform did not ensure the required level of accuracy of sentiment analysis of rep-to-lead messages, which resulted in a significant amount of messages outsourced to a third party for manual analysis
- The solution’s infrastructure demonstrated tight coupling between services, which increased their dependencies and negatively impacted the team performance, causing data quality and data consistency issues
Nitrio’s platform was aimed at efficiently analyzing inbound rep-to-lead messages to extract their intent and to collect useful data about every sales representative’s performance. The collected data was utilized to drive data-proven buy-in strategies for Nitrio’s clients.
The platform analyzed multiple different types of emails and inbound messages, which increased the demands for the accuracy of sentiment analysis. Unless the system could specify the message’s intent with 95% certainty, it was outsourced to a third-party team, which increased service costs. The reliance on manual processes created bottlenecks, caused scalability issues, and stunted Nitrio’s growth.
To elevate their product to a whole new level, Nitrio approached Provectus to design and build a new automated, ML-powered intent extraction platform for sales optimization.
Provectus designed and built a machine learning platform for highly accurate intent extraction.
The accuracy of analysis of inbound rep-to-lead correspondence and inbound messages is ensured by augmenting manually developed regular expressions. Over 4K regular expressions have been replaced by the single model, preserving the same F1.
Development and maintenance of regular expressions have been replaced by crowdsourced data annotation and Active Learning workflow.
The ML platform utilizes advanced neural networks coupled with natural language processing. It is developed using Tensorflow. Data annotation, data training, and data evaluation tasks run in the deep neural network have been automated. The platform is located on a separate EC2 instance, and it works based on hydrosphere.io.
The messages that land in or sent from the ML platform are managed with Amazon SQS. This allowed Nitrio to eliminate the complexity and overhead while dealing with in and out messages.
Continuous monitoring was built with Amazon CloudWatch, ensuring that Nitrio’s team has access to all the required logs, metrics, and events.
Nitrio received an ML-powered intent extraction platform that allowed the team to increase daily throughput by 50%, cut the amount of manual work by 5x, and optimize operational cost by cutting it up to 20%.
The accuracy of intent analysis increased, which allowed Nitrio to deliver higher quality buy-in strategies for their clients while providing sales representatives with advice. Thus, the company’s potential to onboard enterprise clients has increased as well.
The new machine learning platform allowed to mitigate scalability issues, since the team no longer had to heavily rely on third-party sentiment analysis companies. To top it off, structural bottlenecks were eliminated.
Due to the ML platform’s accuracy, sales representatives managed to focus more on communicating with leads rather than processing dozens of emails manually to find out if a prospect was interested in a company’s product. This improved Nitrio’s outlook and allowed to drive more clients.
Looking to explore the solution? Contact Us!