Generation of Statistical, Agent Quality and Product or Service Insights

Step 1: Gathering Data

The process of generating insights begins with gathering relevant data for each Insight Group. This data is tailored based on user preferences, such as region, product group, or predefined issue sets within insight groups. Filtered cases, along with selected metrics and dimensions, are then utilized for generating insights. This targeted approach ensures that the insights align closely with the user's specified parameters, enabling a more customized and relevant analysis of the data.

​

Step 2: Granular Case Level Analysis of Conversations

The Insights Hub employs Sprinklr's in-house fine-tuned large language models for in-depth analysis of conversations at a granular level. During this process, the model aims to extract detailed information from each case, including the specific issues and their duration. By thoroughly processing each case within its scope, the model provides comprehensive insights into the root causes of underlying problems and offers suggestions for improvement.

​

Step 3: Identifying Root Cause Patterns

Following the detailed analysis of individual cases, the next step involves identifying patterns in the root causes observed. This entails aggregating root causes when commonalities are identified, indicating shared issues across multiple cases. The aggregation is facilitated by Sprinklr AI clustering models fine-tuned specifically for this purpose.

​

Step 4: Generating Insights

The grouped root causes, identified as similar, are further categorized to form insights. Each insight undergoes statistical analysis to reveal patterns in associated metadata. Subsequently, these grouped root causes or insights are summarized to derive an Insight Title, Impact Score, and other relevant information.

​Generation of Statistical Insights

Step 1: Gathering Data

The process of generating insights begins with gathering relevant data for each Insight Group. This data is tailored based on user preferences, such as region, product group, or predefined issue sets within insight groups. Filtered cases, along with selected metrics and dimensions, are then utilized for generating insights. This targeted approach ensures that the insights align closely with the user's specified parameters, enabling a more customized and relevant analysis of the data.

​

Step 2: Statistical Analysis of Associated Metadata

Each case comes with various associated fields and metrics, including channel, CSAT score, case duration, region, and the product under discussion. Users have the flexibility to select their preferred metrics and fields when configuring an Insight Group. As the Insights Hub collects relevant data for each insight group, it performs statistical analysis on the amalgamation of these fields, collectively known as associated metadata. This analysis involves examining patterns within the metadata, such as the case count for a specific product experiencing a particular issue or the case count for a specific issue in a particular region. The goal of this statistical analysis is to discern meaningful patterns and insights within the dataset.

​

Step 3: Observing Metadata Patterns

The Sprinklr Statistical Analysis employs an industry-standard approach to uncover noteworthy patterns in the metrics and dimensions of interest within the associated metadata. Cases demonstrating a pattern in metadata undergo detailed analysis to uncover the root causes underlying these patterns. This approach establishes a connection between specific metadata patterns to underlying root causes and actionable suggestions.

​

Observed Patterns
Anomaly: An "anomaly" denotes a deviation from the normal or expected pattern, presenting something that stands out as unusual or different. If any metric for a specific field/dimension shows an anomalous change compared to its usual behavior, it is identified as an anomaly.

Trend: A "trend" signifies the general direction or tendency in which something is evolving or changing over time. It reflects a prevailing pattern or movement that indicates a consistent shift or movement in a particular direction. If any metric for a specific field/dimension exhibits an increasing or decreasing pattern in a given timeframe, it is recognized as a trend.

​

Step 4: Insight Generation

Each identified metadata pattern is summarized into an insight, associated with an impact score, aggregated root causes and suggestions.