Understanding NPS Movement Article

Two persistent challenges often impede organizations from fully harnessing the potential of NPS:

1. Volatility of NPS Scores: NPS can fluctuate unpredictably, often leaving businesses perplexed about the root causes of these changes.
2. Identifying Drivers Behind NPS Movements: Pinpointing the exact experiences or factors that influence the score proves difficult, as traditional analysis methods often fail to capture the nuances of customer feedback.

These challenges place organisations in a frustrating predicament: they have access to valuable customer feedback, but lack the tools to extract meaningful, actionable insights. Conventional methods of analysis frequently overlook the intricate connections between customer experiences and the dynamics of NPS shifts. This is where Aspect-Based Sentiment Analysis (ABSA) steps in as a game-changing solution.

The infamous NPS volatility

The Net Promoter Score (NPS) is known for its volatility, a phenomenon I have explored in several previous articles. While the statistical and mathematical challenges of NPS have been well-documented, this article will focus specifically on identifying and understanding the key drivers behind NPS fluctuations. If you are interested in the NPS volatility have a look at the following article.

Let’s focus on the second question and understand how ABSA can help us identify the drivers behind NPS movements.

ABSA and ASTE understanding the “WHY?”

Aspect-based sentiment analysis (ABSA) and aspect sentiment triplet extraction (ASTE) are two interrelated methodologies within the domain of sentiment analysis, aimed at understanding consumer opinions regarding specific aspects of products and services. ABSA focuses on classifying sentiments associated with distinct features, such as quality or price, thereby allowing businesses to identify strengths and weaknesses in their offerings based on consumer feedback. In contrast, ASTE seeks to extract precise triplets that detail the relationships between target aspects, sentiments, and opinion terms, enabling a more nuanced analysis of consumer sentiments that goes beyond simple categorization.

The significance of ABSA and ASTE lies in their application across various industries, including telecommunications, food and beverage, and media, where they provide valuable insights for market research, customer experience management, competitive analysis, product development, and sentiment monitoring. These techniques empower organizations to make data-driven decisions that enhance customer satisfaction and foster loyalty, by translating complex consumer feedback into actionable strategies

ABSA and ASTE can effectively analyze NPS volatility by breaking down customer feedback into specific aspects and sentiment patterns. ABSA can identify which product/service aspects consistently drive high or low NPS scores, while ASTE provides deeper context by extracting precise opinion triplets that reveal why scores change over time. For example, ABSA might show that “customer service” sentiment dropped before an NPS decline, while ASTE could pinpoint that “long wait times” and “unhelpful responses” were the specific drivers. This granular understanding helps organizations proactively address issues affecting NPS stability by targeting interventions at the most impactful aspects of the customer experience.

ABSA and ASTE are two complementary natural language processing methodologies that analyze customer sentiment at a granular level. Academic research shows how ABSA identifies and classifies sentiment for specific aspects of products or services, while ASTE extracts structured triplets (aspect, sentiment, opinion) from text to provide richer contextual understanding of customer opinions. Let’s go and deep dive into the two methodologies.

What is ABSA?

A Systematic Review of Aspect-based Sentiment Analysis: Domains, Methods, and Trends – Yan Cathy Hua et all – September 2024

Instead of just determining the overall sentiment of a comment, ABSA breaks it down into specific aspects and analyzes the sentiment associated with each. This allows businesses to gain a more nuanced understanding of customer opinions and make more informed decisions.

Aspect-Based Sentiment Analysis is a specialised branch of sentiment analysis that focuses on identifying specific sentiments toward distinct aspects of a product, service, or experience. ABSA goes beyond traditional sentiment analysis by examining customer feedback at a finer granularity. For example, a customer might leave a negative review saying, “The food was great, but the service was terrible.” Traditional sentiment analysis would simply classify this as a negative review. However, ABSA would be able to identify the positive sentiment towards the “food” aspect and the negative sentiment towards the “service”. This granular insight allows organizations to move beyond simple metrics like Net Promoter Score (NPS) and understand the “why” behind customer sentiment. NPS provides a general measure of customer loyalty but doesn’t explain the specific factors driving it.

By identifying the aspects that are driving positive and negative sentiment, organisations can take targeted actions to improve customer experience. For instance, in the example above, the business could focus on improving their service while maintaining the quality of their food.

The Evolution of Composite ABSA: From Simple Pairs to Rich Contextual Understanding

The field of Aspect-Based Sentiment Analysis has undergone a remarkable evolution, moving from basic sentiment detection to increasingly sophisticated information extraction frameworks. This progression reflects the growing need for deeper, more nuanced understanding of customer feedback. Let’s explore this evolution through its key developmental stages:

1. Aspect-Opinion Pair Extraction (AOPE)

Span-Based Pair-Wise Aspect and Opinion Term Joint Extraction with Contrastive Learning – J. Yang et all – October 2023

The foundational step in ABSA’s evolution, AOPE focuses on identifying co-occurring aspects and their associated opinions within text:

• Core Function: Extracts basic {aspect, opinion} pairs
• Example Analysis: Input: “The interface is intuitive and the pricing is reasonable” Extracted Pairs: {interface, intuitive}, {pricing, reasonable}
• Business Value: Enables basic feature-level feedback tracking Helps identify commonly discussed product/service aspects Supports initial categorisation of customer comments

2. Aspect-Sentiment Triplet Extraction (ASTE)

ASTE-Transformer: Modelling Dependencies in Aspect-Sentiment Triplet Extraction – I.Naglik and M. Lango, October 2024

ASTE represents a significant advancement by adding sentiment polarity to the extraction process:

• Enhanced Capability: Creates {aspect, opinion, sentiment} triplets
• Detailed Example: Input: “The app loads quickly but customer support responds slowly” Extracted Triplets: {app, quickly, positive} {customer support, slowly, negative}
• Additional Insights: Captures sentiment direction Enables sentiment trend analysis Facilitates more precise problem identification

3. Aspect Sentiment Quad Prediction (ASQP)

Aspect Sentiment Quad Prediction as Paraphrase Generation – W. Zhang et all – October 2021

Generative Aspect Sentiment Quad Prediction with Self-Inference Template – Y. Qin and S. Lv – July 2024

The most advanced form, ASQP provides a comprehensive view by adding categorical classification:

• Complex Extraction: Generates {aspect term, aspect category, sentiment polarity, opinion term} quadruples
• Rich Example Analysis: Input: “The monthly subscription fee is too expensive” Extracted Quad: {subscription fee, pricing, negative, expensive} Category Classification: Maps specific mentions to broader business categories
• Strategic Benefits: Enables hierarchical analysis of feedback Supports cross-functional improvement initiatives Facilitates systematic tracking of category-level trends

Can ABSA and ASTE explain NPS volatility?

Let me explain how ABSA and ASTE can help businesses understand and manage their NPS scores better, particularly when dealing with score fluctuations.

Think of NPS like your company’s temperature reading – when it fluctuates unexpectedly, you need to know why. Traditional NPS analysis might tell you that your score dropped from 45 to 30, but it often doesn’t explain the complete story behind this change.

This is where ABSA (Aspect-Based Sentiment Analysis) and ASTE (Aspect Sentiment Triple Extraction) come in. ABSA acts like a sophisticated diagnostic tool that breaks down customer feedback into specific aspects of your business – think product quality, customer service, pricing, or website usability. Instead of just seeing an overall decline, you can identify exactly which areas are causing concern.

For example, let’s say your NPS drops significantly in Q2. ABSA might reveal that while product quality sentiment remains stable, customer service sentiment has plummeted. This already gives you a clearer direction for investigation.

ASTE then takes this analysis deeper by extracting specific opinion triplets – essentially the “what-why-how” of customer feedback. It might show that within customer service, “response time – negative – too long” and “support staff – negative – uninformed” are the main drivers. This precise insight tells you not just that customer service is an issue, but exactly what aspects need attention.

When integrated with tools like Insight Narrator and sandsiv+, this analysis becomes even more powerful:

1. Real-time monitoring: The system constantly analyzes feedback across all channels, alerting you to potential issues before they significantly impact your NPS.
2. Pattern recognition: It can identify whether these issues are one-off incidents or part of a larger trend that needs strategic intervention.
3. Actionable insights: Instead of just flagging problems, the system can recommend specific actions based on the identified issues and their impact on NPS.

The real value comes in preventing NPS volatility. By understanding the specific drivers of customer satisfaction and dissatisfaction, you can take targeted action before small issues become major problems that impact your NPS score.

Transforming Customer Feedback into Comprehensive VoC Intelligence

Thanks to integration with Large Language Models and Transformer Architecture, these advanced analyses can now be performed directly within our sandsiv+ solution.

Think of it this way: ABSA and ASTE are like sophisticated diagnostic tools that tell you what’s wrong and why – similar to a doctor using advanced imaging to identify health issues. They’re excellent at pinpointing problems, showing you that maybe your NPS dropped because customers are frustrated with long wait times or confusing billing processes.

But identifying problems is only half the battle. This is where the Complaint-Risk Score (C-Risk Score) comes in, acting as the “treatment plan” that helps prioritize and address these issues effectively.

The C-Risk Score works by:

1. Impact Assessment

• It evaluates each identified complaint or issue based on its potential impact on customer churn
• Considers the “contagion effect” – how likely the issue is to spread to other customers
• Measures the financial implications of each problem

2. Priority Calculation

• Assigns a numerical score (typically 0-100) to each issue based on:Frequency of the complaintSeverity of impact on customer experienceHistorical correlation with customer churnCost of resolution versus cost of inactionTime sensitivity of the issue

3. Solution Effectiveness Prediction

• Analyzes historical data to predict which solutions are most likely to succeed
• Estimates the potential NPS improvement from different intervention strategies
• Calculates the return on investment for various solution approaches

For example, let’s say ABSA/ASTE identify three main issues:

• Slow customer service response times (sentiment: -0.8)
• Confusing billing statements (sentiment: -0.6)
• Website navigation problems (sentiment: -0.4)

The C-Risk Score might reveal that while the customer service issue has the most negative sentiment, the billing confusion actually has the highest risk score because:

• It affects more customers
• Has a stronger correlation with churn
• Creates recurring frustration
• Generates additional customer service costs

This analysis would suggest prioritizing the billing system improvements over hiring more customer service staff, even though service complaints are louder.

Practical Application Example

Let’s say you’re a telecommunications company:

1. ABSA shows negative sentiment around your mobile app
2. ASTE reveals specific issues like “app crashes during payment”
3. The C-Risk Score then:Calculates that this issue has an 85/100 risk scoreShows that customers experiencing app crashes are 3x more likely to churn predicts that fixing the payment system would improve NPS by 8 points Estimates the potential revenue saved from prevented churn suggests this should be prioritized over other improvements

Integration Benefits:

1. Resource Optimization

• Helps allocate budget and resources to solutions with the highest impact
• Prevents overinvestment in low-impact improvements
• Identifies quick wins versus long-term strategic needs

2. Proactive Risk Management

• Identifies potential issues before they become major problems
• Helps prevent NPS volatility through early intervention
• Creates a systematic approach to problem-solving

3. ROI Maximization

• Ensures investments in customer experience improvements deliver measurable returns
• Helps build business cases for major improvements
• Tracks the effectiveness of implemented solutions

1. Strategic Planning

• Provides data-driven insights for long-term improvement strategies
• Helps balance quick fixes with structural improvements
• Enables better forecasting of NPS trends

Coming Soon: Deep Dive into the C-Risk Score

Want to transform how you handle customer feedback and drive real NPS improvements? In my next article, I’ll unveil everything you need to know about the C-Risk Score – a groundbreaking approach to prioritizing and solving customer experience issues.

You’ll discover:

• A comprehensive guide to implementing C-Risk Score in sandsiv+
• Step-by-step integration with major platforms including Medallia, Qualtrics, InMoment and many others.
• How to leverage the Insight Narrator plugin for seamless implementation

Whether you’re using sandsiv+ or another leading CX platform, I’ll show you how to harness the power of C-Risk Score to drive meaningful improvements in your customer experience metrics.

Interested in seeing it in action? Drop a comment below to request a personalized demo of the C-Risk Score implementation. Let’s transform your customer feedback into actionable, prioritized solutions.

Stay tuned – this is one article you won’t want to miss.