Frankwatching: Understanding context as the most important KPI

This article by Qaitbay was published on FrankWatching.

Average Handle Time (AHT), First Contact Resolution (FCR), and Customer Satisfaction (CSAT) have been crucial indicators in customer contact for years. However, these KPIs provide only a partial picture of the situation. While they measure the outcome of an interaction, they offer little insight into the quality of the process that precedes it.

The quality of customer interactions is becoming increasingly important now that artificial intelligence has become an integral part of customer service. Whether a question is answered by a human employee or an AI agent, the quality of the answer ultimately depends on one determining factor: does the person handling the question understand the full context?

Customer inquiries are becoming increasingly complex, and customers expect consistent answers regardless of the communication channel they use. To successfully achieve this, it is essential that all relevant information is available at the right time and is interpreted correctly. This means that the information must be current, accurate, complete, and measurable. The key word in this process is understanding the context.

Current KPIs are rearview mirrors

Most KPIs in customer contact are rearview mirrors:

• AHT tells how long a conversation lasted.
• FCR shows whether a question has been resolved immediately.
• CSAT measures how satisfied the customer was afterwards.

While these indicators are certainly valuable, they do not always answer the question of why an interaction was or was not successful. For example, an employee might have a high average handling time (AHT) because they had to search for the correct technical documentation for twenty minutes. Similarly, a low first contact resolution (FCR) can result from the fact that contract information is scattered across different systems.

Furthermore, low customer satisfaction is often the result of inconsistent responses from different departments. In other words, most KPIs primarily measure symptoms and not the underlying causes.

The real challenge lies before the solution

In many organizations, work often does not begin with answering a question, but with searching for and gathering information. Employees navigate between various platforms such as Customer Service Platform, CRM, SharePoint, email, document management systems, knowledge bases, manuals, and technical documents. Each application offers only a part of the whole.

This process not only takes a lot of time but also increases the risk of errors. The more often an employee has to switch between systems, the heavier the cognitive load becomes, and the greater the chance that important information is overlooked.

The same applies to AI. A language model can only reason with the information at its disposal. If part of the context is missing, the quality of the answer also suffers.

The quality of customer contact is increasingly determined by context

The National Voice Monitor 2026 already indicated that organizations are increasingly dealing with more complex customer inquiries, while at the same time expectations regarding speed and quality continue to rise.

Moreover, the monitor points to the increasing importance of AI, knowledge management, and improved information provision. Not as ends in themselves, but as necessary preconditions to effectively support both employees and AI.

This confirms a development within customer service: the quality of customer contact is increasingly determined less by the speed at which an agent can find information, and more by the completeness of the available context.

What do we mean by context understanding?

Contextual understanding refers to the ability to interpret information within the correct context. This goes beyond thoroughly understanding customer inquiries and having access to a knowledge base or CRM system.

This entails that all relevant information regarding a customer inquiry is automatically collected, analyzed, and interpreted. Which information is important in this regard depends on the specific organization and the nature of the customers. Examples include:

• Customer data and previous contact moments
• Open and closed tickets
• Emails and (internal) notes
• Contract agreements
• Product specifications, manuals, technical data
• Information on the website
• Legislation and regulations

The value depends not only on the quality of information (complete, accurate, up-to-date) but primarily on the interrelationships. A previous service report, for example, can explain why a part was replaced. A contract subsequently determines which solution may or may not be offered. Only when this information is reviewed together does sufficient context emerge to make a sound decision or take proper follow-up action.

Why traditional systems fall short

Many organizations already have modern CRM systems, ticketing applications, and knowledge bases. Yet, information often remains fragmented, and that is not surprising. These systems were developed to store information, not to compile context.

A CRM knows the customer, a ticketing system knows the interaction, SharePoint contains manuals, and the ERP system contains contracts. None of these systems has the primary task of automatically bringing together the full context during a customer conversation. As a result, the employee themselves remains the integration layer.

From knowledge management to context management

For years, CCAAS and case management solutions revolved around capturing customer interactions, and knowledge management revolved around capturing and retrieving information. However, the central question is no longer where information is recorded. The focus is now on how a ticket or case can be handled quickly and with high quality, based on the right information.

New concepts such as Case Intelligence and Case Resolution Management. These approaches do not focus on registering a case, but on analyzing, understanding, and preparing for its full handling. AI plays an important role in this, provided it has access to all relevant context.

Well-known providers such as Topdesk, Genesys, Zendesk, and Salesforce are attempting to take the next step in the case management value chain. However, they are limited because they prefer to keep users within their own ecosystem, which makes a full understanding of context impossible.

In contrast, new providers such as GetMosaic, Chatbase, IrisAgent, and the Dutch Qaitbay are vendor-independent and combine, for example, Genesys with SharePoint, SAP, ServiceNow, and their own DMS. These new solutions are not a replacement for established solutions, but rather an intelligent resolution layer on top of existing systems.

These new developments are drastically changing the role of employees. Instead of starting by searching for, gathering, and interpreting information, they now begin with a fully developed case file. In the case of case resolution, they even have preparatory follow-up actions ready. This enables the employee, or the AI ​​chatbot, to focus entirely on the interaction with the customer.

How do you measure context understanding?

When understanding context becomes so important, a new question naturally arises: how do you make it measurable? Numerous formulas have been developed for this. The foundation for AI was laid in 2017 by Google with the Scaled Dot-Product Attentionformula in the article "Attention Is All You Need".

Besides the fact that this cannot be calculated for an average organization, this model only reflects context understanding within a large language model (LLM). A formula that also considers data fragmentation and search time, and is more practically applicable, is the Context Efficiency & Quality Score (CEQS).

The variables:

1. Context Coverage (CC)

This indicates the extent to which the context, scope, and completeness of data have been taken into account in an analysis. It helps determine whether all relevant parts of a source or text have been used, including the circumstances in which they originated. In this way, the focus is not only on the content but also on the background necessary to correctly interpret the information.

2. Consistency Rate (CR)

How often do similar customer inquiries or cases lead to the same answer, assessment, or follow-up step, regardless of which employee handles the question? A high consistency rate indicates a uniform application of knowledge, processes, and policies. A low score may indicate that information is fragmented, differences in interpretation arise, or employees use different sources.

This KPI therefore says something not only about the quality of the service, but also about the reliability and accessibility of the available knowledge.

3. Search Time (t search)

How much time does an employee spend finding, gathering, and verifying the information needed to handle a customer inquiry or case? This includes searching knowledge bases, manuals, emails, SharePoint, CRM systems, and other applications, as well as assessing which information is current, complete, and reliable.

A high search time often indicates fragmented knowledge, poor findability, or insufficient context, causing delays in processing. This also contributes to determining the ART/AHT.

4. Systems (S)

How many different applications, knowledge sources, and data sources must an employee consult to fully handle a case? The more systems required, the greater the risk of time loss, context shifts, incomplete information, and errors. This KPI provides insight into the degree to which information is fragmented.

Contextual understanding or CEQS are not replacements for existing KPIs, but indicators that explain why performance improves or lags behind, leading to:

• less search time
• Lower Average Handle Time
• higher First Contact Resolution
• more consistent advice
• fewer escalations
• higher customer satisfaction
• less work pressure for employees

Understanding context, the next step in customer contact

The discussion regarding the use of AI in customer contact often focuses on topics such as agent assist, virtual assistants, and agentic AI. While these developments are of great importance, they can easily distract us from a more fundamental issue. The real competitive battle in the coming years will not revolve so much around who has the most advanced AI, but rather on who knows how to create the best context.

Therefore, get started now with the following concrete steps:

• Certain data quality.
  • the degree of data fragmentation (systems, versions, languages)
  • the number of duplicates and contradictions
  • information gaps (content gaps)
• Measure context understanding using CEQS.
• Let a specialist in case intelligence solutions inform you about the potential for improvement.

Perhaps it is time to shift our focus from the speed at which a case is handled to a more important question: does the employee or AI truly understand the situation before providing an answer?