Data Driven Decision Making
Expert-defined terms from the Certificate in Customer Service Analytics course at London School of International Business. Free to read, free to share, paired with a professional course.
A/B Testing – Related terms #
split testing, experimental design, control group. Explanation: A systematic method for comparing two versions of a customer‑service element (such as a script, dashboard layout, or response time policy) by randomly assigning interactions to each variant and measuring performance differences. Example: A call centre tests two greeting scripts to see which yields higher first‑call resolution. Practical application: Enables managers to base policy changes on statistically valid results rather than intuition, fostering continuous improvement. Challenges: Requires sufficient sample size, careful randomisation, and awareness of external factors that may bias outcomes, such as seasonal call volume spikes.
Analytics Dashboard – Related terms #
visualisation, key performance indicator (KPI), reporting tool. Explanation: A real‑time interface that aggregates metrics like average handle time, customer satisfaction (CSAT) scores, and net promoter score (NPS) into interactive charts and tables. Example: A supervisor monitors a live dashboard showing agent adherence and sentiment scores derived from text analytics. Practical application: Provides quick insight for operational decisions, such as reallocating staff during peak periods. Challenges: Over‑loading users with too many widgets, data latency, and ensuring that visualisations accurately reflect underlying calculations.
Attrition Rate – Related terms #
turnover, churn, employee exit rate. Explanation: The proportion of customer‑service staff who leave an organisation within a given timeframe, typically expressed as an annual percentage. Example: A contact centre reports a 15 % attrition rate over the past year, prompting analysis of training adequacy. Practical application: High attrition signals underlying issues—e.g., low morale or inadequate compensation—and informs retention strategies. Challenges: Isolating root causes from external labour‑market trends and linking staff turnover directly to service‑quality metrics.
Churn Prediction – Related terms #
propensity modelling, customer lifecycle, retention analytics. Explanation: The use of statistical or machine learning models to estimate the likelihood that a customer will discontinue using a product or service. Example: A telecom provider applies a logistic regression model to identify high‑risk accounts, then triggers proactive outreach. Practical application: Allows agents to prioritise at‑risk customers with targeted offers, improving revenue retention. Challenges: Data quality (e.g., missing interaction logs), model drift over time, and privacy considerations when handling personal data.
Customer Effort Score (CES) – Related terms #
satisfaction metric, post‑interaction survey, effort‑reduction. Explanation: A single‑question survey that asks customers to rate the ease of resolving their issue on a scale (often 1–5). Example: After a chat session, a customer rates the effort as “2 – low effort.” Practical application: Helps identify friction points in processes such as IVR menus or knowledge‑base navigation. Challenges: Survey fatigue, response bias, and aligning CES improvements with broader business goals.
Data Governance – Related terms #
data stewardship, compliance, data quality framework. Explanation: The set of policies, procedures, and standards that ensure data is accurate, consistent, secure, and used ethically across the organisation. Example: A retailer establishes a data‑governance council to approve data‑access requests for call‑record analytics. Practical application: Guarantees that analytical outputs are trustworthy and that regulatory requirements (e.g., GDPR) are met. Challenges: Balancing accessibility with security, achieving cross‑departmental buy‑in, and maintaining up‑to‑date documentation.
Data Lake – Related terms #
data warehouse, raw data repository, big data storage. Explanation: A centralised storage architecture that holds structured and unstructured data at scale, often in its native format, for later processing. Example: Voice recordings, chat transcripts, and CRM logs are ingested into a cloud‑based data lake for downstream sentiment analysis. Practical application: Enables flexible exploration of diverse data sources without costly ETL pipelines. Challenges: Preventing “data swamp” conditions, ensuring proper metadata tagging, and managing cost‑effective retrieval.
Data Mining – Related terms #
pattern discovery, clustering, association rules. Explanation: The process of extracting hidden patterns, correlations, or anomalies from large datasets using statistical or machine‑learning techniques. Example: Mining call‑detail records reveals a cluster of complaints related to billing errors during a promotional period. Practical application: Generates actionable insights for process redesign or targeted training. Challenges: Over‑fitting models, interpreting spurious patterns, and safeguarding sensitive customer information.
Data Quality – Related terms #
data cleansing, validation, accuracy. Explanation: The degree to which data correctly represents the real‑world entities it describes, measured by dimensions such as completeness, consistency, and timeliness. Example: Incomplete agent‑skill tags lead to misrouted tickets, prompting a data‑quality audit. Practical application: High‑quality data underpins reliable analytics, reducing the risk of erroneous decisions. Challenges: Continuous monitoring, handling legacy data, and reconciling disparate source systems.
Decision Tree – Related terms #
classification model, CART algorithm, predictive analytics. Explanation: A flow‑chart‑like model that splits data based on feature thresholds to predict an outcome, often visualised for interpretability. Example: A decision tree predicts whether a call will result in escalation based on variables like issue type, language, and prior sentiment. Practical application: Provides agents with rule‑based guidance for handling complex queries. Challenges: Sensitivity to noisy data, potential bias if training data is unbalanced, and difficulty scaling to high‑dimensional feature spaces.
Descriptive Analytics – Related terms #
reporting, diagnostic analytics, historical analysis. Explanation: The examination of past data to understand what happened, typically through dashboards, summary statistics, and trend lines. Example: Monthly reports show a 10 % rise in average handle time after a new software rollout. Practical application: Establishes a baseline for performance monitoring and identifies areas requiring deeper investigation. Challenges: Over‑reliance on past patterns without considering future shifts, and the temptation to present data without context.
Dimensional Modeling – Related terms #
star schema, fact table, OLAP. Explanation: A design technique for data warehouses that structures data into fact tables (numeric measurements) and dimension tables (contextual attributes) to optimise query performance. Example: A fact table stores call‑duration metrics; dimension tables include agent, product, and time. Practical application: Enables fast slicing and dicing of service‑level metrics for ad‑hoc analysis. Challenges: Maintaining referential integrity during schema changes and handling slowly changing dimensions.
Employee Net Promoter Score (eNPS) – Related terms #
internal engagement metric, employee advocacy, pulse survey. Explanation: A single‑question survey asking staff how likely they are to recommend their employer to a friend, scored on a 0–10 scale. Example: A contact centre records an eNPS of 30, signalling moderate employee satisfaction. Practical application: Provides early warning of morale issues that could affect service quality. Challenges: Low response rates, cultural differences influencing scoring, and translating eNPS insights into concrete actions.
Feature Engineering – Related terms #
variable creation, data transformation, preprocessing. Explanation: The process of constructing new variables from raw data to improve model performance, such as extracting sentiment polarity from text or calculating time‑since‑last interaction. Example: Adding a “time‑of‑day” feature to a churn model improves prediction accuracy by 5 %. Practical application: Enhances predictive power of classification models used in routing or retention. Challenges: Requires domain expertise, can introduce leakage if future information is inadvertently used, and may increase model complexity.
First‑Contact Resolution (FCR) – Related terms #
service efficiency, repeat call rate, resolution metric. Explanation: The proportion of customer inquiries resolved during the initial interaction without the need for follow‑up. Example: An FCR of 78 % indicates most customers leave the call satisfied. Practical application: Drives staffing decisions, informs training on problem‑solving techniques, and correlates strongly with CSAT. Challenges: Accurately measuring FCR across channels, balancing speed with completeness, and handling complex issues that legitimately require multiple steps.
Forecasting – Related terms #
time‑series analysis, demand planning, predictive modeling. Explanation: The statistical estimation of future values (e.g., call volume, staffing needs) based on historical patterns, seasonality, and external factors. Example: Using ARIMA models, a centre predicts a 12 % surge in inbound calls during a product launch week. Practical application: Supports proactive workforce scheduling and capacity planning. Challenges: Model drift due to unprecedented events (e.g., pandemics), data gaps, and the need for regular model recalibration.
Heat Map – Related terms #
visual analytics, density plot, performance matrix. Explanation: A colour‑coded graphical representation that highlights intensity or concentration of a variable across two dimensions (e.g., time of day vs. queue length). Example: A heat map shows peak wait times in the late‑afternoon slot, prompting schedule adjustments. Practical application: Quickly identifies bottlenecks and informs resource allocation. Challenges: Choosing appropriate colour scales to avoid misinterpretation and ensuring data granularity matches visual resolution.
Key Performance Indicator (KPI) – Related terms #
metric, target, balanced scorecard. Explanation: A quantifiable measure used to evaluate the success of an organisation, department, or individual against strategic objectives. Example: Average handle time (AHT) is a KPI for efficiency, while CSAT measures quality. Practical application: Aligns daily activities with corporate goals, enabling data‑driven performance management. Challenges: Selecting KPIs that are truly actionable, avoiding metric overload, and ensuring they are not gamed.
KPI Alignment – Related terms #
strategic mapping, cascading goals, performance linkage. Explanation: The process of linking individual or team metrics to higher‑level business objectives to ensure cohesive effort. Example: An agent’s resolution‑time KPI is tied to the centre’s overall service‑level agreement (SLA) compliance target. Practical application: Encourages behaviours that directly support organisational priorities. Challenges: Maintaining transparency across layers, preventing conflicting incentives, and updating alignment as strategies evolve.
Logistic Regression – Related terms #
classification algorithm, odds ratio, binary outcome. Explanation: A statistical model that predicts the probability of a binary event (e.g., churn vs. retention) by fitting a logistic function to input variables. Example: Predicting the likelihood that a support ticket will be escalated based on sentiment score and issue category. Practical application: Provides interpretable coefficients for decision‑makers to understand driver impact. Challenges: Assumes linear relationship in log‑odds, sensitive to multicollinearity, and may underperform with complex non‑linear patterns.
Machine Learning (ML) – Related terms #
artificial intelligence, supervised learning, model training. Explanation: A suite of algorithms that enable computers to learn patterns from data and make predictions or classifications without explicit programming. Example: An ML model clusters chat transcripts into topics for knowledge‑base improvement. Practical application: Automates routine analytics tasks, enhances routing decisions, and uncovers hidden insights. Challenges: Data bias, model interpretability, requirement for continuous monitoring, and resource‑intensive training cycles.
Natural Language Processing (NLP) – Related terms #
text analytics, sentiment analysis, entity extraction. Explanation: The computational techniques for analysing, understanding, and generating human language. Example: Using NLP to detect anger in voice transcripts, triggering an immediate supervisor alert. Practical application: Enables automated quality monitoring, real‑time sentiment dashboards, and chat‑bot enhancements. Challenges: Handling slang, multilingual support, and maintaining accuracy across evolving vocabularies.
Net Promoter Score (NPS) – Related terms #
loyalty metric, promoter‑detractor ratio, customer advocacy. Explanation: A single‑question survey asking customers how likely they are to recommend the service to others, scored from –100 to +100. Example: An NPS of 45 indicates a healthy level of promoter sentiment after a recent service redesign. Practical application: Serves as a leading indicator of growth, informs loyalty programmes, and highlights areas needing improvement. Challenges: Cultural response bias, low response rates, and over‑reliance on a single metric without context.
Operational Excellence – Related terms #
continuous improvement, Lean, Six Sigma. Explanation: A philosophy that seeks to optimise processes, reduce waste, and deliver consistent, high‑quality outcomes. Example: Applying Six Sigma DMAIC (Define‑Measure‑Analyse‑Improve‑Control) to reduce average handle time variance. Practical application: Drives systematic, data‑backed enhancements across the service operation. Challenges: Maintaining momentum, aligning cross‑functional teams, and measuring intangible benefits.
Outlier Detection – Related terms #
anomaly detection, statistical deviation, robust statistics. Explanation: Techniques for identifying observations that deviate markedly from the norm, often indicating errors, fraud, or emerging issues. Example: A sudden spike in call‑abandon rates during a specific hour signals a potential system outage. Practical application: Alerts managers to intervene quickly, preserving service quality. Challenges: Defining appropriate thresholds, avoiding false positives, and handling high‑dimensional data.
Predictive Analytics – Related terms #
forecasting, propensity modelling, risk scoring. Explanation: The use of statistical techniques and ML algorithms to anticipate future events based on historical data. Example: A model predicts the probability that a ticket will breach its SLA, prompting pre‑emptive escalation. Practical application: Enables proactive resource allocation and targeted interventions. Challenges: Model validity over time, data integration across silos, and communicating probabilistic outcomes to non‑technical stakeholders.
Process Mining – Related terms #
event log analysis, workflow discovery, conformance checking. Explanation: A method that extracts process models from system logs to visualise actual execution paths and compare them to designed processes. Example: Mining CRM logs reveals that 30 % of tickets bypass the recommended triage step. Practical application: Identifies process deviations, informs redesign, and supports compliance auditing. Challenges: Ensuring log completeness, handling privacy concerns, and interpreting complex process variants.
Quality Assurance (QA) – Related terms #
monitoring, scorecards, compliance testing. Explanation: Systematic activities that assess whether service interactions meet established standards and regulatory requirements. Example: Random sampling of recorded calls yields a QA scorecard rating of 92 % compliance with script adherence. Practical application: Provides feedback for coaching, supports certification, and reduces risk. Challenges: Balancing thoroughness with operational impact, avoiding reviewer bias, and scaling QA across multiple channels.
Root Cause Analysis (RCA) – Related terms #
fishbone diagram, 5 Whys, corrective action. Explanation: A structured approach to identifying the underlying reasons for a problem or defect. Example: An RCA reveals that high first‑call resolution failures stem from outdated knowledge‑base articles. Practical application: Guides targeted remediation, preventing recurrence of the issue. Challenges: Requires cross‑functional collaboration, can be time‑consuming, and may suffer from confirmation bias if not rigorously documented.
Sentiment Analysis – Related terms #
opinion mining, emotion detection, NLP. Explanation: The computational assessment of the emotional tone behind textual or spoken content, often categorised as positive, neutral, or negative. Example: Analyzing chat logs shows a surge in negative sentiment during a new product rollout, prompting immediate escalation. Practical application: Offers real‑time insight into customer mood, informing agent coaching and product improvements. Challenges: Sarcasm detection, language nuances, and ensuring model updates keep pace with evolving expressions.
Service Level Agreement (SLA) – Related terms #
contractual metric, response time, performance target. Explanation: A formal commitment that defines the expected level of service, such as maximum response time or resolution time. Example: An SLA stipulates a 30‑second average speed of answer for inbound calls. Practical application: Sets clear expectations for both provider and client, driving operational planning. Challenges: Negotiating realistic thresholds, monitoring compliance across fluctuating demand, and handling penalty clauses.
Statistical Significance – Related terms #
p‑value, confidence interval, hypothesis testing. Explanation: A determination that observed results are unlikely to have occurred by random chance, typically assessed at a 95 % confidence level. Example: An A/B test shows a 3 % increase in CSAT with p = 0.02, indicating statistical significance. Practical application: Validates that changes are genuinely impactful before scaling. Challenges: Misinterpretation of p‑values, multiple‑testing pitfalls, and ensuring adequate sample size.
Structured Query Language (SQL) – Related terms #
relational database, query, data extraction. Explanation: A programming language used to communicate with relational databases for retrieving, inserting, updating, and deleting data. Example: An analyst writes an SQL query to pull all tickets resolved in the last quarter with a CSAT below 3. Practical application: Enables ad‑hoc data retrieval, supporting rapid insight generation. Challenges: Managing complex joins, ensuring query optimisation for performance, and maintaining data security.
Text Analytics – Related terms #
NLP, content mining, unstructured data analysis. Explanation: The process of extracting meaningful information from textual sources such as emails, chat logs, or survey comments. Example: Topic modelling clusters recurring complaints about billing errors, informing process redesign. Practical application: Turns unstructured feedback into actionable categories for reporting. Challenges: Data cleaning, language diversity, and handling large volumes efficiently.
Time‑Series Analysis – Related terms #
trend decomposition, seasonality, ARIMA. Explanation: Techniques that examine data points collected sequentially over time to identify patterns, trends, and forecast future points. Example: Analyzing daily call volumes reveals a weekly seasonality peak on Mondays. Practical application: Drives staffing schedules, capacity planning, and budgeting. Challenges: Dealing with irregular intervals, outlier spikes, and model selection for non‑stationary series.
Touchpoint – Related terms #
interaction channel, customer journey, omni‑channel. Explanation: Any point of contact between a customer and the service organisation, including phone, email, chat, social media, or in‑person. Example: A touchpoint analysis maps the journey from initial web chat to post‑sale follow‑up email. Practical application: Identifies gaps or redundancies in the service flow, enabling optimisation of the overall experience. Challenges: Consolidating data across heterogeneous systems and attributing outcomes to specific touchpoints.
Value‑Added Services (VAS) – Related terms #
upselling, cross‑selling, premium support. Explanation: Additional offerings that enhance the core service, often delivered for a fee or as part of loyalty programmes. Example: Offering a dedicated account manager as a VAS to high‑value customers. Practical application: Increases revenue per customer and differentiates the service brand. Challenges: Accurately targeting customers who will benefit, avoiding perception of “nickel‑and‑diming,” and measuring ROI.
Voice of the Customer (VoC) – Related terms #
feedback loop, customer insight, survey data. Explanation: A systematic collection of customer preferences, expectations, and aversions, typically gathered through surveys, interviews, and social listening. Example: VoC analysis uncovers a recurring desire for faster self‑service options. Practical application: Guides product development, service redesign, and prioritisation of improvement initiatives. Challenges: Ensuring representative sampling, translating qualitative feedback into quantitative metrics, and acting on insights promptly.
Workforce Management (WFM) – Related terms #
staffing optimization, schedule adherence, forecasting. Explanation: The suite of processes and tools used to predict demand, schedule agents, and monitor performance against planned work. Example: A WFM system uses forecasted call volume to generate a shift roster that meets SLA targets. Practical application: Aligns staffing levels with predicted workload, reducing over‑staffing and under‑staffing costs. Challenges: Incorporating real‑time deviations, handling agent preferences, and integrating with payroll systems.
Zero‑Touch Automation – Related terms #
robotic process automation (RPA), self‑service, workflow orchestration. Explanation: Technology that enables transactions to be completed without human intervention, often through rule‑based bots or AI‑driven processes. Example: An RPA bot automatically updates a CRM record after a resolved ticket is closed. Practical application: Frees agents from repetitive tasks, improving efficiency and reducing error rates. Challenges: Identifying suitable processes, maintaining bot reliability, and managing change‑management resistance.