Wise SLA Decline Analysis

Prakhar Rathi · Lead Data Analyst Application

Overview

This website presents my analysis of the Wise Lead Data Analyst take-home case study. All the notebooks I have used are executed and rendered are here.

The objective is to identify the drivers behind a significant decline in SLA performance observed in the Wise Email support channel during Q2 2023 and provide evidence-based recommendations for operational improvement.

Recommended path: Data Overview → Volume Analysis → SLA Analysis → Friction Analysis → Statistical Evidence

What is this?

The analysis has been published as a Quarto website, allowing the investigation to be reviewed section-by-section while maintaining complete transparency into the underlying methodology, calculations, and statistical testing.

All charts, tables, metrics, and conclusions presented throughout this report are generated directly from the accompanying Python notebooks.

Business Context

It is early July 2023. SLA performance has declined from 83% before Week 11 to 70% after Week 14.

Two operational changes were introduced during this period:

Change Timing Scope
Self-Serve Automation Week 11 (early March) All cases
Skill-Based Routing Week 14 (late March) Internal teams only

Leadership needs to determine:

  1. What caused the SLA decline?
  2. Which operational change had the largest impact?
  3. Are the observed changes statistically significant?
  4. What actions should be taken next?

Analysis Roadmap

The investigation follows a structured approach, moving from descriptive analysis to causal evidence and recommendations.

1. Data Overview
   └── Data quality validation, schema review, and analysis setup

2. Volume Analysis
   └── Understanding changes in case inflow and workload composition

3. SLA Analysis
   └── Measuring performance shifts across channels, teams, and periods

4. AHT and ASA Analysis 
   └── Measuring the time spent on allocating and resolving cases and reviews

5. Friction Analysis
   └── Investigating operational inefficiencies, rework, escalations, and review behaviour

6. Statistical Evidence
   └── Validating observed effects using hypothesis testing and causal inference techniques

Each section concludes with a summary of findings and a transition to the next stage of the investigation.

Reproducibility

This report is fully reproducible.

The source data files provided as part of the Wise take-home exercise are intentionally excluded from the repository to respect data handling best practices.

Required Files

Just the keep the below required files in the same folder as the notebooks. 

├── ops_cases.csv
└── ops_reviews.csv

Notebook Downloads

For convenience, the source notebooks used to generate this report are available directly from this website. You can download the zip version of the notebooks.