29 July 2019

Portfolio of work with Australia Post

Australia Post is a huge and hugely complicated orgaisation. They have amazing data (variety, volume, velicity, …) and to be honest, I was still getting my head aroud it after almost 2 1/2 years working there!

This post is a summary of the work I was involved with during by time with Australia Post.

StarTrack Regional Rates Dashboard

Provides transparency to regional managers over regional delivery agents’ rates
Tools: SAP HANA (SQL), Tableau

Chian of Responsibility Dashboard

Reporting heavy-vehicle safety compliance
Tools: Alteryx, SAP HANA (SQL), Tableau

StarTrack Subaccount Irregularity Analysis

Irregularity analysis of StarTrack subaccount charge data
Quantification of size of irregularities
Tools: Python

Parcel Locker Optimal Placement Analysis

Combined Post and publicly available data to assist decision-making to team charged with Parcel Locker rollout
Tools: SAP HANA, Python, Tableau

StarTrack Sales Opportunity Analysis

Dashboard and analysis for monitoring the progress of sales opportunities
Created feedback-loop and enabled process imporvement
Tools: SAP HANA, Tableau

Site Access and Leave Analysis

Analysis of irregularities in HR data
Tools: Python, Tableau

LPO Revenue Analysis

Analysis of LPO revenue, staffing and business model
Tools: Python

Transaction fraud detection

Random forest model developed to detect fraudulent online transactions
Results: Model had a recall of ~30% at a false-positive rate of 1:2000
Tools: Alteryx, R

Delivery Altering System (DAS) Dashboard

Reporting and monitoring DAS implementation
Strong feedback on requirements elicitation process
Tools: Tableau, SQL

GRC Automated Risk Documentation

Tool to turn data in risk register into risk summary PDFs for ERC and ARC reporting
Turned a very manual and onerous process by the risk managers into a fully automated process that allowed the risk managers to focus on the risk content
Tools: Python, WK HTML to PDF, Markdown (python library)

Injury Prevention Onboarding Process Analysis

Tested effectiveness of new on-boarding processes
Used bootstrapping to estimate measurement error in injury rates
Positive client feedback on data viz
Tool: Python

Happy Parcels

Development of dashboard to summarise parcel happiness (i.e. liklihood to experience delays &c.) for call centre agents
Goals: reduce AHT, reduce repeat calls, increase NPS
Tools: Google Cloud Platform, BigQuery, Tableau

Technical presentations:

Edward Tufte’s philosophy on data viz - I am a massive fan of Tufte’s work
Scatter-density plots - a method for visualizing correlations between continuous variables with both areas of low and high density (see implementation examples here)
Bootstrapping for parameter estimation