Service Fee Discrepancy Analysis

Our transaction ledger hides hundreds of small pricing slips that add up. The job is to surface those slips so the finance team can see exactly where service-level fees drift from policy and how much revenue is at stake.

You will start with raw exports that contain service IDs, transaction dates, listed fees, realised revenue and a few messy quirks—missing cells, outliers and, most of all, inconsistent fee records. Using Python (Pandas, NumPy) together with SQL for joins and aggregations, clean the data, quantify every charge, and spotlight the inconsistencies.

Deliverables
• Tidy dataset with documented cleaning steps
• Jupyter notebook (reproducible, well-commented) plus separate .sql files for any queries
• Statistical profile for each service: average fee, variance, revenue share and count of suspect transactions
• Flag on every row that breaks the expected fee range, plus a ranked list of the worst offenders
• Clear visualisations—histograms, box plots, heat maps—that drop straight into a finance deck
• Short narrative report outlining method, insights and optimisation suggestions

Acceptance criteria
• Code runs end-to-end on our sample without manual tweaks
• Inconsistency flagging shows at least 95 % accuracy on a spot-checked subset
• All figures reconcile with the summary statistics you provide

If pricing audits and data hygiene are your comfort zone, let’s turn our fee structure into a well-oiled revenue engine.