Audit analytics on supplier payments: Bristol supplier spend review

This project analyzes public supplier payment data from Bristol City Council to build an audit analytics workflow around supplier spend, repeated payment patterns, and risk-based review.

The dataset includes three monthly files of payments over £500 and is especially useful for an accounting and audit-focused project because it contains fields such as supplier name, amount, payment date, transaction number, and spending category descriptions.

Source: data.gov.uk — Local authority spend over £500: Bristol City Council

Repository: github.com/FlaviaRossi/data-projects/tree/main/bristol-supplier-spend

Dataset scope

After cleaning malformed and out-of-period rows, the working dataset includes:

• 17,959 payment rows
• 3 months of data
• 1,900 suppliers
• 14,698 transaction numbers
• £189.5M in total analyzed spend

What the review focused on

The analysis was designed around a set of useful audit questions:

• Which suppliers concentrate the largest share of total spend?
• Which spending categories dominate the dataset?
• Which repeated payment patterns should be reviewed?
• Which transactions look unusual relative to each supplier’s normal behavior?
• Can an explainable risk score help prioritize review work?

1. Spend concentration by supplier

Supplier concentration is one of the clearest starting points in a review like this.

Key findings:

• the top 10 suppliers account for about 21.96% of total spend
• the top 25 suppliers account for about 35.25% of total spend

This is useful because concentration risk often matters even when individual transactions are valid. A small number of suppliers can drive a large share of exposure, operational dependence, or review effort.

2. Spend categories and program patterns

The description_1 and description_2 fields provide a workable view of spend categories, programs, or departmental activity.

Among the largest groups by amount were:

• New Construction
• Conversion, Renovation & Improvement
• TPP - Placement Residential
• TPP - Grants
• Services - Fees and Charges
• TPP - Placement Foster Care Agency

This helps frame the project not just as payment analysis, but as a category-level review of where resources are concentrated.

3. Repeated payment patterns

Repeated values are not evidence of error by themselves, but they are very useful review candidates.

The analysis found:

• 1,147 groups with the same supplier + amount + payment date
• 4,152 rows inside those repeated groups
• 925 groups with the same transaction number + supplier
• 4,186 rows inside those groups

In practice, these patterns may reflect:

• valid recurring payments
• split lines under one process
• payment batches
• or cases worth a closer look for duplicate or fragmented payment review

4. Supplier-relative outliers

Looking only at the largest payments in the full dataset is not enough. A more useful audit view compares each transaction to the normal pattern of that supplier.

This approach surfaced notable outliers in suppliers such as:

• Jeff Way Construction Ltd
• Maples
• Oasis Support Ltd
• Constellia Public Limited
• REDACTED

This is a strong example of why supplier-relative analytics can add more value than global ranking alone.

5. Review prioritization with a risk score

A rules-based score was built to create a review queue. The scoring considered flags such as:

• very large absolute amounts
• unusually large amounts relative to supplier history
• repeated supplier + amount + date combinations
• repeated transaction number + supplier combinations
• month-end timing
• redacted supplier names

This produced a small high-priority review set:

• 36 rows with risk score >= 60
• around 0.2% of the cleaned dataset

That is useful because it creates a manageable set of transactions for deeper review instead of applying the same attention to every row.

Why this project matters

This project is a good example of how data analytics can support audit and accounting work in a practical way. It does not depend on a complex black-box model. Instead, it combines:

• real public data
• clear cleaning logic
• supplier and category analysis
• repeated payment checks
• outlier detection
• explainable risk prioritization

That makes it well aligned with audit-oriented work where traceability, interpretability, and reviewability matter.