Procurement analysis case study

Summary

A sports venue operator had to choose between four international catering suppliers. The contract affects £15 to 18 million of annual revenue. Each proposal ran to more than 300 pages, and the board needed a decision within 14 days. The finance director used an AI-assisted process to compare the four proposals in 48 hours. A manual comparison would have taken 2 to 3 weeks. The analysis found a staffing assumption in one proposal that hid more than £200K a year in cost.

Results

Metric Manual AI-assisted Change
Analysis time 2 to 3 weeks 48 hours 95% less
Hidden cost found Not shown in the proposal £200K+ a year Found before negotiation
Output Summary notes 15-slide board presentation, 25-page financial analysis, footnoted sources Board-ready

The problem

Business context

Decision parameters:

  • Industry: sports venue operations
  • Contract value: £15 to 18 million annual revenue
  • Proposals: 4 international suppliers, 300 or more pages each
  • Deadline: executive decision within 14 days
  • Staff: one finance director, with other projects in progress

Document formats

The four proposals arrived as Excel workbooks with linked formulas, PDFs with embedded tables, and Word documents with mixed data. The suppliers used different cost categories, so there was no common baseline to compare them on.

Limits of a manual review

Problem Effect Risk
Time 40 or more hours of manual work Critical
Errors Calculation mistakes in complex models Critical
Consistency Each proposal read and judged on its own terms High
Context No industry benchmark to test the projections against High

Client comment

Client comment on the 48-hour analysis and the £200K+ hidden cost it found

Method

Three phases

Phase 1: context, 30 minutes

graph LR A[Market Research] --> B[Industry Benchmarks] B --> C[Competitive Analysis] C --> D[Context Foundation] style A fill:#e1f5fe style B fill:#f3e5f5 style C fill:#e8f5e8 style D fill:#fff3e0

Research gathered:

  • Stadium catering operations
  • Industry benchmarks for growth and food cost
  • Post-pandemic market conditions
  • The client's own objectives for the contract

Phase 2: document processing, 45 minutes

Processing by format:

Input Method Output
Excel files Formulas kept and recalculated Financial models on one template
PDF documents Table structure recognised Structured data
Word documents Financial figures parsed from prose Cost structures on one template
Each format Mapped to one set of cost categories Common comparison baseline

Phase 3: comparison, 60 minutes

Three lenses:

  1. Lens 1: base case

    • Return at the current activity level

  2. Lens 2: growth

    • Revenue under each supplier's expansion scenario

  3. Lens 3: risk

    • Costs not stated in the proposal
    • Contingency the client would need to hold

Checks and audit trail

  • Source attribution: each figure carries a reference to the document it came from
  • Audit trail: footnotes for each figure, so a reviewer can check it
  • Review: the finance director reviewed each draft and corrected assumptions
  • Iteration: the analysis ran again after each round of corrections

Deliverables

Decision documents

Board presentation

15 slides

Summary and recommendation
Side-by-side supplier comparison
3-year financial impact
Risks and how to handle them

Financial analysis

25 pages

Cost-benefit analysis
Revenue stream breakdowns
Labour cost structure comparison
Costs not stated in the proposals

Implementation plan

8 pages

Negotiation points
Due diligence checklist
Contract terms to change
Performance monitoring

Supporting documents

Document Purpose Pages Audience
Audit trail report Source reference for each figure 12 Compliance and legal
Assumption register Baseline assumptions and the reason for each 6 Finance team
Risk register Likelihood and impact of each risk 8 Executive committee
Action plan Next steps with dates and owners 4 Project management

Findings

Three anomalies

Anomaly 1: staffing assumption

  • One supplier proposed 40% fewer staff than the other three
  • Its model assumed the client would supply the sales staff
  • Effect: £200K or more a year in cost that the proposal did not show
graph LR A[Supplier Proposal] --> B[40% Lower Staffing] B --> C[Hidden Assumption] C --> D[£200K+ Hidden Cost] style A fill:#e3f2fd style B fill:#fff3e0 style C fill:#fef2f2 style D fill:#fef2f2

Anomaly 2: growth projections

Supplier Projected growth Industry benchmark Assessment
Supplier A 15% a year 5 to 7% Above benchmark
Supplier B 8% a year 5 to 7% Near benchmark
Supplier C 3% a year 5 to 7% Below benchmark
Supplier D 12% a year 5 to 7% Above benchmark

Anomaly 3: cost structures

  • Food cost ranged from 28% to 35% of revenue across the four proposals
  • Each supplier allocated labour cost in a different way
  • Ranking the suppliers required mapping the four proposals to the same cost categories

Recommendations

Ranking by lens

Base case
Best: Supplier B
Highest return at current activity

Growth
Best: Supplier A
Most credible expansion plan

Risk
Best: Supplier C
Most conservative projections

Negotiation points

  • One labour cost structure for each bidder
  • Defined revenue growth milestones
  • Performance penalty clauses
  • Consequences for service level failures

Aim:
Contract terms that close the three anomalies above

Timeline: manual and AI-assisted

Timeline comparing the 2 to 3 week manual process with the 48-hour AI-assisted analysis

Business impact

Direct results

  • Time: 40 or more hours of the finance director's time saved
  • Decision basis: four proposals compared on one set of figures
  • Cost: £200K or more a year in hidden cost found before negotiation
  • Board: one document set that finance, legal, and the executive committee worked from
  • Reuse: the framework applies to the client's next supplier review

Technical architecture

Processing stages

graph TD A[Context Engine] --> B[Document Intelligence] B --> C[Financial Modeling] C --> D[Report Generation] A --> E[Automated Research
Industry Analysis] B --> F[Multi-Format Extraction
Data Standardization] C --> G[Scenario Analysis
Sensitivity Testing] D --> H[Professional Documents
Audit Trails] style A fill:#e3f2fd style B fill:#f3e5f5 style C fill:#e8f5e8 style D fill:#fff3e0

Security and confidentiality

  • Anonymisation: supplier identities anonymised throughout
  • Audit: each output figure traces to a source document
  • Access: restricted by permission
  • Retention: stored with backup and recovery in place

Conclusion

Four 300-page proposals in three file formats were compared in 48 hours against a 14-day deadline. The comparison found a staffing assumption worth £200K or more a year that the proposal did not show, and it gave the board a 15-slide presentation with a source footnote on each figure. The finance director reviewed each draft. The framework is ready for the client's next supplier review.

made with bernard

Cookie settings