Case Studies

Business problems. Practical solutions.

Explore how I’ve used data and business analysis to investigate problems, improve processes and support better decisions. Interact with the dashboards and process maps to explore the work.

Interactive visuals use representative sample data to protect confidential information.

Case 01Finance · Procurement

Investigating missed supplier credits

Differences between supplier invoices and agreed contract terms meant money owed to the business was being missed. I investigated those discrepancies and helped introduce a more consistent validation process.

Tools: SQL · Excel · Stakeholder workshops

What I did

  • 01Used SQL to reconcile supplier invoices against contracts, rebate percentages, buy prices and costs.
  • 02Investigated discrepancies with Finance, Vendor Management and Customer Success.
  • 03Supported the resolution of outstanding amounts and cash collection.
  • 04Introduced invoice validation checks and an audit trail for follow-up.

Result

Supported the recovery of approximately $3 million in supplier credits within one year.

Supplier invoice reconciliation

Illustrative reconstruction using representative sample data.

Invoiced (sample)

$425,566

Per contract

$413,702

Total variance

$11,864

Invoiced above contract terms

Outstanding

$3,012

Flagged + under review

Recovered (sample)

$8,852

16 exception invoices

Invoiced vs contract value by supplier (sample, USD)

  • Discrepancy
  • Invoiced
  • Per contract
NorthlineJuniperSouthportCascadeElmfieldHarbor$0k$30k$60k$90k$120k

Exception invoices — invoices above the contracted terms, with their current status.

RecoveredUnder reviewFlagged / unresolvedCleared
InvoiceSupplierMonthInvoicedPer contractVarianceTypeStatus
INV-2402Northline Supplies2024-08$19,123.50$18,300.00+$823.50Rebate not appliedRecovered
INV-2403Northline Supplies2024-09$4,389.00$4,200.00+$189.00Rebate not appliedUnder review
INV-2405Juniper Packaging2024-11$27,192.00$26,400.00+$792.00Price mismatchRecovered
INV-2406Juniper Packaging2024-12$4,389.00$4,200.00+$189.00Rebate not appliedUnder review
INV-2408Juniper Packaging2025-02$8,151.00$7,800.00+$351.00Rebate not appliedRecovered
INV-2409Southport Timber2025-03$26,928.00$26,400.00+$528.00Quantity errorUnder review
INV-2411Southport Timber2025-05$4,326.00$4,200.00+$126.00Price mismatchRecovered
INV-2412Southport Timber2025-06$40,755.00$39,000.00+$1,755.00Rebate not appliedRecovered
INV-2414Cascade Ceramics2024-08$4,326.00$4,200.00+$126.00Price mismatchRecovered
INV-2415Cascade Ceramics2024-09$4,326.00$4,200.00+$126.00Price mismatchRecovered
INV-2417Elmfield Glass2024-11$27,588.00$26,400.00+$1,188.00Rebate not appliedRecovered
INV-2418Elmfield Glass2024-12$27,984.00$26,400.00+$1,584.00Missing creditRecovered
INV-2420Elmfield Glass2025-02$8,151.00$7,800.00+$351.00Rebate not appliedFlagged
INV-2421Harbor Freight Co2025-03$27,192.00$26,400.00+$792.00Price mismatchRecovered
INV-2423Harbor Freight Co2025-05$40,755.00$39,000.00+$1,755.00Rebate not appliedFlagged
INV-2424Harbor Freight Co2025-06$27,588.00$26,400.00+$1,188.00Rebate not appliedRecovered

Case 02Retail · E-commerce

Understanding why customers returned products

Returns were rising, but the reasons were spread across order, product and returns data that nobody looked at together. The teams who could act on it had no clear view of which products, categories or reasons were driving the problem.

Tools: Power BI · SQL · Excel · Data modelling

What I did

  • 01Cleaned and joined returns data with order, product and reason data.
  • 02Analysed returns by product, category, reason and channel to find the real drivers.
  • 03Built Power BI reporting tracking return rates and reasons over time.
  • 04Shared findings with the teams responsible so actions targeted the biggest causes.

Result

My Power BI reporting helped reduce customer returns by 75%.

Customer returns analysis

Illustrative reconstruction using representative sample data.

Units delivered

153,565

Sample units

Units returned

8,601

Return rate

5.60%

Returned ÷ delivered

Top reason (problem area)

Wrong size

34% of returns

Return rate trend by month

24-0724-0824-0924-1024-1124-1225-0125-0225-0325-0425-0525-060%3%6%9%12%
Return rate (lower is better)

Returned units by reason

0750150022503000Wrong sizeChangedmindFaultyNot asdescribedLate delivery
Returned unitsLargest reason — problem area

Return rate by category (%)

HomeElectronicsApparelBeauty0%0.45%0.9%1.35%1.8%
Return rate by categoryHighest rate — problem area

Units delivered and returned, filtered by period, category and reason.

MonthCategoryReasonReturned unitsDelivered unitsShare of month-category returns
2024-07Home & LivingWrong size ordered6832522.1%
2024-07Home & LivingChanged mind4432521.4%
2024-07Home & LivingFaulty on arrival3632521.1%
2024-07Home & LivingNot as described2832520.9%
2024-07Home & LivingLate delivery2432520.7%
2024-07ElectronicsWrong size ordered11136923.0%
2024-07ElectronicsChanged mind7236922.0%
2024-07ElectronicsFaulty on arrival5936921.6%
2024-07ElectronicsNot as described4536921.2%
2024-07ElectronicsLate delivery3936921.1%
2024-07ApparelWrong size ordered18045074.0%
2024-07ApparelChanged mind11645072.6%
2024-07ApparelFaulty on arrival9545072.1%
2024-07ApparelNot as described7445071.6%
2024-07ApparelLate delivery6345071.4%
2024-07BeautyWrong size ordered4532501.4%
2024-07BeautyChanged mind2932500.9%
2024-07BeautyFaulty on arrival2332500.7%
2024-07BeautyNot as described1832500.6%
2024-07BeautyLate delivery1532500.5%
2024-08Home & LivingWrong size ordered6934802.0%
2024-08Home & LivingChanged mind4434801.3%
2024-08Home & LivingFaulty on arrival3634801.0%
2024-08Home & LivingNot as described2834800.8%
2024-08Home & LivingLate delivery2434800.7%
2024-08ElectronicsWrong size ordered10838052.8%
2024-08ElectronicsChanged mind7038051.8%
2024-08ElectronicsFaulty on arrival5738051.5%
2024-08ElectronicsNot as described4438051.2%
2024-08ElectronicsLate delivery3838051.0%
2024-08ApparelWrong size ordered12733773.8%
2024-08ApparelChanged mind8233772.4%
2024-08ApparelFaulty on arrival6733772.0%
2024-08ApparelNot as described5233771.5%
2024-08ApparelLate delivery4533771.3%
2024-08BeautyWrong size ordered2821561.3%
2024-08BeautyChanged mind1821560.8%
2024-08BeautyFaulty on arrival1421560.6%
2024-08BeautyNot as described1121560.5%
2024-08BeautyLate delivery921560.4%
2024-09Home & LivingWrong size ordered6534961.9%
2024-09Home & LivingChanged mind4234961.2%
2024-09Home & LivingFaulty on arrival3434961.0%
2024-09Home & LivingNot as described2634960.7%
2024-09Home & LivingLate delivery2334960.7%
2024-09ElectronicsWrong size ordered8933482.7%
2024-09ElectronicsChanged mind5833481.7%
2024-09ElectronicsFaulty on arrival4733481.4%
2024-09ElectronicsNot as described3633481.1%
2024-09ElectronicsLate delivery3133480.9%
2024-09ApparelWrong size ordered18451793.6%
2024-09ApparelChanged mind11951792.3%
2024-09ApparelFaulty on arrival9751791.9%
2024-09ApparelNot as described7551791.4%
2024-09ApparelLate delivery6551791.3%
2024-09BeautyWrong size ordered4839141.2%
2024-09BeautyChanged mind3139140.8%
2024-09BeautyFaulty on arrival2539140.6%
2024-09BeautyNot as described1939140.5%
2024-09BeautyLate delivery1739140.4%

Case 03Operations · Reporting

Replacing manual reporting with automated workflows

Recurring reports were rebuilt by hand every cycle — downloading exports, copying, cleaning and re-checking totals. The work was slow, easy to get wrong and depended on whoever remembered to do it.

Tools: PowerShell · Excel · AI-assisted workflows

What I did

  • 01Mapped the manual steps in each reporting workflow and assessed what could safely be automated.
  • 02Automated data collection, refresh and distribution using PowerShell.
  • 03Used AI-assisted workflows to speed up building and validating the automation.
  • 04Validated outputs against the manual versions before switching over, and documented the new process.

Result

Recurring manual reporting was replaced with validated, documented automated workflows. No time-saving figure is claimed.

Reporting automation workflow

Illustrative reconstruction using representative sample data. No time savings are claimed — only the workflow is shown.

4 manual steps replacedControl retained: exception queue

Previous manual process (replaced)

Download exportsCopy & pasteClean by handRebuild tablesRe-check totalsEmail the team

Inputs

Processing

Validation

Outputs

Automated step (new)Control retainedPrevious manual processSelected step

PowerShell: collect & merge

A scheduled PowerShell script collects the source files, merges them and standardises columns and formats. Replaces manual download, copy-paste and cleaning steps.

Click any step to see what it does, which manual steps it replaces and which controls are kept. In my experience this kind of automation removes repetitive manual steps, reduces copy-paste errors and frees time for the analysis itself.

Case 04Supply Chain · Sales

Making supplier delivery problems visible

Late deliveries were affecting sales, but purchase, delivery and sales data sat in separate places. Without a shared view, supplier follow-up relied on anecdote rather than evidence.

Tools: SQL · Power BI · Excel

What I did

  • 01Consolidated purchase, delivery and sales data into one model.
  • 02Measured delivery performance and late orders by supplier, segment and period.
  • 03Connected supplier delivery issues to the sales they affected.
  • 04Delivered reporting that gave teams a shared evidence base for supplier follow-up.

Result

Identified delivery issues and improved visibility for supplier follow-up.

Supply chain & sales performance

Illustrative reconstruction using representative sample data.

Orders

12,805

Sample orders

Late deliveries

1,295

On-time delivery

89.9%

Below 95% target

Sales

€6,602,387

Sample revenue

Margin

27.3%

Revenue-weighted

Sales and margin trend by month

  • Margin % (dashed)
  • Sales (€)
24-0724-0824-0924-1024-1124-1225-0125-0225-0325-0425-0525-06€0k€200k€400k€600k€800k0%8%16%24%32%

Late orders by supplier

0150300450600NorthlineJuniperSouthportCascadeElmfieldMeridian
Late ordersMost late ordersOn time ≥ 95% (favourable)

Supplier detail (filtered)

SupplierOrdersLateOn-timeSalesMargin
Northline224540582%€1,178,65927.6%
Juniper225433585.1%€1,197,47227.1%
Southport219918591.6%€1,137,93527.9%
Cascade208815992.4%€1,059,45427.1%
Elmfield209913293.7%€1,038,77226.5%
Meridian19207995.9%€990,09527.4%

Delivery performance against sales and margin, filtered by period, supplier and segment.

Case 05Finance · Process Improvement

Designing the invoice validation process and audit trail

Finding discrepancies was only half the problem — without a defined process, flagged invoices could still be paid and outstanding amounts forgotten. The business needed clear checks, owners and follow-up.

Tools: Process mapping · Requirements gathering · Stakeholder discussions · SQL · Excel

What I did

  • 01Agreed the validation checks with Finance, Vendor Management and Customer Success.
  • 02Defined responsibilities: who validates at intake, who reviews exceptions, who approves recoveries.
  • 03Designed exception handling so flagged invoices enter a queue with a status and an owner.
  • 04Implemented the audit trail recording each check, finding, status and outstanding amount.

Result

Invoice validation checks and an audit trail were implemented, identifying discrepancies and tracking outstanding amounts to resolution.

Invoice validation & audit trail

Illustrative reconstruction based on implemented work. AS-IS shows the process before the change; TO-BE shows what was introduced.

Current process stepBottleneckException pathProposed improvement / control
Procurement
Finance
Vendor Management

Click any step in the map to see what happens, who is responsible and where the bottlenecks and exception paths are.

Want reporting like this for your business?

Tell me what you need — you’ll get a clear scope and a fixed price before anything begins.

Contact

Let’s talk.

A short description of your situation is enough. You'll get a considered reply with a clear scope and a fixed price before anything begins.

Your enquiry is sent only to me and is never published.