Skip to content
Vega Data

Case study · Retail and e-commerce · 2026

Pay-as-you-save programme

Saving EGP 15.9M a year for a fast-growing beauty brand

A pay-as-you-save data and AI programme for an Egyptian beauty brand with EGP 109M in annual revenue: a full leakage audit, an AI agent that reconciles storefront, ERP and courier data every day, and EGP 15.9M a year in savings the finance team can check line by line.

Waterfall chart of annual savings by lever, adding up to EGP 15.9 million
The seven levers behind EGP 15.9M a year.
saved a year, 14.6% of the brand's EGP 109M revenue
EGP 15.9M
Measured in delivery or live datasaved a year, 14.6% of the brand's EGP 109M revenue
online orders a year audited end to end
44,000
Measured in delivery or live dataonline orders a year audited end to end
gap between gross and net online sales traced to its causes
29%
Measured in delivery or live datagap between gross and net online sales traced to its causes
in orders missing from the ERP, found by the AI agent in its first 48 hours
EGP 265k
Measured in delivery or live datain orders missing from the ERP, found by the AI agent in its first 48 hours
  • Measured in delivery or live dataMeasured in delivery or live data
  • Measured in prototype testsMeasured in prototype tests
  • Modelled estimateModelled estimate

The starting point

The brand turns over about EGP 109M a year through its own online store, five retail branches and business buyers. Online, about 85% of orders are paid cash on delivery and two-thirds go to Cairo and Giza. Orders flow from the storefront to the ERP and out through two couriers, and the three systems did not agree with each other.

Online discounts had grown from 8.2% of gross sales in 2024 to 19.7%, cancellations and returns removed another 9.2%, and the ledger showed cash differences nobody could explain. The owners could see margin leaking but not where.

What we did

  • Fixed the data first. Repaired the storefront-to-ERP sync, added a nightly sync, validate and repair job, and removed 44 duplicate SKUs and 16 duplicate barcodes.
  • Put an AI agent on reconciliation. A validation agent built on the Model Context Protocol compares storefront, ERP and courier records every day and writes a plain-language exception report. In its first two days it found 365 orders worth EGP 265,000 that had never reached the ERP.
  • Reconciled every cash-on-delivery payment. Courier remittances are matched line by line against the ERP, so cash differences are explained the day they happen.
  • Routed deliveries by zone. An address-keyword zoning model sends each order to the cheaper courier for its zone and prints barcoded labels through the courier's API.
  • Tightened stock and costing. Redesigned warehouse locations, added barcode scanning, kit bills of materials for bundles and cash controls at the tills. An audit of four high-volume SKUs had found about 189 units unaccounted for.
  • Put discounts and marketing on evidence. Discount rules by channel, customer segmentation and basket analysis for bundles, and ad spend checked against verified store orders. One agency reported a return on ad spend of 13.6; verified orders showed 0.40.

Where the EGP 15.9M comes from

Every line below starts from the brand's own order, courier and ledger data, so the finance team can challenge each one. Inputs marked (E) are estimates used to split the total across levers.

Annual savings against the 12 months before the programme
LeverBaselineChangeEGP a yearUSD a year
Discount rules, online store19.7% of EGP 41.2M gross salesDown to 12.0%3,172,00063,440
Discount rules, branches and B2BEGP 79.7M of sales4.5 points lower (E)3,587,00071,740
Cash-on-delivery and cash reconciliationEGP 3.74M a year of unexplained cash differences85% recovered or prevented (E)3,175,00063,500
Stock loss and procurementCost of goods about EGP 49.1M (E)Shrinkage 3.5% to 1.0%, supplier and kit costing 3% lower (E)2,698,00053,960
Delivery fees and courier routingEGP 1.49M a year of delivery fees not recovered; 13,459 orders outside Cairo and Giza75% of fees recovered (E); half of those orders moved from EGP 120 to EGP 831,366,00027,320
Marketing spendAbout EGP 4.0M a year40% with no verified return cut (E), net of margin on lost sales1,216,00024,320
Back-office automationManual order entry, reconciliation and reportingFive roles redeployed at EGP 12,000 a month (E)720,00014,400
Total15,934,000318,680
The total is 14.6% of annual revenue. Rate used: EGP 50 = USD 1.

Pay-as-you-save

Vega Data is paid from the savings. We receive the measured savings for the first three years; from the fourth year the brand keeps all of them. If the savings do not appear, we are not paid, so our incentives match the owner's from the first day.

Platform and infrastructure

Azure Functions
Run the scheduled sync, validation and reconciliation jobs.
Application Insights
Monitors every job run and alerts on failures.
Azure Virtual Machines
Host the integration services and the customer-service ticketing system.
Azure Storage
Holds data exports and the daily exception reports.

Retail and e-commerce · 2023 to present

Turning 41 retailers' shelves into one comparable catalogue

How Fattuta, Vega Data's price-intelligence product, resolves about 465,000 store listings into 343,000 comparable products across Egypt, Saudi Arabia and the UAE, refreshed every night.

41retailers in three countries, refreshed every night

Read the case study

Government · 2020 to 2026

Ranking 460 crash hotspots with geospatial analysis and deep learning

Geospatial analysis and deep-learning risk models that place every crash on a national road network, score risk for every segment, and rank 460 priority hotspots on 142 roads, each with candidate countermeasures.

460priority hotspots identified on 142 national roads

Read the case study

Tell us the decision you need to make.

Describe the decision, the data you hold and your timeline. A senior engineer reads every message.