Case Studies

Challenge

bol.com had invested significantly in building sophisticated fraud detection systems using machine learning. However, despite having great tools in place, there was no visibility into whether these systems were actually working and how to quantify their impact. Fraud owned the numbers, but the analytics team owned the tools to stop fraud. Without owning the numbers, it was difficult to measure, improve, or stay the course on fraud prevention initiatives.

Solution

• Metric Definition and Alignment: Defined clear metrics for fraud detection effectiveness and worked to align ownership between the fraud team and analytics team.
• Ground Truth Analysis: Deployed ground truth analysis to verify whether blocks were actual fraud cases, understanding what the ML flags meant and systematically verifying blocked transactions.
• Visibility Dashboard: Built comprehensive dashboards that showcased the great work being done by the fraud detection team, making the impact visible to stakeholders across the organisation.

Results

Hard numbers showed fraud went from €900K in 2024 to €500K in 2025, demonstrating a €400K reduction in fraud losses. Established clearer ownership lines and confirmed the effectiveness of the ML-based fraud detection through ground truth analysis.

"Ahad brought tremendous energy to the project. What really set him apart was his commitment to alignment—he made sure we were all on the same page before development started, and then kept us aligned throughout. The metrics we defined together aren't just sitting in a dashboard, they're actually being used properly and don't get lost. That's rare."

— Atanas Trodowski, Product Owner, bol.

Challenge

Fastned didn't have a dedicated Data Warehouse to perform Business Intelligence on. They were using a copy of the transactional database from AWS to MySQL. The Backend was based in AWS while the Databases for Tableau were based on MySQL, creating a fragmented IT infrastructure. The current reporting only showcased charging sessions and sales without actionable insights.

Solution

• Data Warehousing: Set up AWS Redshift to make integration between backend and analytics systems seamless.
• Actionable Insights: Instead of dashboards showing state of business, started answering business problems and supporting decision making.
• Advanced Analytics: Leveraged regression analysis to find the next best station locations using sales data, EV data, and charging station accessibility.

Results

Improved data accessibility through centralisation. Developed a utilisation metric understanding concurrent charging to identify bottlenecks. Using data-supported location hunting, Fastned grew from 114 stations in 2019 to 297 stations today.

"KYD Analytics helped revamp our data infrastructure. Ahad's drive and eagerness to learn and solve problems really helped us in unlocking the value in our data. We highly recommend KYD Analytics to any organization seeking to harness the power of their data."

— Martijn Benjamin, Head of Software Development, Fastned

Challenge

Floward is a leading e-commerce platform in the Middle East specialising in flowers and gifts delivery. The existing batch ETL processes had latency measured in hours, making it impossible to respond quickly to business events in the fast-moving e-commerce space.

Solution

• CDC Pipeline: Replaced batch ETL with real-time Change Data Capture using Debezium. Built pipeline from PostgreSQL through Debezium to MSK Kafka, then Kinesis Firehose to S3 Iceberg.
• Segment Events Integration: Created parallel pipelines for mobile and web events through Segment for comprehensive customer journey tracking.
• Medallion Architecture with DBT: Implemented Bronze-Silver-Gold transformation layers using DBT with 36 models across 15 threads.

Results

Reduced data latency from hours to under 2 minutes. Deployed 11 CDC tables, 3 Segment tables, and 36 DBT models in fully automated, Terraform-managed infrastructure. CDC Pipeline median latency of 63.8 seconds, Segment Pipeline median latency of 78.7 seconds.

"The speed at which Ahad works is exceptional. We had real-time streaming operational in just 3 weeks—something we thought would take months. He didn't just deliver fast, he delivered right. The infrastructure he built has been rock-solid since day one."

— Amr, Engineering Lead, Floward

Challenge

ABN AMRO is the 3rd largest bank in The Netherlands. Due to ECB regulations, every bank needs to comply with certain IT Standards identified through internal surveys on ServiceNow. The deadline was approaching with no visibility on survey completion status or where surveys were stuck. A diverse group of stakeholders meant knowledge was scattered.

Solution

• Visibility and Accountability: Added survey data into the data model and brought KPIs into main reporting at leadership level, with details to identify where surveys were stuck.
• Enhanced Feedback Loop: Set up regular connects with stakeholders to consolidate information in one place for easier audit-related access.

Results

The organisation is now better equipped to identify potential bottlenecks in survey processes and address them proactively. Setting up regular stakeholder connections created a more organized approach to communication around audit-related matters.

"Ahad joined the team during a period when there was quite some pressure to deliver reporting in a short period of time. The reporting requirements were not yet defined. Ahad delivered actionable reports that were usable for the business in a very short period of time. A few meetings with stakeholders were enough for Ahad to translate an idea to a product."

— Paul Ten Kaate, Product Owner – Tower Insights, ABN AMRO

Challenge

Allianz Direct is a B2C subsidiary of Allianz Insurance dealing directly with customers for Auto and Home Insurance. Their EDM based on star schema was slowing down as the model grew. The centralised structure was causing bottlenecks, and brokers/agents were disconnected from their commission reports.

Solution

• Infrastructure as Code: Facilitated building a decentralised Data-Mesh infrastructure designed by the in-house team.
• Broker Portal: Built frontend reports using React and PowerBI Embedded after PowerApps implementation seemed unpolished. Implemented Row Level Security through Single Sign On.

Results

The decentralisation empowered teams to be responsible for their own data with dedicated Data SPOCs. Brokers gained access to their filtered data with RLS, enabling them to upsell and reach targets more easily.

"He consistently delivered exceptional results. He demonstrated strong technical skills, and contributed significantly in achieving teams goals. Ahad's professionalism and excellent communication skills made him a standout team member. The positive energy and enthusiasm that he brings to the team, is unmatched."

— Vaibhavi Rege, Chief Data Officer, Allianz Direct

Challenge

Buffering is one of the biggest KPIs for VodafoneZiggo for 2024/2025, directly impacting customer experience and retention. Customers were experiencing buffering on their Mediaboxes, but there was no systematic way to predict when buffering would occur or identify root causes before customers started suffering.

Solution

• Telemetry Model Design: Analysed requirements and designed rule-based logic models to identify patterns in telemetry data correlating with buffering events.
• Machine Learning Implementation: Collaborated with the data science team to build ML models that predict buffering events using historical telemetry data.
• Self-Service Flow Integration: Identified leading indicators (packet size, internet quality) and connected insights to self-service flows for proactive customer outreach.

Results

Successfully identified leading indicators that predict buffering events before they impact customers. Enabled self-service flows that automatically reach out to customers with solutions before buffering becomes a significant issue. The ML models achieved 81% accuracy in predicting buffering events, enabling proactive intervention.

"Working with Ahad on the telemetry analytics was a great experience. His vibrant personality made him a real asset to the team—he was always present and ready to answer stakeholder questions about our data products, no matter how technical or business-focused. He brought clarity to a complex domain and helped us build models that actually work in production. The buffering prediction system he helped develop has become a cornerstone of our customer experience strategy."

— Peter Schweigmann, Product Owner Telemetry In-Home, VodafoneZiggo

Challenge

Metito Group is a leading utilities provider based in Dubai with operations spanning 28 entities across multiple regions. Financial data was spread across Infor LN and SAP systems with no unified view. They planned to decommission Infor LN, requiring complete data lineage and migration of 10 years of historical financial data.

Solution

• AWS Data Warehouse: Set up AWS Redshift as central data warehouse, using AWS Glue for extraction and S3 as landing zone.
• Data Lineage Framework: Established complete data lineage from source systems through transformation layers to final Power BI reports.
• Dimensional Modeling: Transformed raw financial data into dimensional format optimised for reporting across all 28 entities.

Results

Consolidated financial reporting across 28 entities into a single, scalable platform. Established foundation for Infor LN decommissioning with complete data lineage and 10 years of historical data preserved in the cloud.

Challenge

Benerail is the technical provider of rail tickets to the biggest rail operators in Europe, operating since mid-2000s. They had gut-based decision making without data support. Their invoicing service for suppliers was manual, costing thousands of euros yearly and couldn't scale.

Solution

• Data Infrastructure: Built Analytical Database using Microsoft SQL Server coupled with PowerBI, creating a single source of truth.
• Automated Invoicing: Leveraged automation to build an automated invoicing tool using the Data Infrastructure coupled with Microsoft Business Central.

Results

The Data Model became the single source of truth for the whole organisation. Benerail now provides Data Products such as Automated Invoicing as a Service, Reporting as a Service using APIs, SQL Endpoints, and PowerBI Dashboards.

"Ahad joined our team right after we chose to implement Power BI as our reporting tool. He made sure he understood our new data-model, in order to maintain and to further develop it. Ahad was able to bring a lot of knowledge and skills about data-modelling, DAX, visualisations, adding new sources via datafactory and databricks, all while thinking along about the future of our data organisation and culture."

— Sanne Boone, Product Owner Business Development, benerail

Challenge

GulfMerc Exchange is a trading and chemicals exchange company in the Middle East. Trading and financial data was scattered across SFTP files, Excel spreadsheets, PDFs, APIs, and databases. Monthly reporting took 6 days of manual effort with no self-service analytics capability.

Solution

• Medallion Architecture: Built Bronze-to-Gold data pipeline using Microsoft Fabric with Bronze layer as landing zone for all sources.
• Dimensional Modeling: Designed Snowflake schema with facts and dimensions optimised for trading analytics.
• Self-Service Enablement: Delivered 10 Power BI dashboards with training for 20 users.

Results

Monthly reporting time reduced from 6 days to less than 1 day (95% reduction). Complete data infrastructure and 10 dashboards delivered in just 8 weeks. 20 business users trained and enabled for self-service analytics.

Challenge

Shopforward is a marketing agency running thousands of campaigns at their own cost, only getting paid on sales. They had no collective insight into campaigns running on multiple channels, and relied on database administrators for any data extraction, leading to slow turnover time.

Solution

• Data Model: Built first iteration of Analytical Database in MySQL using Campaigns, Sales, and Geodata sources, transformed into a Dimensional Model.
• Marketer Training: Provided training to marketers to build their own reports using PowerBI and extract insights for strategy optimisation.

Results

The first iteration showed such promising value that Shopforward decided to transition to Databricks soon after. They also onboarded an internal data scientist to support building better insights.

"We had the pleasure to work with Abdul and his team, which really paid off with getting PBI integrated in our daily routine. The whole integration went really smoothly, not only because of their knowledge, but also due to the fact that their response time was really fast and their ability to cope with business specific issues we faced."

— Dirk Verzijden, Founder, Shopforward