Factory control center: image triage to annotate near incidents
This demonstration application shows how Fabric apps can be used in conjunction with Azure IoT Operations. On the edge of a local network, of a factory, building, or vessel, Azure IoT Operations can work with local devices. It can send data to Microsoft Fabric Real-Time Intelligence conditionally, using a secure connection via Azure Arc. What if we use a vibe-coded Fabric App for triage on images, taken at the factory floor while a vibration sensor detects abnormal behavior? On the Edge, an IP camera video stream and a LoRaWAN vibration sensor telemetry stream are combined using custom logic running in a Kubernetes pod. That data is then sent to an Eventstream custom endpoint and forwarded to an Eventhouse KQL Database table. Using table update policies, the table rows are cleaned and exposed in a Fabric Semantic model. Using vibe-coding in Visual Studio Code together with GitHub Copilot, the Fabric app is created. It shows how the latest images with the abnormal vibration deviation are presented. The user can then do triage and annotate the reason. This can then be shared with colleagues, customer or used for training agents. The application is built in ~1 day and demonstrates various features. Check this blog post. You can read about the full story on how the demonstration solution is built. Source of the custom code is available on GitHub.Executive Performance Dashboard
Executive Performance Dashboard About This Dashboard Executive Performance Dashboard - Built for Data-Driven Decision Making The dashboard covers: Overview – Key performance indicators and metrics Analytics – Deep insights and trend analysis Reporting – Automated and customizable reports Monitoring – Real-time data tracking and alerts What sets this apart? It's built on the principle that effective analytics should be both powerful and accessible. I focused on creating a user-centric design by: Highlighting the most actionable insights with clear visualizations Using intuitive navigation and clean interface design Providing interactive features for deeper exploration Ensuring fast performance and reliable data updates Making complex data accessible to all stakeholders Thank you, eyJrIjoiM2VkNWQxODYtMmRjOS00ZjZmLTliZTUtMWFhZDNiMzkyMDZhIiwidCI6ImQ4ZTFiMDVlLTcwYWEtNGVmNy1iODc4LTQ2NmI2ODhmOTUyZiJ9
African Gig Economy & Digital Wallet Risk Analysis
For the August #DataDNA challenge, I developed a Power BI report focused on understanding fraud exposure, fraud drivers, and operational risk across a digital wallet ecosystem serving the African gig economy. The objective was to move beyond reporting metrics and answer three business questions: What is the current risk exposure? What factors are driving fraud risk? Where should controls and monitoring be strengthened? To support this analysis, the report was structured into three sections: Executive Risk Overview Assess transaction performance, fraud exposure, disputes, and reversals. Fraud Driver Analysis Identify the customer, behavioural, and transaction characteristics associated with elevated fraud risk. Operational Risk & Controls Evaluate dispute patterns, reversal behaviour, transaction timing, and market risk indicators to identify opportunities for intervention. Key Observations Fraud exposure was more closely associated with customer behaviour, transaction type, and channel usage than demographics alone. USSD channels and Cash-In transactions consistently recorded elevated fraud rates. Month-end periods showed increased cash-out activity and reversal risk. Observed fraud exposure did not always align with expected market risk levels. The report also incorporates recent Power BI enhancements, including Donut Chart Center Value and Drop-down Slicer features introduced in the August Power BI update. Source: https://datadna.onyxdata.co.uk/challenges/august-2026-datadna-african-gig-economy-and-digital-wallet-analytics-challenge/
Mercado Español de Turismos: Evolución y Electrificación
🇬🇧 Spanish Passenger Car Market: Evolution & Electrification This Power BI project analyzes the evolution of the Spanish passenger car market and the expansion of vehicle electrification using more than 6 million official vehicle registration records from Spain's Directorate-General for Traffic (DGT). The projet goes beyond tracking registration volumes. Its purpose is to explore the structure and evolution of the market through several analytical perspectives: How is the passenger car market evolving? Which brands lead the market? Which are gaining or losing ground? How quickly is electrification expanding across Spain? Which technologies are driving this growth? And which brands and models are contributing most to the transition? The report analyzes the period from 2024 to August 2026, while 2023 data is also incorporated into the data model to provide the historical context required for year-over-year comparisons. 📊 Analysis Market Overview Provides a high-level view of passenger car registrations through KPIs, annual and monthly trends, geographical distribution and evolution by propulsion type. Market Analyzes the competitive landscape through brand rankings, market share, year-over-year variation and monthly trends. A scatter plot combines market share and YoY growth, making it possible to contextualize the position and evolution of the main brands. Electrification Examines the evolution of electrified passenger cars, distinguishing between PHEV, REEV, BEV and FCEV technologies. The analysis explores registration volumes, technology mix, monthly evolution and the geographical penetration of electrification across Spanish provinces. Electrified Brands & Models Identifies the brands and models contributing most to electrified registrations. The analysis combines registration volume, share and year-over-year growth with the technology mix of leading brands and a ranking of the most registered electrified models. Conclusions Brings together the main findings from the report, highlighting the growth of electrification, changes in the technology mix, leading brands and models, and territorial differences in electrification penetration. ⚙️ Data Preparation & Modelling One of the main challenges of the project was transforming the original DGT registration files into a consistent analytical dataset suitable for Power BI. Power Query was used to build the data preparation process, including: Integration and transformation of several years of registration data. • Classification of passenger cars and propulsion technologies. • Standardization of brand names. • Normalization of vehicle model names to reduce inconsistencies in the original registration data. • Creation of analytical categories for electrified vehicles. • Preparation of geographical dimensions for province-level analysis. • Creation of the temporal structure required for year-over-year and YTD analysis. The semantic model incorporates DAX measures for registration volumes, YTD calculations, previous-year values, year-over-year variation, market share and electrification share. Special attention was given to equivalent-period comparisons. Since the latest available 2026 data covers January through August, 2026 YTD indicators are compared against January-August 2025, preventing partial-year results from being compared with a complete previous year. This approach allows the same report to support both full-year historical analysis and comparable YTD analysis for the current year. 🎨 Report Design & User Experience The project also places strong emphasis on data storytelling, visual consistency and usability. The interface was designed to create a cohesive analytical experience across the different sections of the report, using a consistent visual hierarchy, navigation system and color language. The report includes custom navigation, interactive year selection and contextual help overlays that explain how to use filters and interpret key visualizations while keeping the underlying report visible. The final Conclusions page transforms the main analytical findings into a concise storytelling layer, connecting market growth, electrification, technology adoption, brand performance and geographical differences. The objective was to build a report that could be explored and understood independently, even by users who were not involved in its development. Tools: Power BI · Power Query · DAX · Figma Data source: Dirección General de Tráfico (DGT) Period analyzed: 2024 - August 2026 Dataset: 6M+ vehicle registration records 🔗 MORE ABOUT THE PROJECT 🚗 Full project & case study: https://cristina-mg.github.io/projects/turismos-electricos/ 📊 Data Analytics Portfolio: https://cristina-mg.github.io/ 🇪🇸 Mercado Español de Turismos: Evolución y Electrificación Este proyecto de Power BI analiza la evolución del mercado español de turismos y el avance de la electrificación a partir de más de 6 millones de registros oficiales de matriculaciones de la Dirección General de Tráfico (DGT). El proyecto va más allá del análisis del volumen de matriculaciones. Su objetivo es estudiar la estructura y evolución del mercado desde distintas perspectivas: ¿Cómo evoluciona el mercado de turismos? ¿Qué marcas lo lideran? ¿Cuáles están ganando o perdiendo terreno? ¿A qué ritmo avanza la electrificación en España? ¿Qué tecnologías están impulsando este crecimiento? ¿Y qué marcas y modelos están contribuyendo en mayor medida a esta transición? El informe analiza el periodo comprendido entre 2024 y agosto de 2026, incorporando también los datos de 2023 al modelo para disponer del histórico necesario para realizar comparaciones interanuales. 📊 Análisis Visión General Ofrece una perspectiva global de las matriculaciones de turismos mediante KPIs, evolución anual y mensual, distribución geográfica y comportamiento según el tipo de propulsión. Mercado Analiza la estructura competitiva mediante rankings de marcas, cuota de mercado, variación interanual y evolución mensual. Un gráfico de dispersión combina cuota de mercado y crecimiento interanual, permitiendo contextualizar la posición y evolución de las principales marcas. Electrificación Profundiza en la evolución de los turismos electrificados diferenciando las tecnologías PHEV, REEV, BEV y FCEV. El análisis estudia el volumen de matriculaciones, el reparto entre tecnologías, su evolución mensual y la penetración territorial de la electrificación por provincia. Marcas y Modelos Electrificados Identifica las marcas y modelos que más están contribuyendo a las matriculaciones electrificadas. El análisis combina volumen, cuota y crecimiento interanual con la distribución tecnológica de las principales marcas y un ranking de los modelos electrificados con mayor número de matriculaciones. Conclusiones Integra los principales hallazgos del informe, conectando la evolución de la electrificación, los cambios en el reparto tecnológico, el comportamiento de las principales marcas y modelos y las diferencias territoriales en la penetración de los turismos electrificados. ⚙️ Preparación y Modelado de los Datos Uno de los principales retos del proyecto fue transformar los ficheros originales de matriculaciones de la DGT en un conjunto de datos consistente y preparado para el análisis en Power BI. Mediante Power Query se desarrolló el proceso de preparación de los datos, incluyendo: Integración y transformación de varios años de matriculaciones. • Clasificación de turismos y tecnologías de propulsión. • Estandarización de las marcas. • Normalización de los nombres de los modelos para reducir las inconsistencias presentes en los datos originales. • Creación de categorías analíticas específicas para los vehículos electrificados. • Preparación de las dimensiones geográficas necesarias para el análisis provincial. • Creación de la estructura temporal necesaria para los análisis interanuales y YTD. El modelo semántico incorpora medidas DAX para calcular volumen de matriculaciones, YTD, valores del año anterior, variación interanual, cuota de mercado y cuota de electrificación. Se prestó especial atención a las comparaciones entre periodos equivalentes. Al disponer de información de 2026 entre enero y agosto, los indicadores YTD de 2026 se comparan con enero-agosto de 2025, evitando comparar los resultados de un año parcial con los de un año completo. Este planteamiento permite utilizar el mismo informe tanto para analizar años históricos completos como para realizar un seguimiento YTD comparable del año en curso. 🎨 Diseño y Experiencia de Usuario El proyecto también presta especial atención al data storytelling, la coherencia visual y la usabilidad. La interfaz se diseñó buscando una experiencia analítica coherente entre las distintas páginas del informe, manteniendo una jerarquía visual, navegación y lenguaje de color consistentes. El informe incorpora navegación personalizada, selección interactiva del año y un sistema de ayuda contextual que explica el funcionamiento de los filtros y la interpretación de las principales visualizaciones manteniendo visible la página sobre la que se está trabajando. La página final de Conclusiones transforma los principales resultados del análisis en una capa de storytelling que conecta la evolución del mercado, el avance de la electrificación, las tecnologías, el comportamiento de las marcas y las diferencias territoriales. El objetivo final fue construir un informe que pudiera ser explorado y comprendido de forma autónoma, incluso por usuarios que no hubieran participado previamente en su desarrollo. Herramientas: Power BI · Power Query · DAX · Figma Fuente de datos: Dirección General de Tráfico (DGT) Periodo analizado: 2024 - agosto de 2026 Dataset: +6 millones de registros de matriculaciones 🔗 MÁS INFORMACIÓN 🚗 Proyecto completo y caso de estudio: https://cristina-mg.github.io/projects/turismos-electricos/ 📊 Portfolio de Data Analytics: https://cristina-mg.github.io/ 💼 LinkedIn: https://www.linkedin.com/in/cristina-mart%C3%ADnez-garc%C3%ADa-dataWindFingerprint: pointing at air pollution sources using nothing but wind data
WindFingerprint: pointing at air pollution sources using nothing but wind data For any EEA air quality monitoring station, WindFingerprint joins hourly pollutant concentration to hourly wind direction and speed and draws the two plots atmospheric scientists use to work out where pollution is coming from, including, as of this week, the sharper of the two overlaid on a street map so you can see what's sitting in the direction it points to. What it does Hourly pollutant concentration (NO2, PM10, PM2.5) joined to hourly wind direction and speed, EEA + Open-Meteo, 20 stations in the Netherlands, 2020–2025. A bivariate polar plot: mean concentration by wind sector (24) and wind speed bin (6), a hot spot at the centre means a source right next to the station, a lobe further out means a distant one in that direction. A conditional probability function (CPF) rose: share of hours exceeding a threshold, by direction, the sector reaching furthest out points at the source. The CPF rose overlaid on a real OpenStreetMap basemap, centred on the station, so you can pan around and check what's there. A threshold slider that recomputes the CPF rose instantly from a precomputed histogram, no query per move Why nothing calls a live backend The first version queried a Fabric Lakehouse SQL analytics endpoint live, one query per threshold change, from a Rayfin Function using `NTILE(100)` to approximate a 90th-percentile threshold. It worked, right up until deployment, which returned a plain 400: Fabric does not yet support running Rayfin Functions in production. The function is still in the repo, parked. Everything the app shows now comes from a Spark notebook that runs once against the gold Lakehouse tables and writes one small JSON file per station and pollutant, a binned concentration grid, a default threshold, and a histogram per wind sector with shared bucket edges. Moving the slider does a linear interpolation inside whichever bucket contains the new value, client-side, no request, faster than the live version it replaced. Two details were wrong in ways that were easy not to notice. The EEA's validity flag has values 1 through 3, and 2 and 3 both mean "valid, but below the detection limit", filtering on exactly 1 quietly drops legitimate readings. And the boundary between the EEA's verified and unverified datasets isn't the fixed year the documentation implies; some countries submit the previous year's verified data early, so the pipeline tries the verified dataset first and only falls back to unverified if that comes back empty, per year, per country. What it deploys into your workspace Entra sign-in, via Fabric's own brokered SSO A Fabric SQL database, holds saved views and per-sector annotations; the pollution data itself ships as static files, not through this database A static web app Data The EEA's official Air Quality Download Service, joined against Open-Meteo's historical ERA5 wind reanalysis. Twenty stations in the Netherlands, 2020–2025, 2.5 million-plus joined hourly readings, the pipeline generalises to the rest of the EEA's network, but only the Netherlands is loaded so far. Credits Air quality: European Environment Agency, Air Quality Download Service. Reuse permitted with attribution — "Source: European Environment Agency (EEA)." Wind: Open-Meteo historical weather API, CC BY 4.0, non-commercial tier. Basemap for the map overlay: © OpenStreetMap contributors, ODbL. Try it Run your own copy: git clone https://github.com/AncovandenBerg/FabricApps.git cd FabricApps/apps/windfingerprint npm install npx rayfin up --workspace-id <your-workspace-guid> --tenant <your-tenant-guid> Source: https://github.com/AncovandenBerg/FabricApps/tree/main/apps/windfingerprintEuropean Summer Getaways 2026
This report explores European Summer Getaways 2026, helping commercial and portfolio teams identify the strongest destination opportunities, understand the drivers of destination attractiveness, and evaluate how holiday packages can be positioned for different traveller needs. Through a three-page analytical journey, the report moves from opportunity prioritisation to destination analysis and concludes with traveller-focused package positioning recommendations. Portfolio: Where Should We Focus? The first page identifies the strongest portfolio opportunities by balancing destination appeal with package availability and portfolio depth. The analysis moves from destination types to countries and cities, helping identify where commercial attention and promotion may deliver the greatest opportunity. Outcome: Prioritise the destinations and markets that deserve greater commercial focus. Drivers: What Drives Destination Attractiveness? The second page explores why some destinations perform better than others. Weather conditions, temperature and tourism pressure provide context around destination attractiveness, while the performance matrix highlights the trade-off between strong appeal and visitor pressure. Outcome: Understand the factors behind destination performance instead of relying on ranking alone. Travellers: Which Holiday Is Right for Which Traveller? The final page shifts from destination performance to package value and positioning. The analysis compares traveller spending, accommodation pricing, travel convenience and city-level value to understand how attractive destinations can translate into commercially relevant holiday packages. Outcome: Support package positioning around value, convenience and different traveller needs. Final Takeaway One of the biggest challenges with such a rich dataset was deciding what not to show. Rather than trying to analyse every available field, I focused each page on a clear business question and built the report around a simple journey: Opportunity → Understanding → Action A rich dataset does not need a crowded report. Every visual should earn its place by helping answer the business question. Data Source: FP20 Analytics ZoomCharts Challenge 40 | Summer 2026 Holiday Dataset
Veterans Affairs (Standardized Design.VA.gov)
A Power BI theme aligned to the current U.S. Department of Veterans Affairs Design System (VADS), which builds on the U.S. Web Design System (USWDS). Colors are mapped to VADS semantic tokens, typography, palettes, and styles to allow reports to closely match modern VA.gov digital styling standards. What's included: the full VADS-derived data color sequences, semantic good/neutral/bad mapping for conditional formatting, and per-visual defaults for cards, tables, matrices, slicers, KPIs, gauges, charts (bar, column, line, area, scatter, waterfall, pie/donut, treemap, funnel, ribbon, etc.), reference/trend/error lines, tooltips, and the filter pane. Titles, axes, gridlines, and legends are pre-styled defaults to ensure a clean, consistent look, with white canvas backgrounds and a very subtle tint on visuals to separate them from the page, while font sizing favors readability. Designed for VA analysts, programmers, designers, or employees who want VADS/USWDS-consistent Power BI reports. Notes: Licensed under Apache-2.0. & based on the VADS color palette as published on https://design.va.gov. A colorblind-friendly companion theme is available separately in the Microsoft Fabric theme gallery.Human Resource Workforce Analytics Dashboard
Human Resource Workforce Analytics Dashboard In today’s dynamic business landscape, managing talent effectively requires more than intuition. This dashboard delivers real-time insights into workforce trends, diversity & inclusion, compensation, and retention, all in one clear view. It empowers leadership with data-driven visibility to: - Make smarter talent decisions - Track diversity and inclusion progress - Identify risks and opportunities early - Shift HR from a reactive to a proactive strategy People Analytics is now a strategic necessity. Visit Website: https://arbazahmad-bi.netlify.app/ Thank you, eyJrIjoiNWVmYjhkNDYtYmJjNS00MDA5LTkwOWMtOTZiZDMyNjRhNTEyIiwidCI6ImQ4ZTFiMDVlLTcwYWEtNGVmNy1iODc4LTQ2NmI2ODhmOTUyZiJ9
Live Weather Dashboard
Live IoT Weather Dashboard This is a real-time weather dashboard powered by a custom-built DIY IoT weather station. While many weather dashboards available online use static datasets or merely change color schemes on the same template, this project is fundamentally different. It is a live system that bridges the gap between physical hardware and digital analytics. Technical Overview Data Collection: Unlike dashboards that rely on CSV uploads, this station utilizes an ESP8266 microcontroller paired with DHT11 and BMP280 sensors. Live readings for Temperature, Humidity, and Pressure are logged directly to Google Sheets in real-time. Region-Specific Calibration: Specifically calibrated for the unique tropical climate of the Trichy, Tamil Nadu region, ensuring localized accuracy that global APIs often miss. Advanced Analytics: Using only four raw data points (Timestamp, Temp, Humidity, and Pressure), I have developed a custom suite of DAX measures. These calculate complex weather trends, "feels like" temperatures (Heat Index), and atmospheric dew points. Dynamic Visuals: The interface isn't just a static skin; it features weather-dependent iconography and a custom refresh schema that ensures the data you see is as fresh as the air outside. Why this stands out Most dashboards you see are "re-skins" of existing templates. This project demonstrates a full-stack data engineering pipeline: Sensor → Microcontroller → Cloud Storage → Power BI Logic → End-User Visualization. eyJrIjoiYTVjNzg4MzMtZDJlMS00NmY5LWI1ZDItNzRhMjRmNDdkZWIwIiwidCI6IjBjM2QwNTc2LTFkOWYtNGM4Ny05OTNjLTg2YjQ0MGE1YjA3OCJ9HR Analytics Dashboard
Project Overview This end-to-end Power BI project analyzes a workforce of nearly 60,000 employees to uncover the hidden drivers behind attrition and employee satisfaction. The goal was to transform raw HR data into actionable insights for leadership to improve retention and organizational health. Dashboard Pages & Key Insights 1. Executive Overview Purpose: High-level KPIs for leadership. Metrics: Total Headcount, overall Attrition Rate, Average Salary, and Tenure. Insight: Provides an immediate snapshot of the company’s stability and scale. 2. Attrition & Retention Analysis Purpose: Identifying "Why" and "Where" people leave. Metrics: Attrition by Job Role, Overtime impact, and Distance from Home. Insight: Highlights high-risk departments and the correlation between work-life balance and turnover. 3. Compensation & Performance Purpose: Analyzing pay equity and growth. Metrics: Monthly Income by Gender and Job Level, and the link between Performance Ratings and Promotion frequency. Insight: Ensures transparency in pay scales and evaluates if the company is "promoting from within" effectively. 4. Demographics & Work Environment Purpose: Understanding workforce diversity and culture. Metrics: Remote work adoption, age distribution, and sentiment analysis (Job Satisfaction & Company Reputation). Insight: Assists in tailoring culture-building initiatives based on actual employee feedback and demographic needs. Technical Toolkit (DAX & Modeling) Advanced DAX: Created measures for Attrition % using CALCULATE and DIVIDE to handle dynamic filtering. Data Transformation: Cleaned and structured 24+ columns for optimized performance. Visualization: Used a mix of heatmaps, distribution charts, and KPI cards to ensure scannability. eyJrIjoiMjVkNmJhMDktYTE4My00YTk3LTk5YWEtOTAwNjk2NzMyNjgzIiwidCI6IjBjM2QwNTc2LTFkOWYtNGM4Ny05OTNjLTg2YjQ0MGE1YjA3OCJ9
