Three versions of revenue walk into the board meeting.
Finance, sales and operations each bring a number. All of them are "right". None of them match.
Data & analytics strategy · 19 years
I fix the data problems dashboards get blamed for.
Most dashboard arguments aren't dashboard problems. They're ownership problems wearing a dashboard costume. I work on what sits underneath: the definitions, models and governance that decide whether leaders trust a number. Then I make it ready for AI.
- 19 yrs
- in data, analytics and delivery
- 10 → 1–2 days
- reporting turnaround after an enterprise rebuild
- 10+
- enterprise systems unified into one model
- 20+
- governed semantic models in production
- 6
- countries of delivery: BE, NL, DE, UAE, US, India
Sound familiar?
If this is your Monday, the problem isn't the dashboard.
Month-end takes two weeks.
Analysts stitch spreadsheets together while the decisions wait for them.
Hundreds of reports, and still a request queue.
Everyone built a dashboard. Nobody trusts one enough to stop asking for a new one.
The AI pilot stalled at the data.
The model was fine. The definitions underneath it weren't, and nobody owned them.
Services
Strategy you can act on, from someone who has built it.
I don't hand over a deck and leave. Every engagement ends with something running: a model, a glossary, a forum, a team that knows how to keep it going.
01
Analytics & BI transformation
From fragmented reports to one governed system.
I assess your reporting estate, design the target architecture (warehouse, marts, semantic layer) and lead the rebuild, starting with the reports people already depend on.
- Current-state audit and verdict
- Target architecture and roadmap
- Delivery leadership through migration
02
KPI governance & semantic layers
One definition, one owner, one number.
A KPI glossary with named owners, conformed dimensions, and Power BI semantic models built to be reused rather than rebuilt, with security and certification baked in.
- KPI glossary and ownership model
- Semantic model standards and reviews
- Model optimisation and clean-up
03
AI-readiness with AIR Foundation
Get the data ready before the use case.
Conformed data, governed KPIs, quality rules and semantic definitions, then a readiness score for each AI use case on your list, based on what actually exists.
- Readiness assessment by domain
- Use-case scorecard
- Remediation plan, in order
04
Fractional data leadership
A senior data leader, without the full-time hire.
A part-time head of analytics for growing businesses: set the strategy, coach the team, run the governance forum and sit with leadership when the numbers are on the table.
- Data strategy and operating model
- Team hiring and coaching
- Leadership-level sparring partner
Ways to work together
2–3 weeks · fixed scope
Data health diagnostic
I trace your critical KPIs from source to slide and interview the people who argue about them. You get a written verdict and a 90-day plan.
3–6 months
Transformation programme
Architecture, governance and delivery leadership, working with your team rather than around it, until the new system is the one people use.
Monthly
Advisory retainer
A senior sparring partner for your data leader or CXO. Architecture reviews, hiring, vendor choices, and a second opinion before the expensive decision.
Approach
Boring foundations, first. Then the shiny things work.
-
1
Diagnose
Sit with the people who disagree about the numbers. Trace three critical KPIs from source system to board slide, and find where they break.
-
2
Define
Agree one definition and one owner for every KPI that matters. Write it down somewhere people will actually find it.
-
3
Build
Warehouse, conformed dimensions, semantic models. Automate the reports people already rely on before inventing new ones.
-
4
Adopt
Data Champions in every function, self-service with guardrails, and a governance forum that actually meets.
Selected work
Programmes I've led, and what changed.
B2B commerce & supply chain · India · 2019–present
Enterprise BI, built from scratch
Reporting ran straight off transactional systems, and every function had its own version of the truth. I built the analytics function from the ground up and led a team of 10–15 to replace it with a central warehouse and a semantic layer now used from the CXO level through operations, by finance, investors and auditors.
- Pan-India reporting framework: 434 decks and 82 Power BI workspaces for 200+ users
- 50+ Data Champions driving self-service adoption across functions
- Standardised business-unit P&L visibility and KPI definitions across leadership and operations
- 10 → 1–2
- days to produce business reporting
- 10+
- platforms integrated: ERP, CRM, finance, operations
- 20+
- governed semantic models
- 70+
- executive dashboards
Semantic model modernisation
Half the model, all the meaning
Rebuilt enterprise Power BI semantic models with AI-assisted tooling (the Microsoft Power BI MCP Server), simplifying years of growth into models that are faster to refresh and easier to maintain.
- 1.5 GB → 700 MB
- model size
- 90+ → 48
- tables
Retail · Belgium
A full BI landscape for a European retailer
Owned India delivery of the entire BI estate for Zeb, a Belgian fashion retailer, as the single point of contact for its business. Set up the team's methods onsite in Brussels, owned master data, and ran the SQL Server, SSAS Tabular and Power BI stack.
Financial services · global
Ratings data into the global model
Led the data team for S&P Global Ratings across two streams: moving LATAM surveillance data out of spreadsheets into the global data model, and designing the data model for Green Evaluations, with the governance rules to keep it clean.
Consumer goods · multi-country
One reporting standard, many markets
Managed Reckitt Benckiser's multi-country health and hygiene reporting migration to the cloud: defined close to 50 KPIs, led development of 100+ reports, and secured sign-off from every market.
Programmes delivered for
- S&P Global Ratings
- Reckitt
- HeidelbergCement
- Genpact
- Bose
- Zeb
- Jubilant FoodWorks
- Arvind
AI-readiness
AI can't fix numbers your business doesn't agree on. It just gets the wrong answer faster.
So I build the governed foundations first. AIR Foundation is my framework for getting a business domain ready for AI before any use case is built. I built a working reference version on a fictitious distributor with synthetic data, with one hard rule: no AI in it.
Ask about an AI-readiness assessment- 01
Raw data in
Sales, finance and operations data exactly as it lands today.
- 02
Conformed facts & dimensions
One customer, one product, one calendar, everywhere.
- 03
KPIs with a definition and an owner
Every number traceable to a rule and a person.
- 04
Quality rules, checked against the data
Not a policy document. Tests that run.
- 05
A readiness check per AI use case
Go, fix first, or not yet, based on what actually exists.
Hands-on
I build with AI, not just talk about it.
Using Claude, Codex and Antigravity. Not everything I try works, and I learn the most from what breaks. I'd rather advise leaders from what I've shipped than from what I've read.
Framework · reference build
AIR Foundation
Takes a business domain from raw data to an AI-readiness verdict: conformed model, governed KPIs, quality rules, use-case scoring.
Content pipeline
Data Monk Shorts
A pipeline that turns my scripts on data interviews into finished YouTube Shorts. One rule: every query result on screen is executed for real.
Product · reference build
ekparchi
A phone-first tool for Indian micro-businesses. The whole spec fits in one line: a PDF in a WhatsApp chat in under two minutes.
Live
in5.bio
A personal intro in five seconds.
The Lab · free tools
Try the thinking before you talk to me.
Run a BI health check on your own setup, sketch a star schema, or get a tricky stakeholder message right. They run in your browser, no sign-up needed.
Mentorship
I also teach what took me 19 years to learn.
Most data careers stall on a roadmap problem, not a talent problem. Random courses, nothing connecting. I teach the structure underneath.
Course · by BI7 Analytics
DA51: from BI Analyst to BI Architect
51 hours. 6 modules. One clear outcome.
The path I wish I'd had: built on modern data architecture, taught the way real projects run, from the source system to the dashboard a CXO actually opens.
- AFoundations and the BI7 framework
- BUnderstanding OLTP
- CSetting up loads
- DData mart creation
- ESuccess mapping with stakeholders
- FPower BI reporting
1:1 mentorship
For data people taking the next step
A small number of mentees at a time, working on your real situation rather than a syllabus.
- Analyst → architect
- Modelling, warehouse design and semantic layers, reviewed against your actual work.
- Into data consulting
- For senior practitioners moving into consulting: framing problems, scoping engagements, running stakeholders, and proving value.
- Leading a data team
- Operating models, governance forums, hiring, and talking to a CFO about a number.
About
Nineteen years on the unglamorous side of data.
I'm Atul Rai. I started in an insurance MIS team in 2007, building a Pan-India reporting mechanism before the warehouse existed. Since then I've been a business analyst, a project manager and a delivery lead on client programmes across Europe, the Middle East and the US, and I built an enterprise BI function from nothing.
Along the way I learned that the hard part is rarely the tool. It's getting a room of smart people to agree what a number means, who owns it, and what happens when it moves. That's the work I do.
Data Monk Labs is my practice: the consulting, the frameworks, and the teaching.
Onsite delivery in
Brussels · Amsterdam · Heidelberg · Dubai
Working stack
- Power BI
- DAX
- SQL Server
- SSIS
- SSAS Tabular
- BigQuery
- Informatica
- Redshift
- Grafana
- Power Automate
- Claude
- Codex
- Antigravity
- 2019 – now
Associate Director, Analytics
B2B commerce & supply chain. Joined as Senior Manager to build enterprise BI from scratch.
- 2018 – 2019
BI Functional Analyst, Team Lead
Hansamukh Software · Zeb, Belgium
- 2017 – 2018
Manager, Business Analysis (Data & BI)
Ness Technologies · S&P Global Ratings
- 2012 – 2017
Project Manager, BI & Data Transformation
Acuvate · Reckitt, HeidelbergCement, Genpact, Bose
- 2009 – 2012
Business Analyst
CITS · consulting and delivery programmes
- 2007 – 2009
Assistant Manager, MIS
SBI Life Insurance
Questions
Before you write.
Do you advise, or do you actually build?
Both. I'm most useful when I set the direction and then stay close to delivery: reviewing models, unblocking the team and making the hard calls with leadership. If you only need a second opinion, the advisory retainer covers that.
Do you work with companies outside India?
Yes. Much of my career has been client delivery for European, Middle Eastern and US businesses, including onsite work in Brussels, Amsterdam, Heidelberg and Dubai. Most engagements run remotely with onsite workshops where they help.
We're not on Power BI. Does that matter?
Not much. The hard problems (definitions, ownership, conformed dimensions, governance) are the same on any stack. I've delivered on Microsoft, Google, AWS and Birst, and I'll tell you honestly if a tool change isn't what you need.
We want to "do AI". Where do we start?
With a readiness check on the domain you care about most. It's quicker than a pilot and it tells you which use cases can go now, which need the data fixed first, and which aren't worth it yet.
Contact
Tell me which number nobody trusts.
A short note on what's going on is enough. I reply personally, usually within two working days, and the first call is free: 30 minutes to see whether I can help.