Case study · Architecture & economics
One number, two histories
When two systems disagree about a shared metric, the consolidation starts with its definition.
- Architecture
- Governance
- Stakeholder alignment
Anil Thapa
Data & AI Platform Leader
Both halves of the data problem.
Ask what the data platform cost last quarter and someone has the number. Ask what it changed and you get a list of dashboards. I have built platforms that gave that answer and led teams that had to do better, and the distance between the two answers is the subject of everything here. Every case study starts from a decision that was going to be made with or without the data, says what the platform had to get right for the evidence to arrive in time, and names what that cost. The essays carry the same question into the team: who owns a definition, what a recommendation has to carry before anyone acts on it, and how judgment is handed to the next person. Two short exercises give you the choice first and the cost after.
A platform is a cost until it changes a decisionanda career in data grows the same way.
For anyone paying for a data platform, or building a career on one.
What each half has to get right
The platform
Can we trust the number?
- Definitions that agreeOne meaning from source to report.
- Pipelines proven before useChecked before anyone relies on the answer.
- Costs someone can explainFreshness and reliability with an operating price.
Reliable numbers can still leave every decision unchanged.
The people and decisions around it
Can it still change the choice?
- An owner for the decisionResponsible for the choice, and for the definitions it rests on.
- Evidence while options are openInvolved before the decision is settled.
- Recommendations with a tradeoffA choice, an alternative, and a reason to reconsider.
A persuasive recommendation can still rest on the wrong number.
The decision sets the requirement. The evidence shapes the choice.
What someone needs to decide determines which definitions, reliability, and cost the platform must support. What the platform can establish determines which recommendations are defensible.
The interface may be a dashboard, a briefing, or an AI assistant. The definition, the evidence, and the accountability still have to hold.
The organization needs evidence that informs the choice before it is settled. The data professional earns trust by making the recommendation and its tradeoffs clear.
Two exercises
Choose first. Then read what it costs.
Every case study here names the decision the work changed and what the choice cost. These two hand you the choice first.
Compare the two01A fictional quarterFour demands. Capacity for two.Four reasonable requests to a central data team, room for two. Choose a pair and read what it buys, what it assumes, and what it leaves exposed.Start 02A decision of your ownWhat would this work change?Six questions about who decides, when, on what evidence, and what the next step costs. Leave with a question to ask next and a small step to take.Start
The experience behind the writing
About- Built from zero
- Central data functions, hired and structured from scratch
- In the room
- Analysis that reaches the decision while it is still open
- Hands on
- Still in the query plan when that is what the problem needs
- The tradeoff, written down
- Every case study names what its decision cost
What I write about
Each subject crosses both halves. None of them is only about the platform, and none is only about the career.
Architecture & economics What the platform must deliver, and what that promise costs.4 case studies · 2 essaysTrust & governance Who owns the definition, and what happens when a number changes.2 case studies · 3 essaysData in the business How evidence reaches a choice while alternatives are still open.1 case study · 5 essaysLeading data teams How hiring, ownership, and review develop judgment across a team.2 case studies · 6 essaysAgentic & AI systems What an assistant can answer, and how its use is evaluated and governed.1 case study · 5 essays
Selected case studies and essays
All writingCase study · Data in the business
Data in the boardroom
When analysis arrives after the decision, the work starts with being in the room while the options are still open.
- Exec partnership
- Stakeholder alignment
Essay · Leading data teams
When every decision still needs you
When judgment stays with one person, hand someone a bounded decision and review their reasoning.
Case study · Agentic & AI systems
What the assistant was allowed to know
When an assistant answers from the platform, launch it narrower than feels reasonable and let the unanswered questions set the roadmap.
- Architecture
- Governance
- Enablement
Recent essays
All essaysAgentic & AI systems
When the assistant costs more than the queue
How to decide whether an AI data assistant should be widened, narrowed or retired: four numbers against the queue it replaced, and what retirement keeps.
Agentic & AI systems
When the agent writes the pipeline, review is the platform
When AI drafts the pipelines and models, review becomes the constraint: what changes in the queue, what a reviewer checks, and what it costs a data team.
Leading data teams
An interview that asks for a decision
Hiring for judgment in data: an interview built around one bounded decision, what to listen for, how to score it, and what the format costs.
Leading data teams
Count the decisions, not the dashboards
The evidence of judgment in a data career is a record of decisions, not deliverables: four columns to keep, how to read them, and what the log costs.
For a talk, a podcast, or a different view on something written here, the writing is where to start. The contact page says what an invitation should include.