No one designs a data estate. It accumulates.
A warehouse from one era, a lake from the next, a mesh someone announced and half-built, and now a scramble of vector indexes and agents bolted to the side. Every layer was a reasonable decision at the time. Together they are a landscape nobody can hold in their head — and you are asked to draw a target architecture for it by Thursday.
Data Darśana is a map for that moment.
Not a maturity model that scores you, and not a reference architecture that assumes you are starting from nothing. A map: the whole territory laid out at once, so you can find where you actually stand, see what genuinely connects to what, and make the next decision with your eyes open.
| 248 | 6 | 20 | 128 | 13 | 1,888 | 112 |
|---|---|---|---|---|---|---|
| capabilities | pillars | domains | archetypes | families | links | diagrams |
Most capability models tell you what you must be able to do. Reference architectures tell you what good looks like in one shape. Almost nothing connects the two — and that gap is precisely where architecture lives, where the arguments happen, and where budgets are won or lost.
Data Darśana holds both halves and joins them:
-
Capabilities — what must we be able to do? 248 capabilities across 20 domains and 6 pillars, each placed in the stage of maturity where it first genuinely matters.
-
Archetypes — how do we actually assemble it? 128 architecture patterns in 13 families. Each one carries when to use it, how data flows through it, its governance model, its honest limitations, and where it typically evolves next.
-
The matrix — 1,888 graded links. Choose a pattern and read off the capabilities it demands as Core or Supporting. Choose a capability and see which patterns depend on it. That join is what most models leave to the reader.
The past is often still the right answer. Kimball and Inmon sit in this map beside the lakehouse and the semantic layer — not as history, but as live options. Batch ETL remains correct for a great deal of regulated reporting. A map that treats everything older than five years as a mistake will fail you in exactly the estates that matter most.
Honesty is more useful than enthusiasm. Every pattern here states what it costs. Data mesh is described as most programmes actually experienced it. "Self-healing pipelines" are not oversold. Where a technique is genuinely unproven, it says so — because the fastest way to lose a room of senior architects is to tell them something they know is untrue.
A diagram should argue, not decorate. Each drawing is made to show what the definition cannot say: where the control point sits, which arrow is the expensive one, what the shape quietly costs you. Sixteen archetypes have no diagram at all, deliberately — a picture that merely restates its definition is padding, and padding is the tell of a deck with nothing to say.
A map is only useful if you can find yourself on it. This one is built to be located in, not admired. Every archetype is something an estate somewhere actually runs, named the way practitioners name it, so the first useful move is recognition: that row is us. Everything after that — the gaps, the next stage, the honest cost — follows from standing somewhere real.
Sarasvatī is Vīṇā-vādinī — she who plays the seven svaras, the notes Sa Re Ga Ma Pa Dha Ni. Seven notes are not seven songs. They are one music, heard from seven positions.
The map is built the same way. One taxonomy, seven vantages: change the view and no fact changes — only where you are standing.
| Svara | View | What it answers | |
|---|---|---|---|
| ◉ | Sa · the ground | Graph | the home note — everything sounds in relation to the whole web |
| ✧ | Re | Constellation | every capability and archetype as one boundary-free field of stars |
| ✳ | Ga | Sunburst | the hierarchy blossoming outward from a single centre |
| ↗ | Ma | Roadmap | the journey — each capability in the stage where it enters |
| ▦ | Pa · the dominant | Heatmap | where intensity concentrates across pillars and stages |
| ◈ | Dha | Radar | the honest look inward — where you stand against the model |
| ▤ | Ni · leads home | Tables | the return to detail that resolves everything back to Sa |
This is offered as homage, not scripture: no claim on the tradition, only the truth that a single reality is richer when you can hear it more than one way.
The same 128 archetypes, sorted by the work they serve and the generation they belong to — proven, mainstream, frontier, side by side.
Read a row across and you are often reading how a kind of work evolved. Conformed dimensions become the semantic layer, which becomes the contract an agent queries. The ESB's service registry becomes the API gateway, which becomes the tool gateway an agent calls. Nothing was replaced so much as renamed and re-let.
Empty cells are left visible on purpose. They mark where the map is genuinely thin — and where an era simply has not happened yet.
"The table format war has largely resolved. The contested layer is the catalog above it, because that is what owns access control, lineage and the commit protocol."
"Cheap for the first three integrations and quietly fatal by the thirtieth: five endpoints is ten interfaces; thirty is four hundred and thirty-five."
"Only one node exits the loop. Everything else re-plans — so it needs a hard iteration cap, which is a product decision, not a tuning knob."
That is the register throughout: what it is, and what it will cost you.
- New to it — open the map and go to ▤ Tables → Architecture Eras. It is the fastest way to see the whole landscape and locate what you already run.
- Assessing an estate — use ◈ Radar to score yourself per pillar against the model's targets. The gap it draws is your roadmap.
- Planning a build — find your pattern in the Archetype Atlas, then read its Core capabilities. That is your minimum viable scope, before anyone negotiates it down.
- Preparing a board paper — the cheatsheet PDF is the whole model as a printable field guide, including the full era grid.
Darśana (दर्शन) means seeing — and in Indian philosophy, an entire way of seeing; a school of thought, a stance from which the world becomes legible.
The emblem is Sarasvatī, goddess of knowledge, music and flow — chosen because what she presides over is exactly what a map like this is for: knowledge held in the open, and free to learn from. The philosophy page tells that properly — why a goddess, why seven notes, and where the metaphor stops.
Data Darśana is a sibling to Bodhi Map — Awakening to AI.
Created by Sanjeev Azad
If it helps you see your own estate more clearly, it has done its job.
This repository holds the published site. The capability model, build pipeline and
diagram sources live in a private repository and are published from there.
Licensed under the terms in LICENSE.