WHY PROPPERLY · YOUR ANALYTICS CAPABILITY, AI-USABLE

You already have the stack. What is still missing?

A warehouse. A semantic layer. A catalog. Search or RAG. An AI agent that can reach all of them. Each of those is valuable, and Propperly does not ask you to replace any of them.

The gap they leave is the same one every analytics team lives with: the conclusions and decisions change over time, and they live across those systems rather than inside any one of them.

Propperly maintains what your analytics organization has actually concluded — what is current, what it superseded and why, where sources conflict, and what is still unresolved — so your agent doesn't confidently repeat a decision you already changed.

Complementary to your stack. No platform migration to understand where it fits.

what you already have
warehousesemantic layercatalogsearch / RAGAI agent
the object Propperly maintains

What your organization has concluded — and still stands behind — for a defined scope.

currentsupersededconflictunresolvedevidence
Your systems can find the sources. What should the agent rely on now?
Where it fits

Each system governs a different object.

The systems you already run each focus on a different object. Propperly works alongside them, maintaining the evolving analytical conclusions and decisions that span those systems.

Warehouse / data platformPrimary object · the data
Stores and computes the rows and tables an answer is built from.
Semantic layerPrimary object · metric definitions
Defines how a metric is computed, so the same question returns the same number.
Catalog / governancePrimary object · metadata
Governs data assets, lineage, ownership and policy across the estate.
Search / RAGPrimary object · findable content
Retrieves the existing documents and passages most relevant to a question.
AI agentPrimary object · reasoning
Reasons over, generates from and acts on whatever those systems expose.
PropperlyPrimary object · analytical conclusions & decisions
Maintains the evolving conclusions the organization expects people and agents to build on — the reviewed current state, with what it replaced, where it conflicts, and what is still open.
currentsupersededconflictunresolvedevidence

These are primary roles, not walls — some systems hold a little of a neighbour's object. Propperly focuses on maintaining the analytical conclusion as it changes across the stack.

One concrete example

A definition changed. Who knows what now?

A single, ordinary analytics decision — the kind that happens somewhere in your organization every week.

active_customer — one decision over time
  1. A 30-day active-customer window was the working definition.metric doc · jan 12
  2. A new analysis challenged it.analysis · apr 28
  3. A person decided to move product reporting to a 28-day window.decision · may 04 · approved
  4. The 30-day definition is now superseded — but only for that scope.supersedes · reporting scope
  5. Three dashboards still calculate the old way.conflict · not yet propagated
  6. Whether to restate history is still unresolved.open · owner: analytics
What your systems may know
  • WarehouseThe rows behind both windows.
  • Semantic layerHow active_customer is currently computed.
  • CatalogOwners, lineage and metadata for the tables.
  • Search / RAGDocuments that mention active_customer.
  • AgentCan retrieve and reason over all of the above.
What Propperly maintains
  • Which definition governs which scope now.
  • What it replaced, and that the old one still holds elsewhere.
  • Why it changed — the evidence and the decision behind it.
  • What hasn't propagated yet (the three dashboards).
  • What remains unresolved (the historical restatement).
  • Who approved the change, as reviewed state — not an inference.
The object, in four states

Conflict and unresolved are states, not errors.

Propperly represents a conclusion in whichever of four states it is actually in. Conflict and unresolved are first-class states the agent can see — not failures to hide, and not things Propperly silently resolves on its own.

current

Current

What the organization currently stands behind for the defined scope — the conclusion an agent should rely on now.

superseded

Superseded

What used to govern, what replaced it, why it changed, and where the old state may still legitimately apply.

conflict

Conflict

Multiple reasonable sources or interpretations still disagree — kept visible, not collapsed into a single winner by default.

unresolved

Unresolved

The organization has not yet made a supported decision. The honest state is "open" — and the agent is told so, rather than guessing.

Each state carries its supporting evidence and, where relevant, the human decision and the scope it applies to.

Human authority

Who gets to turn evidence into organizational state?

Propperly can reconstruct and propose. It does not decide what is true. A model's inference never silently becomes authoritative — a person does.

On every proposed conclusion, a reviewer can:

Approve.Edit.Reject.Leave unresolved.

The point isn't only that a human reviews. It's that a reviewed decision becomesdurable state the next person or agent can build on — versioned, evidence-linked, and scoped — instead of a one-off answer that disappears after the chat.

Grounding proves the evidence exists. Humans decide what the organization stands behind.

How this becomes an engagement

Establishing this reviewed current-state for one bounded analytics domain — before any large integration — is exactly what the Agent Readiness Sprint does, from work your team already has.

Object by object

The same question, across categories.

Not a feature scoreboard. Each row is an object; each cell is that category's primaryrelationship to it. Where a neighbour partly overlaps, this shows it.

ObjectWarehouseSemantic layerCatalogSearch / RAGPropperly
Primary governed objectThe dataMetric definitionsMetadata & policyFindable contentAnalytical conclusions
Retrieves the underlying evidencePrimary purposeNot primary objectCan supportPrimary purposeLinks evidence to each conclusion
Defines metric semanticsNot primary objectPrimary purposeCan supportNot primary objectDefers to your semantic layer
Data & metadata governancePartialCan supportPrimary purposeNot primary objectDefers to your catalog
Evolving analytical conclusionsNot primary objectNot primary objectNot primary objectNot primary objectPropperly focus
Current vs superseded, with why & scopeLatest overwritesPartial — deprecationPartial — lifecycleNot primary objectKept, with why + scope
Conflict & unresolved as durable stateNot primary objectResolved by constructionPartial — open issuesNot primary objectFirst-class state
Human-reviewed revision as maintained stateNot primary objectPartial — in codePartial — certifies the assetNot primary objectThe decision is the object

Each cell describes where a category's authority primarily lives, not a limit on what it can do — "partial" and "not primary object" don't mean a tool can't support the capability. Propperly is additive above the systems you keep; it doesn't claim to beat search or RAG on retrieval accuracy — it changes what the agent is told currently governs.

What it does not replace

Bring your agent. Keep your stack.

Propperly is additive. It fits around the systems you already run, and it doesn't require you to swap any of them out.

Warehousekeep it.
BIkeep it.
Semantic layerkeep it.
Catalog / governancekeep it.
Model / agent platformyour choice.
Propperlythe reviewed conclusions those systems and agents build on.

We describe how Propperly fits alongside these systems; it doesn't claim a pre-built connector into every one of them.

Why analytics first

Analytics is where conclusions pile up.

Analytics teams accumulate exactly the kind of work this object is made of — years of it, spread across tools and people. An agent can often retrieve the artifacts; it still needs to know what currently governs.

metric definitionsanalysis conclusionsexceptionsscope decisionsrejected approachesdashboardsSQL & notebooksbusiness decisionsunresolved disagreements

See the Analytics solution — or start with one domain.

The Analytics solution holds the engagements, deliverables and evidence. A bounded reconstruction lets a team test the approach on one domain before a larger integration.

Read the research behind reliable AI analytics