Docs/Building with Celiq/Decision Trails

Decision Trails

Log the decisions you make from your data, link them to the metrics that drove them, and track how they turned out — an audit trail of why, not just what.

TracerWeaverLuminaryUpdated Jun 2026 · 6 min read

✦

In this section

Part of Celiq's semantic layer platform. Connect your warehouse, model your data once, query it everywhere.

💡

Tip

What it is: A running log of the decisions your team makes from data — each one linked to the Orion question, the answer, and the metric values that drove it.
Why use it: Build institutional memory of why you did things, not just what the numbers were. Track outcomes to learn what actually worked.
Where to find it: The Log decision button under any Orion answer, and the Decisions pill in the floating nav (/decisions).
Who can use it: All roles. Decisions are scoped to your workspace and never shared across workspaces.

Most BI tools tell you what the data said. Decision Trails capture what you did about it. After Orion answers a question, you can record the decision you made — and Celiq automatically attaches the question, Orion's answer, and the relevant metric values. Over time you accumulate a searchable, outcome-tracked history of your team's choices, grounded in the data that informed them.

What is a Decision Trail

A decision is a short record of a choice your team made, linked to the data behind it. Each decision stores:

A title — what you decided (e.g. "Launched re-engagement campaign")
Optional context — why you decided it
The Orion question you asked and Orion's answer
A metric snapshot — the numeric values from the response at the moment you decided
Optional links to the mosaic, reveal, and domain involved
Tags for grouping and filtering
An optional outcome, added later

Decisions are created from a real Orion conversation, so the trail is always anchored to governed data rather than a vague memory of "what the dashboard showed last quarter".

Logging a decision from Orion

Every Orion response has a subtle 📌 Log decision button, alongside the follow-up suggestions. Logging never blocks or changes Orion's answer — it's an additive action.

After Orion answers, click 📌 Log decision.
A quick-capture drawer opens.
Enter What did you decide? (required).
Optionally add Context explaining the reasoning.
Confirm the pre-attached details — Orion's answer and the metric values are already filled in for you.
Add Tags if you want (e.g. marketing, q2, inventory).
Click Log Decision.

Celiq pre-fills the drawer with:

Orion question — the question you asked
Orion answer — the text of Orion's response
Metric snapshot — any numeric values from the response

This means a decision is almost always one short sentence away — the data context comes for free.

The decisions timeline

Open the Decisions pill in the floating nav, or go to /decisions, to see the full history as a timeline, most recent first:

Jun 12, 2026
  📌 Launched email campaign
     "CAC was $44 → decided to launch re-engagement campaign"
     Context: Revenue was flat 2 weeks, CAC rising
     Orion said: "CAC increased 14% WoW, driven by paid search"
     [View Orion answer] [Add outcome] [Edit] [Delete]

Jun 8, 2026
  📌 Paused Footwear inventory orders
     "Return rate hit 11% — paused new POs"
     Outcome: ✅ Return rate dropped to 7% by Jun 15

Each entry shows the title, date, a context snippet, tags, and an outcome badge if one is set. Expand an entry to read the full Orion answer that informed it. You can filter the list by domain or tag.

Decisions can also be viewed in the context of a mosaic: when you log a decision while looking at a mosaic, it's linked to that mosaic so you can see every choice made off that dashboard.

Adding an outcome

The real value of a decision trail comes from closing the loop. Any decision can have an outcome added after the fact — click Add outcome on the card. Choose one of:

Outcome	Badge	Meaning
Worked as expected	✅	The decision had the intended effect.
Didn't work	❌	It didn't pan out.
Still monitoring	⏳	Too early to tell; checking back.
Changed course	🔄	We adjusted or reversed the decision.

The outcome appears as a colored badge on the card. You can revisit and update it as the situation evolves — a decision that starts as ⏳ Still monitoring can become ✅ Worked as expected weeks later.

Use cases

Board and investor prep — pull up exactly what you decided last quarter and how it turned out, with the supporting numbers attached.
Postmortems — when a metric moves, look back at the decisions tagged to that area and see which ones contributed.
Team alignment — a shared, durable record so new team members understand why the current strategy exists.
Learning what works — filter by outcome to see your hit rate on a class of decisions (e.g. pricing changes).

API

All decision endpoints are workspace-scoped and require authentication. Decisions are never visible across workspaces.

Method	Endpoint	Purpose
`POST`	`/api/decisions`	Create a decision linked to the current Orion conversation.
`GET`	`/api/decisions`	List decisions for the workspace. Supports `limit`, `offset`, `domain`, and `tag` query params.
`PATCH`	`/api/decisions/:id`	Update a decision — e.g. add an `outcome`, refine `context`, or change `tags`.
`DELETE`	`/api/decisions/:id`	Delete a decision.

Creating a decision accepts a body like:

json

{
  "title": "Launched re-engagement campaign",
  "context": "Revenue flat 2 weeks, CAC rising",
  "orionQuestion": "What's driving the change in CAC?",
  "orionAnswer": "CAC increased 14% WoW, driven by paid search",
  "metricSnapshot": { "cac": 44, "cac_change_pct": 14 },
  "mosaicId": "…",
  "revealName": "cac",
  "domainName": "marketing",
  "tags": ["marketing", "q2"]
}

To record how it turned out later:

json

PATCH /api/decisions/:id
{ "outcome": "worked" }

Limitations

Decisions are workspace-scoped. They are never shared or visible across workspaces.
The metric snapshot is a point-in-time capture. It records the values as they were when you logged the decision; it does not update if the underlying data later changes. That's intentional — it preserves the basis for the decision.
Logging is manual. Celiq pre-fills the data context, but it won't infer a decision for you — you decide what's worth recording.
Outcomes are self-reported. Celiq does not automatically verify that a decision "worked"; you mark the outcome based on your own judgment.