> ## Documentation Index
> Fetch the complete documentation index at: https://docs.blobrouter.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Storage Decision Engine

> How BlobRouter reasons about storage — decide(), callouts, segment economics, and when not to move data.

<Info>
  **BlobRouter v0.1** · Last updated: September 2026 · Architecture version: **2.0**
</Info>

BlobRouter’s **Storage Decision Engine** turns observed (and carefully labeled) workload signals into a structured **decision report** — not a black-box “AI said migrate.”

The engine’s job is to help you decide what to do with storage estates: move cold data, keep hot data where it is, or **explicitly do nothing** until a monitor trigger fires. Paid plans buy **continuous analysis** of that estate; they do not buy migrations. See [Plans & billing](/billing/plans).

## What a decision report contains

A completed audit may include an additive decision report alongside the legacy per-object waste recommendations. Typical surfaces:

| Surface                   | Meaning                                                             |
| ------------------------- | ------------------------------------------------------------------- |
| **Callouts**              | Highest-signal findings first (where money goes, what not to touch) |
| **Segment economics**     | Cost and risk by workload segment, not only by bucket               |
| **Scenarios**             | Comparable “what if” paths grounded in the same pricing inputs      |
| **DO NOT MOVE / monitor** | Honest non-moves with re-evaluation triggers                        |

Deterministic core logic (`decide()` in `@blobrouter/core`) produces the report. The dashboard renders it on the Storage Audit page when present.

## Live audit vs marketing demo

| Source                       | What you get                                                                                                                                                                           |
| ---------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Live AWS audit**           | Starts from observed S3 inventory via a read-only IAM role. Modeled assumptions (for example access patterns not fully measured yet) are labeled — they are not sold as measured fact. |
| **Marketing / product demo** | Uses a fully populated representative workload profile so you can click through a complete report. It is **not** a real customer’s data.                                               |

Do not treat demo dollars as a production guarantee for your account.

## How this fits the product

```text theme={null}
Free AWS audit (STS, read-only)
  → Legacy waste + ranked recommendations
  → Decision report (when available on the scan)
  → Optional: approve a migration
  → Optional: smart router for new uploads
  → Paid: continuous re-evaluation of the estate
```

Upload **routing** (where a new object goes) is a separate concept — see [Routing logic](/concepts/routing-logic). The decision engine is about the **estate you already have**.

## Related

* [How BlobRouter works](/concepts/how-it-works)
* [Architecture](/architecture)
* [REST API — scan results](/sdk/rest-api#scan-results)
* [Plans & billing](/billing/plans)