Governed logistics decision infrastructure

Logistics decisions, made computable.

Raindex gives logistics teams and AI agents one governed environment to model possibilities, test strategy, and turn complex decision-making into infrastructure the business can trust. Keep the systems that run logistics. Govern the intelligence above them.

Model Compare Govern Decide

Explore the platform

01ThinkSpaceOrganize the initiative.An Account-scoped project for one logistics initiative and the studies it needs.
02WorkSpaceModel exact worlds.The analytical laboratory: distinct worlds, policy versions and their evidence, kept together.
03DecisionSpaceExpose the choices.For each shipment, expose every alternative supported by its exact inputs and rules.

01 / From entropy to governance

Messy logistics data is normal. Governed logistics reality is not.

Logistics evidence arrives in different schemas, grains, units, codes and vocabularies. Raindex governs the meaning instead of hard-coding around every source.

The Problem

Every source describes the same logistics world differently.

1.1Same concept. Different schema.

Your schema is not your logistics model.

// Source A
{
  "shipment_id": "7812",
  "svc_lvl": "2DA",
  "weight_lb": 12,
  "zone": "208"
}

// Source B
{
  "tracking_no": "7812",
  "service_code": "02",
  "weight_kg": 5.44,
  "zone_code": "8"
}

// Source C
{
  "shipment": "7812",
  "service": "2ND DAY AIR",
  "wt": "12 LBS",
  "dest_zone": "Z8"
}

Equivalent concepts arrive under different names.
Types and units change.
Required fields appear, disappear or move.
Every source-specific transformation becomes another thing to maintain.

Traditional ETL tends to encode the source’s quirks into code.

1.2Same-looking values. Different meaning.

Clean strings do not guarantee clean meaning.

Carrier	Service	Zone
UPS	Ground	"008"
UPS	2nd Day Air	"208"
FedEx	Ground	"8"
USPS	Ground Advantage	"8"

Zone 8, written four ways. "208" only means zone 8 once you know the carrier is UPS and the service is 2nd Day Air.

The same business concept may be encoded differently.
Similar values may not mean the same thing.
Carrier and service context can change interpretation.
Blind normalization can create clean-looking but economically wrong data.

Logistics governance has to understand context, not just syntax.

1.3A row is not always a Shipment.

Even the meaning of a row changes.

// Shipment-level
{
  "shipment_id": "7812",
  "service": "2DA",
  "weight_lb": 12,
  "total_charge": 14.47
}

// Charge lines · shipment 7812
[
  { "charge": "BASE", "amount": 8.40 },
  { "charge": "FUEL", "amount": 1.12 },
  { "charge": "RES",  "amount": 4.95 }
]

shipment_count: 12 One proved-equivalent record can represent twelve Shipments.

One row may be one Shipment.
Several source lines may describe one Shipment.
A proved-equivalent record may represent multiple Shipments.
Charges, package facts and observed economics keep their declared grain.

If grain is wrong, every downstream calculation can look precise and still be wrong.

1.4Economics arrive fragmented too.

The economics inherit the same fragmentation.

// Source A
base           8.40
fuel           1.12
residential    4.95

// Source B
transportation 14.47

// Source C
base           8.40
accessorials   [ … ]

12 LB · 5.44 KG · "12 LBS" The same weight, three ways.

Charges may be rolled up or itemized.
Units and dimensions vary.
Service definitions affect which charges are valid.
Contract structure determines how landed economics are evaluated.

Cleaning a column is not the same as governing an economic definition.

1.5And then the source changes again.

The integration is never really finished.

// March
- "svc_lvl"
+ "service_level"

// June
- "service_level"
+ "service"

// September
- "weight": 12
+ "weight": "12 LBS"

// Also this quarter
+ new column      "accessorial_code"
- missing column  "billed_weight"
~ changed type    "zone": integer → string

Schemas drift.
New carriers and systems appear.
Source owners change exports.
Exceptions accumulate.
Hard-coded fixes become permanent maintenance.

The more sources you add, the more brittle source-by-source logic becomes.

Raindex

The source can vary. The governed meaning should not.

2.1Resolve what is already known.

Start with certainty.

Source A · mapping revision 7

"shipment_id": → Shipment identity; ✓ exact
"svc_lvl": → Service Level; ✓ accepted mapping
"weight_lb": → Shipment Weight; ✓ declared unit
"zone": → ?; needs context

Exact declarations first.
Existing compatible mappings next.
Deterministic normalization where meaning is already governed.
No model call when Raindex already knows the answer.

2.2Escalate only what is ambiguous.

Unknown stays unknown until it is governed.

"zone": "208" · illustrative proposals

UPS 2nd Day Air zone 8likely
Same row carries carrier UPS and service 2DA
Customer region codepossible
Matches a customer-defined field pattern
Generic parcel zone 208unlikely
Outside common parcel zone ranges

Ambiguous fields are surfaced instead of silently guessed.
Ranked proposals narrow the review.
Reasons make each proposal inspectable.
People review; authorized agents can prepare proposals.
Acceptance is explicit.

2.3Preflight turns meaning into admissible evidence.

Governance becomes computation.

Preflight

Mapping: ✓
Grain: ✓
Units: ✓
Required facts: ✓
Network definitions: ✓
Unresolved: 0

map → normalize → validate → admit

Mapping resolves source meaning.
Normalization preserves exact grain and units.
Network validates governed logistics definitions.
Economic inputs can be evaluated consistently.
Unresolved evidence blocks the claims it cannot support.

Definitions become computational against the source.

2.4Remember the work.

Every accepted mapping reduces future friction.

// First delivery
"svc_lvl"
  → unresolved
  → reviewed
  → Service Level · revision 7

// Next compatible delivery
"svc_lvl"
  → revision 7 → deterministic

"accessorial_code"
  → new drift → review

Exact mapping revisions are retained.
Compatible future deliveries reuse them.
New drift is isolated instead of restarting the exercise.
Review becomes exception-driven rather than repetitive.

Raindex gets easier to operate as governed knowledge accumulates.

The Result

Governed evidence the rest of Raindex can reason over.

3.1Clean and governed.

Clean data is useful. Governed data is reusable.

Shipment SHP-7812

Identity

Shipment: SHP-7812
Represented Shipments: 1

Service

Carrier: UPS
Service: 2nd Day Air
Zone: 8 · native “208”

Measurement

Weight: 12 lb

Evidence

Source revision: 18
Mapping revision: 7

Consistent identities.
Explicit grain.
Exact units.
Accepted meaning.
Retained lineage and versioned mappings.

The next team, agent or WorkSpace does not have to rediscover what the record means.

3.2Computable logistics reality.

Clean data is the beginning. Computable logistics reality is the result.

Governed evidence

Network

Carrier
Carrier Account
Contract
Service
Facility
Geography
Charges
Business Attributes
Intelligence

Computable

Landed economics
Supported options
Exact analytical inputs

Calculate governed Contract Charges.
Resolve landed economics.
Evaluate Business Attributes and Intelligence.
Determine which options the evidence supports.
Build exact analytical worlds from the result.

Govern the reality once. Then build worlds from it.

02 / WorkSpace / exact analytical worlds

Model the world. Then change the assumptions.

Organize an initiative in ThinkSpace. Keep its studies in WorkSpace, with distinct Roots for different worlds and Versions for different selection policies.

Inside WorkSpace

Explore WorkSpace

Exact worlds. Inspectable choices.

Worlds & Versions

Keep the baseline. Change a world with a new Root, or its selection policy with a new Version.

A question or a chart changes neither Root nor Version.

ThinkSpace / initiative
WorkSpace / analytical study
Root A → A1 baseline · A2 revised policy
Root B → B1 alternate world

03 / Rulebooks / readable source

Make the logic as inspectable as the data.

Write the logic. Preserve its exact source and revision. Bind it explicitly to the owner that uses it.

selection.sqlRAK / illustrative source

01RULE CASE WHEN TRUE
02THEN (
03  SELECT option.option_id
04  FROM decision_space AS option
05  ORDER BY option.postoverlay_rate ASC
06  LIMIT 1
07) END

Current RAK syntax. postoverlay_rate is the exact technical field for the evaluated Contract Charges result.

Exact UTF-8 source · revision · hashExplicit Save / Revise

Consumer binding / conceptual

Source travels.
Authority stays with the owner.

Bind: Exact source revision and hash
Resolve: Exact Coordinate, parameters and dependencies
Review: Fresh owner Preview and required confirmation
Effect: Owner Apply → receipt → verified state

Validation does not Save. Source publication does not activate a consumer. RAK selects an existing option or returns no-decision.

SelectionChargesConditionsGoalsRoot constructionAnalysisWorkFlow source

Explore Rulebooks

04 / DecisionSpace / the inspectable option field

Materialize every choice the governed world can support.

For one Shipment at one exact Coordinate, preserve the supported alternatives, economics and comparison bases behind the selection.

Shipment SHP-DEMO-001

WorkSpace · Root A · Version 2 · Coordinate A2

Illustrative values / USD
Not product data or a quote

Inspect a rank basis

The basis changes emphasis. All five options and the policy selection stay fixed.

Scroll across to inspect economics, ranks and selection →

Five illustrative options; OPT-03 remains policy-selected for every comparison basis.
Option / Carrier · Service	Base	Charges	Landed	Transit	Intelligence	Composite	Basis rank	Policy
OPT-01Carrier A · Ground	8.00	2.40	10.40	4 days	81	2.10	#2	Retained
OPT-02Carrier A · Two-day	13.00	2.60	15.60	2 days	88	1.70	#4	Retained
OPT-03Carrier B · Ground	7.50	2.00	9.50	3 days	92	1.40	#1	Selected
OPT-04Carrier C · Ground	9.00	2.20	11.20	4 days	76	2.80	#3	Retained
OPT-05Carrier B · Next day	20.00	3.10	23.10	1 day	92	2.90	#5	Retained

Ranks have a basis. Intelligence and Composite values are illustrative, with support dependent on the governed inputs. No score is a universal definition of best.

Inspect the selection evidence

In this example, A2 selects OPT-03 by lowest landed cost. Product evidence pins the exact Contract revision, Charge Breakdown, Rulebook revision and selection receipt. This illustrative table supplies none of those receipts.

When policy cannot select an eligible existing option, the result is explicit no-decision.

05 / Goal Search → Raindex Research

Search the governed space. Research the frontier.

An exact finite world turns a logistics question into a computational problem. Goal Search frames the product path; Research explores the harder frontier.

Platform / classical Goal Search

Define what should improve.
Protect what must hold.

Goal: Reduce modeled cost
Protected: Service threshold · eligibility · option population
Allowed change: Selection logic

01Qualify
02Search
03Replay
04Review
05Publish

Explore the WorkSpace model

06 / People + agents / one governed world

Give AI a governed world to reason over.

A force multiplier needs exact context and bounded capabilities. People and agents work against the same world, permissions and owner contracts.

A request with a reviewable result.

Inspect exact context

Read the world, policy and retained evidence.

Launch bounded work

Use supported owner capabilities with authorization.

Review before effect

Prepare intent and inspect a fresh Preview.

Verify the result

Follow receipts to the resulting state.

Meet Raindex Agent Build with API and MCP

07 / Enterprise context / who it serves

A decision layer between your logistics evidence and execution systems.

Different teams inspect different parts of it. They work from the same governed model.

01 / Systems + evidence

The evidence you already have.

Contracts · TMS / WMS / ERP · Shipment feeds · operational files

↓ Evidence admission

02 / Raindex decision layer

Alongside your systems.

Sources
Network
WorkSpace
Rulebooks
Streams
Agent / Developer interfaces

Govern definitions, inspect alternatives and preserve decision evidence.

↓ Customer pull / read

03 / Customer execution systems

Your systems carry out the work.

TMS · WMS · ERP · carriers / downstream automation

Logistics analysts

Can I compare exact worlds and preserve the alternatives?

Operations leaders

Can I consume governed decisions alongside my execution systems?

Finance & procurement

Can I inspect the Contract basis, Charges and assumptions?

IT & security

Can I verify Account scope, interfaces, retention and authority?

Developers & agent builders

Can my applications use the same exact capabilities?

Executive buyers

Can this become reusable infrastructure for the next decision?

08 / A focused evaluation

Bring one difficult logistics decision.

Start with the question. We can map the evidence, governed definitions, alternatives and analytical scope needed to make it inspectable.

01
The question
What are you trying to decide, compare or explain?
02
The evidence
What currently informs that decision?
03
The useful result
What does your team need to inspect or prove?

Read Product Docs

Logistics decisions, made computable.

Model Compare Govern Decide

Carrier

Service

Zone

UPS

Ground

"008"

UPS

2nd Day Air

"208"

FedEx

Ground

"8"

USPS

Ground Advantage

"8"

Five illustrative options; OPT-03 remains policy-selected for every comparison basis.

Option / Carrier · Service

Base

Charges

Landed

Transit

Intelligence

Composite

Basis rank

Policy

OPT-01Carrier A · Ground

8.00

2.40

10.40

4 days

2.10

Retained

OPT-02Carrier A · Two-day

13.00

2.60

15.60

2 days

1.70

Retained

OPT-03Carrier B · Ground

7.50

2.00

9.50

3 days

1.40

Selected

OPT-04Carrier C · Ground

9.00

2.20

11.20

4 days

2.80

Retained

OPT-05Carrier B · Next day

20.00

3.10

23.10

1 day

2.90

Retained

Messy logistics data is normal. Governed logistics reality is not.

Your schema is not your logistics model.

Clean strings do not guarantee clean meaning.

Even the meaning of a row changes.

The economics inherit the same fragmentation.

The integration is never really finished.

Start with certainty.

Unknown stays unknown until it is governed.

Governance becomes computation.

Every accepted mapping reduces future friction.

Clean data is useful. Governed data is reusable.

Clean data is the beginning. Computable logistics reality is the result.

Model the world. Then change the assumptions.

Worlds & Versions

Make the logic as inspectable as the data.

Source travels.Authority stays with the owner.

Materialize every choice the governed world can support.

Search the governed space. Research the frontier.

Define what should improve.Protect what must hold.

Give AI a governed world to reason over.

Inspect exact context

Launch bounded work

Review before effect

Verify the result

A decision layer between your logistics evidence and execution systems.

The evidence you already have.

Alongside your systems.

Your systems carry out the work.

Logistics analysts

Operations leaders

Finance & procurement

IT & security

Developers & agent builders

Executive buyers

Bring one difficult logistics decision.

The question

The evidence

The useful result

Messy logistics data is normal. Governed logistics reality is not.

Your schema is not your logistics model.

Clean strings do not guarantee clean meaning.

Even the meaning of a row changes.

The economics inherit the same fragmentation.

The integration is never really finished.

Start with certainty.

Unknown stays unknown until it is governed.

Governance becomes computation.

Every accepted mapping reduces future friction.

Clean data is useful. Governed data is reusable.

Clean data is the beginning. Computable logistics reality is the result.

Model the world. Then change the assumptions.

Worlds & Versions

Make the logic as inspectable as the data.

Source travels.Authority stays with the owner.

Materialize every choice the governed world can support.

Search the governed space. Research the frontier.

Define what should improve.Protect what must hold.

Give AI a governed world to reason over.

Inspect exact context

Launch bounded work

Review before effect

Verify the result

A decision layer between your logistics evidence and execution systems.

The evidence you already have.

Alongside your systems.

Your systems carry out the work.

Logistics analysts

Operations leaders

Finance & procurement

IT & security

Developers & agent builders

Executive buyers

Bring one difficult logistics decision.

The question

The evidence

The useful result

Source travels.
Authority stays with the owner.

Define what should improve.
Protect what must hold.

Source travels.
Authority stays with the owner.

Define what should improve.
Protect what must hold.