---
title: Logistics and Supply Chain on AWS (2026): Visibility, Fleet Tracking, and Planning Tiers
description: For a mid-market 3PL (~2.1M shipments/yr, OTIF 82%), layering AWS Supply Chain on ERP ingest plus Location Services fleet tracking moved ETA accuracy from 71% to 87% in 90 days — without replacing Manhattan WMS.
url: https://www.factualminds.com/blog/aws-logistics-supply-chain-reference-architecture-2026/
datePublished: 2026-07-02T00:00:00.000Z
dateModified: 2026-07-02T00:00:00.000Z
author: palaniappan-p
category: Cloud Architecture
tags: aws, supply-chain, logistics, iot, location-services, eventbridge, data-analytics, architecture
---

# Logistics and Supply Chain on AWS (2026): Visibility, Fleet Tracking, and Planning Tiers

> For a mid-market 3PL (~2.1M shipments/yr, OTIF 82%), layering AWS Supply Chain on ERP ingest plus Location Services fleet tracking moved ETA accuracy from 71% to 87% in 90 days — without replacing Manhattan WMS.

**AWS Supply Chain** added **N-Tier Visibility** — trading-partner onboarding, shared purchase orders, and in-app collaboration — extending upstream beyond your four walls. The **Well-Architected Supply Chain Lens** (2025–2026) codifies shipment tracking, fleet visibility, and resilience practices that previously lived only in logistics SI playbooks.

This post is the **logistics and supply chain reference architecture** — visibility tier, planning tier, and integration patterns on AWS. It is **not** [manufacturing OT/IT](/industries/aws-manufacturing/), **not** [retail peak traffic](/blog/aws-retail-architecture-black-friday-peak-traffic/), **not** [media OTT delivery](/blog/aws-media-ott-streaming-architecture-live-vod-2026/), and **not** [generic event throughput](/blog/aws-high-throughput-event-processing-tier-selection-2026/) (though high-volume scan events may need that guide).

Artifacts: [architecture decision matrix](https://www.factualminds.com/examples/architecture-blog-2026/logistics-supply-chain/architecture-decision-matrix.md), [KPI baseline worksheet CSV](https://www.factualminds.com/examples/architecture-blog-2026/logistics-supply-chain/kpi-baseline-worksheet.csv).

> **Benchmark pattern (not a cited client)** — Mid-market **3PL**, **~2.1M shipments/year**, **14 DCs**, Manhattan WMS + legacy TMS APIs, **OTIF 82%**, **ETA accuracy 71%**. Phase 1: nightly ERP/WMS → S3 → **AWS Supply Chain** ingest (**visibility latency 180 min → 22 min**). Phase 2: **Amazon Location** + driver mobile on **500** owned routes (**ETA accuracy 71% → 87%** in **90 days**). WMS unchanged.

## Three tiers — pick one primary pain

| Tier                  | Question                                | Default AWS path                                   |
| --------------------- | --------------------------------------- | -------------------------------------------------- |
| **Visibility**        | Where is inventory and inbound freight? | **AWS Supply Chain** SCDL + risk map               |
| **Execution signals** | Where are trucks and last-mile stops?   | **IoT Core** + **Location Services** + API Gateway |
| **Planning**          | Will we stock out in 14 days?           | AWS Supply Chain ML lead-time + watchlists         |

**Opinionated take:** **Buy visibility before building a data lake.** Teams that start with Kafka + Flink for logistics usually have not fixed **OTIF baselines** — see the [KPI worksheet](https://www.factualminds.com/examples/architecture-blog-2026/logistics-supply-chain/kpi-baseline-worksheet.csv).

## Reference architecture

```
ERP (SAP/Oracle) ──┐
WMS (Manhattan)  ──┼──► Connectors / S3 landing ──► AWS Supply Chain (SCDL)
TMS APIs         ──┘         │                           │
                             │                           ├── Risk map / insights
Telematics/ELD ──► IoT Core ─┼──► Timestream / Kinesis   ├── N-Tier partner chat
Mobile driver app ─► Location Services                   └── Export to S3 → ERP
                             │
                             └──► EventBridge ──► Lambda (exception workflows)
                                       │
                                       └──► QuickSight (OTIF, fill rate, ETA)
```

### AWS Supply Chain layer

AWS Supply Chain sits **above** operational systems — it does not replace WMS picking logic. Ingest ERP/WMS/TMS on a schedule or stream; the application contextualizes inventory health and surfaces stock-out / overstock risks with ML lead-time projections (per AWS Supply Chain product documentation).

Use when:

- Multiple DCs and suppliers create **conflicting inventory snapshots**
- Planners spend days merging spreadsheets before S&OP meetings
- You need **partner-facing** PO status without building a portal from scratch

### Fleet and last-mile layer

For **owned fleet** or contracted carriers with mobile apps:

- **Amazon Location Service** — maps, geofences, route segments
- **AWS IoT Core** — MQTT from telematics or phone location streams
- **API Gateway + Lambda** — appointment scheduling, exception webhooks
- **EventBridge** — fan-out to customer notification SNS/email

Well-Architected Supply Chain Lens **SCREL02-BP01** recommends IoT + logistics API integration for end-to-end shipment visibility — align geofence entry/exit to automated workflows (delay notifications, dock reassignment).

> **What broke** — Week 5 of a TMS API polling pipeline. Carrier rate limit (**120 req/min**) caused **4-hour** position gaps; customer ETA widget froze. Fix: switch high-value lanes to **IoT MQTT** telematics; keep API poll for long-tail carriers at **15-min** cadence. **ETA accuracy recovered 61% → 84%** on instrumented lanes only.

## Integration patterns — latency vs complexity

| Source              | Pattern                   | Freshness      | Pitfall                    |
| ------------------- | ------------------------- | -------------- | -------------------------- |
| ERP nightly extract | S3 + Glue → Supply Chain  | Hours          | Silent extract failure     |
| WMS event stream    | Kinesis / EventBridge     | Minutes        | Duplicate location IDs     |
| TMS webhook         | API Gateway → EventBridge | Near real-time | No retry/idempotency       |
| ELD/telematics      | IoT Core rules            | Seconds        | Cert expiry at 2k+ devices |

For event volume above **~8k TPS** on scan streams, read [high-throughput tier selection](/blog/aws-high-throughput-event-processing-tier-selection-2026/) before FIFO or shard mis-sizing.

## KPIs — measure before tooling

| Metric                 | Why it matters                                           |
| ---------------------- | -------------------------------------------------------- |
| **OTIF**               | Customer-facing outcome — not "data lake row count"      |
| **Days of supply**     | Ties inventory $ to planning quality                     |
| **ETA accuracy**       | Validates fleet layer ROI                                |
| **Visibility latency** | Minutes from scan to dashboard — exposes pipeline health |

Model targets in the [KPI worksheet](https://www.factualminds.com/examples/architecture-blog-2026/logistics-supply-chain/kpi-baseline-worksheet.csv) before RFPing AWS Supply Chain or IoT vendors.

## When NOT to escalate architecture

| Situation                     | Stay lighter                            |
| ----------------------------- | --------------------------------------- |
| Single DC, < 10k SKUs      | Athena + QuickSight on nightly extracts |
| No owned fleet                | TMS vendor ETA only                     |
| Partners refuse cloud sharing | EDI/SFTP batch; defer N-Tier            |
| Cold-chain not in scope       | Skip Greengrass sensor path             |

## What to do this week

1. Export **90-day OTIF, fill rate, ETA accuracy** — [worksheet CSV](https://www.factualminds.com/examples/architecture-blog-2026/logistics-supply-chain/kpi-baseline-worksheet.csv).
2. Inventory ERP/WMS/TMS extract paths; alarm on **stale snapshot > 2 hours**.
3. Run [decision matrix](https://www.factualminds.com/examples/architecture-blog-2026/logistics-supply-chain/architecture-decision-matrix.md) for visibility vs fleet vs planning.
4. Pilot **one DC + one carrier lane** on Location + IoT before enterprise mobile rollout.
5. If multi-supplier risk is the pain, evaluate **AWS Supply Chain** POC with top **20%** spend suppliers.

> **Reproduce this** — Download [kpi-baseline-worksheet.csv](https://www.factualminds.com/examples/architecture-blog-2026/logistics-supply-chain/kpi-baseline-worksheet.csv). Fill `baseline_before` from your TMS/WMS exports. Set `month_3_target` at **+5–8 pts** OTIF only if visibility latency is under **30 minutes** — otherwise fix pipeline first.

## What this post doesn't cover

- **Warehouse robotics / AMR** — WMS vendor domain.
- **Customs and cross-border trade compliance** — legal and broker integrations.
- **Manufacturing plant OEE** — [manufacturing industry hub](/industries/aws-manufacturing/).
- **Full transportation management system (TMS) replacement** — AWS complements, does not replace.

**Related:** [Data analytics on AWS](/services/aws-data-analytics/) · [Managed services](/services/aws-managed-services/) · [Event-driven messaging](/blog/aws-event-driven-async-messaging-boundaries/)

## FAQ

### When should we use AWS Supply Chain vs a custom data lake?
Use AWS Supply Chain when you need enterprise-wide inventory risk maps, ML lead-time predictions, and N-Tier partner visibility across multiple ERP/WMS/TMS sources. Build Glue + Athena + QuickSight when you have a single ERP, one region, and under ~10k SKUs — a nightly S3 extract may be sufficient without the application license.

### When should we NOT deploy real-time fleet tracking?
Skip IoT + Location for fleets under ~50 daily routes if your TMS vendor API already provides adequate ETA webhooks. Real-time tracking adds device cert lifecycle, mobile app maintenance, and geofence tuning — overhead that small operators cannot operationalize.

### What breaks when ERP extracts fail silently?
Inventory health maps show stale green donuts while DCs are stock-out. Symptom: OTIF drops before dashboard alerts fire. Fix: EventBridge rules on failed Glue/connector jobs, row-count anomaly alarms, and SLA on extract freshness (e.g., alert if WMS snapshot older than 2 hours).

### How does this differ from manufacturing IoT on AWS?
Manufacturing posts focus on OT/IT convergence, PLCs, and OEE on the plant floor. Logistics focuses on shipment movement, partner networks, and last-mile ETA — different data sources (TMS/telematics vs OPC-UA), different KPIs (OTIF vs OEE). See the manufacturing industry hub for plant patterns.

### Can AWS Supply Chain replace our WMS?
No. AWS Supply Chain is an intelligence layer on top of ERP, WMS, TMS, and OMS — it ingests data into a supply chain data lake (SCDL) and surfaces risk insights. Execution (picking, putaway, slotting) stays in Manhattan, Blue Yonder, SAP EWM, or similar.

### What could go wrong with partner N-Tier visibility?
Suppliers refuse data sharing, purchase order formats do not match, or chat-based collaboration bypasses audit trails. Mitigate with phased onboarding (top 20% of spend suppliers first), standardized PO export to S3, and written data-sharing agreements before enabling N-Tier invites.

---

*Source: https://www.factualminds.com/blog/aws-logistics-supply-chain-reference-architecture-2026/*