How Do You Migrate from AWS Lookout for Equipment Before October 2026?

When Does AWS Lookout for Equipment Shut Down?

AWS Lookout for Equipment stops accepting new customers on October 7, 2025 and will be fully retired on October 7, 2026. Existing customers can continue using the service until the shutdown date, but no new models can be created after the new-customer cutoff.

AWS announced this retirement as part of a broader consolidation of their industrial ML services. The recommended migration paths from AWS are Amazon IoT SiteWise (which added native anomaly detection in July 2025) and Amazon SageMaker (for teams that want to build custom models).

Both AWS-recommended paths require significant re-architecture. SiteWise requires adopting the full asset modeling framework. SageMaker requires ML engineering expertise and ongoing model management. Neither provides a simple API migration.

What Are Your Migration Options from AWS Lookout?

You have three practical options: migrate to Amazon SiteWise, build custom models in SageMaker, or switch to a standalone anomaly detection API like Canary Edge. Each option involves different levels of effort and lock-in.

Amazon SiteWise is the closest in-ecosystem replacement. It added anomaly detection in July 2025 specifically to absorb Lookout customers. However, it requires you to model your equipment as SiteWise assets with defined properties and hierarchies before anomaly detection works.

Amazon SageMaker gives you full control but requires your team to select, train, and deploy ML models. You take on all the MLOps burden that Lookout previously handled.

Canary Edge is a standalone REST API. You send JSON time series data, you get anomaly scores back. No asset modeling, no ML expertise, no cloud lock-in.

How Does Lookout's Architecture Differ from Canary Edge?

AWS Lookout for Equipment uses an asynchronous, S3-based workflow where you upload CSV files to an S3 bucket and wait for batch results. Canary Edge uses a synchronous REST API that returns results in a single HTTP response within 50 milliseconds.

Lookout's workflow: 1. Upload training CSV to S3 (minimum 6 months of historical data) 2. Create a model via CreateModel API — training takes 30 minutes to 24 hours 3. Start an inference scheduler via CreateInferenceScheduler 4. Upload new data to S3 on a schedule 5. Poll for results via ListInferenceEvents

Canary Edge's workflow: 1. Send baseline data via POST /v1/baseline/multivariate — training completes in seconds 2. Send new readings via POST /v1/detect/multivariate — results return in the same HTTP response

The fundamental architectural difference is that Lookout was designed for batch processing on hourly or daily cadences, while Canary Edge is designed for real-time detection on sub-second data. Canary Edge uses a LeWM (Latent Embedding World Model) architecture that learns machine behavior in a compressed 192-dimensional latent space, enabling fast inference without the multi-hour training cycles.

How Do You Export Your Data from AWS Lookout?

Your training data and inference results are already in S3, so export is straightforward. Download your original training CSVs and any inference output files from your Lookout S3 buckets before the October 2026 deadline.

Step 1: Identify your S3 buckets. In the AWS Console, go to Lookout for Equipment, open each model, and note the S3 URI for both training data and inference output.

Step 2: Download training data. Use the AWS CLI:

aws s3 sync s3://your-lookout-training-bucket/dataset/ ./lookout-export/training/

Step 3: Download inference history. These results contain your historical anomaly detections:

aws s3 sync s3://your-lookout-inference-bucket/ ./lookout-export/inference/

Step 4: Export model metadata. Save your model configurations using the Lookout API:

aws lookoutequipment list-models --output json > models.json
aws lookoutequipment list-datasets --output json > datasets.json

Keep these exports even after migration. They serve as an audit trail and can be used to validate that your new system produces comparable detection results.

How Do Lookout API Calls Map to Canary Edge?

The API mapping is not one-to-one because Lookout uses an asynchronous model management paradigm while Canary Edge uses synchronous request/response. However, every Lookout capability has a Canary Edge equivalent.

The biggest conceptual change is eliminating the scheduler. With Lookout, you set up an inference scheduler that periodically reads from S3. With Canary Edge, you call the detect endpoint whenever you have new data — from a cron job, a Lambda function, or directly from your application code.

What Does a 30-Day Migration Timeline Look Like?

A complete migration from AWS Lookout to Canary Edge can be executed in 30 days or fewer. Most of the time is spent on parallel testing rather than development work.

Week 1: Setup and baseline creation - Day 1-2: Sign up at canaryedge.com, generate an API key - Day 3-5: Export healthy operating data from your Lookout S3 buckets - Day 5-7: Create baselines on Canary Edge using POST /v1/baseline/multivariate for each machine

Week 2: Integration development - Day 8-10: Replace Lookout SDK calls with Canary Edge REST calls (typically 10-20 lines of code per integration point) - Day 11-14: Update your data pipeline to send JSON payloads to Canary Edge instead of uploading CSVs to S3

Week 3: Parallel testing - Day 15-21: Run both Lookout and Canary Edge simultaneously, comparing detection results - Verify regime classifications match your operational expectations - Tune sensitivity settings if needed (Canary Edge uses 0-99 sensitivity scale)

Week 4: Cutover - Day 22-25: Route production traffic to Canary Edge - Day 26-28: Decommission Lookout inference schedulers - Day 29-30: Clean up S3 buckets and IAM roles

Cost comparison: Lookout for Equipment pricing is based on inference hours and data ingested. Canary Edge pricing is based on data points processed. The free tier includes 1,000,000 points per month and up to 5 machines, which covers most pilot deployments.