Delivering results to your warehouse
Push a published run's segment-labeled results into your own data platform via Delta Share, cloud storage, or sFTP.
Warehouse delivery is the production route for activation: Neuralift pushes a published run’s segment-labeled results directly into your data platform, where you join them to your customer table and feed every downstream system. An admin configures the destination once for the organization; after that, when a published run is ready to go out, delivery is run with your Neuralift team, and you monitor its status from the run’s Exports tab.
Before you begin
- Configuring delivery is an admin task; see Roles & permissions.
- The run you want to deliver is published. See Publishing.
- Delivery is available on contracted plans; see the exporting overview.
Choose a delivery method
The first two methods ride Delta Sharing, the open protocol for secure data sharing developed by Databricks (branded OpenSharing in Databricks’ documentation). It’s the same governed mechanism as inbound Databricks connections, running in the opposite direction. The other two write Parquet files to a destination you control — and on Google Cloud, the files are also registered as queryable BigQuery tables.
| Method | How results arrive | Best when |
|---|---|---|
| Delta Share (Databricks recipient) | A shared table readable from your Unity Catalog metastore | Your warehouse is Databricks |
| Delta Share (open protocol) | A share you register with a one-time activation URL | You’re on Snowflake or another open-protocol reader |
| Cloud-storage push (Preview) | Segment-labeled Parquet files copied to your S3 or GCS bucket; on GCS, also registered as BigQuery tables | You’re on Redshift, BigQuery, or run a lake-based stack; see Release stages |
| sFTP (Preview) | Parquet files written to a private gateway folder you retrieve over sFTP | You can’t use Delta Share or cloud federation |
What to provision in your own warehouse or cloud account for each method — service users, buckets, policies, and the SQL to mount or register what Neuralift shares — lives with each warehouse’s connection page: Databricks, Snowflake, BigQuery, and Redshift.
Configure delivery
Delivery is configured once per organization, in Settings → Data Connections, under the Share your results panel. Click Configure delivery to open the setup form, then follow the section below for your method.
Delta Share to Databricks
- Under Delivery method, choose Delta Share; under Recipient type, choose Databricks.
- Enter your Customer Databricks metastore ID, found in your Databricks workspace under Unity Catalog settings.
- Click Configure delivery.
The panel shows a setup spinner while the share is provisioned, then an Active badge with the share and recipient names. On your side, the share appears under Delta Sharing → Shared with me; mount it as a catalog following Receiving results in Databricks.
Note. The recipient is bound to the metastore ID you entered, and it can’t be edited afterwards. To move delivery to a different metastore, use Reset & re-provision on the panel.
Delta Share via open protocol (Snowflake)
- Under Delivery method, choose Delta Share; under Recipient type, choose Snowflake (open protocol).
- Click Configure delivery. A Delta Share activation URL appears in a dialog.
- Click Copy URL and store it securely, then use it to register the share in Snowflake as a catalog integration with a linked database — the exact SQL is in Receiving results in Snowflake. Once it’s registered, you should see the shared table on your side. (On another open-protocol reader, follow the standard Delta Sharing consumer flow for your platform.)
Warning. The activation URL is shown exactly once and cannot be retrieved later. If it’s lost, remove the delivery configuration and configure it again to generate a new one.
Cloud-storage push (Preview)
- Under Delivery method, choose Cloud storage.
- Enter the Storage destination: an S3 URI such as
s3://<bucket>/<prefix>, or a GCS URI such asgs://<bucket>/<prefix>. - Click Configure delivery.
What happens next depends on the cloud:
- S3. The panel shows a bucket policy JSON statement. Have whoever administers your AWS account add it to the bucket’s policy. It grants Neuralift’s dedicated delivery role the bucket-metadata and list access, plus the object write, delete, and multipart-upload operations it needs to manage deliveries under the configured prefix. Deliveries will fail until the policy is in place. Full detail, including loading delivered files into Redshift with
COPY, is in Receiving results in Redshift. - GCS. The panel shows the email of Neuralift’s delivery service account and the grant commands to run in your project. Create the bucket, dataset, and BigLake connection first, save the configuration, run the grants, then click Reset & re-provision — the full sequence is in Receiving results in BigQuery.
sFTP (Preview)
sFTP has its own sFTP delivery panel below the main configuration:
- Click Set up sFTP delivery.
- Paste your public key into the Customer SSH public key field (for example
ssh-ed25519 AAAA... user@host). Only the public half of the key pair is submitted; no password is stored. - Neuralift provisions a private gateway. When the panel shows Active, it lists the Host, gateway User, and your key’s fingerprint.
Connect with your matching private key using any sFTP client. Delivered results appear under the outbound/ folder. Use Update key to rotate to a new key at any time.
Deliver a run’s results
Once the destination is configured, delivery itself is run with your Neuralift team: when a published run is ready, they push its results to the configured share or storage destination. You follow along from the run’s Exports tab: open the published use case, select Exports in the sidebar, and watch the Result delivery section.
The status shows Sharing… or Delivering… while the push runs, then a Shared (Delta Share) or Delivered badge with the delivery date. You should now see the run’s segment-labeled results on your side: a queryable shared table in your metastore, Parquet files at your storage destination, or files in your sFTP outbound/ folder.
To push again (for example after a change on your side), ask your Neuralift team to re-share or re-deliver. For Delta Share, a run can also be unshared, which removes its data from the share so your platform can no longer read it.
What you receive
Delivered results are segment-labeled rows keyed by the same ID column as your prepared dataset, with each customer’s segment assignment, ready to join straight back to your customer table.
Troubleshooting
- The Exports tab says “Configure delivery in Org Settings to push results to your warehouse.” No active delivery configuration exists yet; an admin needs to complete the setup above.
- It says “Publish this run before delivering results.” Delivery only works on published runs; see Publishing.
- A Failed badge appears. For cloud-storage push, first confirm the bucket policy was added on your side, then ask your Neuralift team to retry the delivery. If it keeps failing, contact support@neuralift.ai.
- You can’t remove the Delta Share configuration. Shared runs are still using it. Ask your Neuralift team to unshare them first, then remove the configuration.
- You removed a cloud-storage configuration but files are still in the bucket. Removing the configuration stops future deliveries; it doesn’t recall files already delivered. Delete them from the bucket yourself if needed.
Next steps
- Join the delivered labels to your customer table, then compare notes with the segment landscape.
- For quick, one-off audience files alongside delivery, see CSV exports.