Messaging data,
exported on schedule to your S3 bucket.
Compressed Parquet files, Hive-partitioned by date and hour, written to a bucket your team owns. AWS S3 natively, or any S3-compatible store: Backblaze B2, Cloudflare R2, Wasabi, DigitalOcean Spaces, MinIO.
THE BEFORE
Your messaging data shouldn’t live in a vendor’s UI.
Every report your team builds, every model your data scientists train, every audit your compliance lead runs depends on raw events. Not screenshots. Not CSV exports. Continuous, schema-stable, in your warehouse.
Everything your data team expects from a real warehouse pipeline.
Three guarantees. Each one removes a question your engineering review will ask.
01 / SCOPED ACCESS
Your bucket. Your key. Your control.
You create an access key scoped to one bucket with a write-only policy. We encrypt the secret at rest, bind it to that destination, and validate it with a probe write. Revoke the key any time.
02 / STABLE DATA SHAPES
Versioned. Documented. No glue code.
Stable, versioned data shapes under a v1 prefix. Same fields next quarter as this quarter. Backwards-compatible additions only, so your pipelines never break on a release.
03 / SCHEDULE AND SCOPE
Hourly, daily, or on demand. Incremental or full.
Hourly, daily at 08:00 UTC, or on demand. Each scheduled run picks up exactly where the last one ended. A chunked full export backfills up to two years of history.
Eighteen data sets. Stable. Documented. Versioned.
The grain matches how data teams actually model messaging. A new destination starts with the four core texting data sets, and you tick the rest.
Texting
sends, replies, opt-outs, clicks and survey answers
Every text with status, cost and carrier. Every project in any status with its current totals. Opt-outs, including when someone texts START. Link clicks with device and bot flags. Survey answers with the option they matched.
- message_events
- project_summary
- opt_out_events
- click_events
- survey_responses
every recipient, click, campaign and suppression
Each email to each recipient with delivery status, engagement and cost. Link clicks with bot flags. Campaigns with their live counters. The compliance list of addresses you can no longer email.
- email_messages
- email_clicks
- email_campaigns
- email_suppressions
People
everyone you have reached, and how they are tagged
Everyone you have texted, emailed or tagged, with name and brands, and every tag applied to a phone number or address, with when it was applied.
- people
- contact_tags
Money
gifts and charges
Donations from your connected fundraising pages, with refunds and source codes, and every charge, payment, refund and adjustment on your account balance.
- donations
- billing_ledger
Reference lists
the names behind the ids, complete on every run
Tags, contact lists, brands, campaigns and organizations, so every id in the other data sets resolves to a name without a lookup outside the warehouse.
- tags
- contact_lists
- brands
- campaigns
- organizations
Backwards-compatible additions only. Your downstream queries never break on a release. A destination limited to specific brands does not receive the data sets that belong to the whole organization: email, people, donations and the billing ledger.
Same outcome. A real pipeline instead of a spreadsheet.
What used to be a weekly export ritual is one connection and a schedule.
Manual pulls. Stale data. Format drift.
- 1Log in to vendor UI
- 2Export CSVs by hand
- 3Reconcile schema drift
- 4Re-import to the warehouse
Lost weeks. Audit gaps. Decisions on yesterday’s data.
Continuous. Schema-stable. Auditable.
- 1Add a scoped access key once
- 2Pick data sets and a schedule
- 3Query in the warehouse
One scoped access key. Hourly refresh. Parquet in your bucket. Done in an afternoon.
Snowflake direct. Every other engine from one bucket.
Snowflake is a destination of its own and loads straight into your tables. Every other engine reads the standard Parquet files in Hive-partitioned paths from an S3-compatible bucket. No connector to install on our side.
Snowflake
Loaded straight into tables in your account, kept current. No bucket.
Amazon Athena and Glue
Crawl the prefix or declare a partitioned table.
Databricks
Auto Loader on the prefix, landing Delta tables.
BigQuery
An external table over the Hive-partitioned path.
DuckDB
read_parquet with hive_partitioning straight from S3.
Answered before you have to.
This is the defining platform. See how it's measured.
Every message delivered. At a known cost. On demand. At any scale.
