California DROP implementation · fixed scope · customer-owned

Own the pipe before the next cycle.

We turn your schema into a tested download → match → action → report pipeline in 48 hours. It runs in your cloud. Consumer identifiers never enter ours.

Successful sandbox dry run in 48 hours—or your build fee is refunded.
Aug 1processing began
45 daysmaximum access interval
6 listsofficial identifier types
0 identifiersconsumer values sent to DeleteRail
Current implementation basis: CalPrivacy DROP overview · broker process · technical specifications

Finished infrastructure, not advice

One narrow job. Completely built.

Broad privacy platforms span many workflows. DeleteRail is a fixed implementation for the new DROP cycle, built around your existing stack and handed to you as source code.

01 / RETRIEVAL

API client + CSV fallback

Authenticated download, retry handling, hash-chained intake evidence, and response upload using the current official interface.

02 / IDENTITY

Six deterministic standardizers

Email, phone, MAID, CTVID, name/DOB/ZIP, and name/VIN—normalized, hashed, and covered by test vectors.

03 / ACTION

Review-gated routing

Match evidence, exemptions queue, ambiguous-identity opt-out path, deletion adapters, and processor task exports.

04 / MEMORY

Ongoing suppression gate

Hash-only suppression state, official Removed-file reconciliation, and a repeatable screen for newly collected records.

05 / EVIDENCE

Status + evidence pack

Accepted-format status files, per-request disposition evidence, exception log, and 45/90-day operating runbook.

06 / OWNERSHIP

Your repository. Your cloud.

Dockerized source, deployment configuration, tests, runbook, and no annual license or custody of consumer records.

Founding-cycle scope

The expensive middle is included.

One source platform with up to three tables or resources, plus two downstream processor action manifests. Supported starting points: PostgreSQL, Snowflake, BigQuery, S3/CSV, and documented REST APIs.

  • Schema and identifier mapping
  • DROP sandbox configuration
  • Official-format download and upload
  • Hashing and composite-match tests
  • Human decision queue for exemptions
  • Processor action manifest
  • Suppression screening gate
  • Docker, environment template, and runbook

Data boundary by design

Your records stay where they are.

DeleteRail asks for field names and architecture—not consumer values, request lists, keys, or production credentials. The generated system performs every sensitive operation inside your environment.

No annual platform contract

Buy the implementation. Keep the code.

The founding cycle is deliberately fixed: one broker, one source platform, up to three resources, two processor manifests, one deployment-ready repository.

Founding build

$9,950

One-time. Paid upfront.

  • 48-hour repository delivery
  • Sandbox dry-run assurance
  • Unlimited factual corrections for 14 days
  • No consumer-data transfer
Reserve a founding build →

Larger scopes are quoted separately only after the founding cycle.

Questions a privacy or engineering lead will ask

Do you decide which records are legally exempt?

No. The implementation exposes an evidence-backed review queue. Your designated privacy or legal owner makes exemption decisions and approves production actions.

Does DeleteRail receive DROP request files or consumer identifiers?

No. The central site accepts only architecture and schema metadata. Request files, identifiers, API keys, and production records remain inside your environment.

What does “sandbox dry run” mean?

The delivered repository must execute its official-format download, standardization, hashing, match, and status-output test path against authorized sandbox fixtures. If it cannot within 48 hours of complete intake, the build fee is refundable.

Is this legal advice or a compliance certification?

No. DeleteRail implements an operational workflow against the technical requirements you select. Your company remains responsible for legal interpretation, exemptions, approvals, and compliance.

What if we already use a privacy platform?

The implementation can sit behind it or export actions into it. You keep the DROP-specific code and avoid replacing your existing privacy program.