Migration Guide

The migration importer takes an export from your current attribution platform and loads it into Measura as historical events, installs and attribution records. Your cohort charts and channel breakdowns then include the period before you switched.

Imported attributions are preserved as the source reported them. We do not re-run our matching over historical data, because the click level signals needed to do that are not in any platform’s export. Imported records carry a confidence score of zero and a written reason recording where they came from, so they are never confused with attributions Measura resolved itself.

File requirements

Larger histories are imported as a sequence of uploads. Split by date range so a failed batch is easy to identify and re-run.

Which columns to include

Export everything your platform offers. Unmapped columns are ignored rather than rejected, so a wider export costs nothing. At minimum, include the event timestamp, the event name, and at least one device identifier.

Use the raw data export. These columns are recognised automatically.

Branch exports carry no dedicated install time column, so the event timestamp is used for install rows.

Choosing custom applies no column translation. Your file must already use Measura field names.

[
  {
    "event_timestamp": "2026-05-14T08:31:02Z",
    "installed_at": "2026-05-14T08:31:02Z",
    "event_type": "install",
    "gaid": "38400000-8cf0-11bd-b23e-10b96e40000d",
    "channel": "whatsapp",
    "campaign_name": "lagos-launch-may",
    "attribution_model": "deterministic",
    "revenue": 0,
    "currency": "USD",
    "device_model": "Samsung SM-A245F",
    "os_name": "android",
    "os_version": "14",
    "app_version": "2.1.0"
  }
]

This is also the route for a platform we do not map natively. Rename the columns in your export tooling, then import as custom.

Channels

Source network names are translated to Measura channels. Anything unrecognised is lowercased with spaces replaced by underscores and carried through as is.

Event types

Attribution models

The developer dashboard provides this flow under Migrate. To automate it, call the three endpoints directly.

1. 1.Create the job. Returns a job identifier and the column mapping that will be applied, so you can confirm the mapping before committing the data.
2. 2.Upload the file. Processing begins immediately, in batches of 100 rows.
3. 3.Poll for status. Reports rows processed, a percentage, and any row level errors.

# 1. Create the job
curl -X POST "$MEASURA_HOST/functions/v1/migration-importer/start" \
  -H "Authorization: Bearer $SESSION_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"app_id":"7d2e5b1a-...","source_platform":"appsflyer"}'

# → {"job_id":"c4a9...","column_mapping":{...}}

# 2. Upload the export
curl -X POST "$MEASURA_HOST/functions/v1/migration-importer/upload?job_id=c4a9..." \
  -H "Authorization: Bearer $SESSION_TOKEN" \
  -H "Content-Type: text/csv" \
  --data-binary @appsflyer-export.csv

# → {"job_id":"c4a9...","total_rows":24500,"status":"processing"}

# 3. Poll until complete
curl "$MEASURA_HOST/functions/v1/migration-importer/status?job_id=c4a9..." \
  -H "Authorization: Bearer $SESSION_TOKEN"

# → {"status":"completed","processed_rows":24500,"progress_pct":100,"errors":[]}

Full request and response detail is in the API Reference.

Success and failure

A job completes when fewer than five percent of rows produced errors. At or above that threshold the job is marked failed, on the basis that a systematic problem such as a wrong platform selection or a malformed file should not be quietly half applied.

Rows that error are listed individually in the status response with the reason, so you can correct and re-upload just those.

What imported records look like

• Marked as already processed, so they do not enter the live attribution queue.
• A confidence score of 0 and a reason recording the source platform. Imported history is never mixed with scores Measura computed.
• Postback status skipped, so importing history never fires conversion notifications to advertising networks.
• A user agent of migration-import/{platform} and a fraud score of zero.
• An idempotency key derived from the job and row position, so re-uploading the same file into the same job does not duplicate rows.

This guide is mostly about coming in, so it is worth being just as clear about going out. You can export everything Measura holds for your account at any time, without asking us, and without a fee.

curl -X POST "$MEASURA_URL/rest/v1/rpc/export_my_account_data" \
  -H "apikey: $ANON_KEY" \
  -H "Authorization: Bearer $SESSION_JWT" \
  -H "Content-Type: application/json" \
  -d '{}' > measura-export.json

The export contains your apps, campaigns, short links, installs, attributions with their full confidence breakdown, raw events, fraud flags and payment records, plus a totals block so you can check nothing was truncated.

What to do about the SDK

Remove the Measura dependency and the initialisation call. Nothing is left behind on the device except the local event queue, which is cleared when the app is uninstalled. There is no server call required to deactivate.

Deleting what we hold

Exporting does not delete. When you have your data and have confirmed it is complete, raise a deletion request from the dashboard and the account is erased on approval. Deletion is irreversible, which is why it is a separate, deliberate step rather than a side effect of exporting.

1. 1.Check your totals. Compare installs by month in the Measura dashboard against the same period in your previous platform. Small differences are expected, since deduplication and event type normalisation are not identical. Large ones are worth investigating before you switch over.
2. 2.Run both platforms in parallel. Keep the old SDK alongside Measura for two to four weeks. Comparing live numbers is the only reliable way to build confidence in the switch.
3. 3.Expect attribution differences. Measura declines to report matches below its confidence floor and reports them as organic instead. Your attributed share may look lower than your previous platform reported, with a corresponding rise in organic. That gap is the point rather than a defect, and the confidence score page explains the arithmetic.
4. 4.Then remove the old SDK. Once the numbers reconcile and your team trusts the new figures.

Migrating a large history, or coming from a platform not listed here? Talk to us and we will work through it with you.