How Identity Stitching Works
Deterministic Resolution
When an event arrives with a known identifier (MSISDN):
- Normalize (E.164 format)
- Hash for privacy-safe storage
- Query identity graph for existing DTX_ID
- Found → associate event | Not found → generate new DTX_ID
100% confidence — relies on exact matches.
Match Rules
| Rule | Match Keys | Priority | Confidence |
|---|---|---|---|
| MSISDN Exact | MSISDN | 1 | 100 |
| Email Exact | Email (hashed) | 2 | 100 |
| Account + MSISDN | Account ID + MSISDN | 3 | 100 |
| IMEI + Recent MSISDN | IMEI + MSISDN (30d) | 4 | 95 |
Probabilistic Resolution
When deterministic identifiers unavailable — ML models score candidates:
- ≥90% → auto-merge
- 50-89% → human review queue
- <50% → create new DTX_ID
Confidence Scoring
| Factor | Weight | Description |
|---|---|---|
| Match type | 40% | Deterministic = 1.0, Probabilistic varies |
| Data recency | 25% | Recent activity increases confidence |
| Source reliability | 20% | BSS > Web > Third-party |
| Corroborating evidence | 15% | Multiple identifiers matching |
Anonymous-to-Known Stitching
Anonymous visitors accumulate events under cookie ID. On identification (login, registration, OTP), anonymous behavior stitches to known profile if consent granted. 14-day event backfill.
Stitch Methods
LOGIN · REGISTRATION · FORM_SUBMIT · SSO · MSISDN_CAPTURE
Retention Rules
- Anonymous identifiers: 90 days default (configurable 30-180)
- Stitching requires explicit consent (configurable per tenant)
- DSAR includes stitched anonymous history