A unified telco data foundation still needs a shared join lesson
Agree the namespaces and mapping revision before teaching a cross-domain network use case.
One platform, several identities
Putting subscriber, radio and network observations on one platform does not make their identifiers interchangeable. A cross-domain learning session should begin with the join contract, before anyone demonstrates an AI answer. Truthound can check the reference links in a prepared exercise. It cannot establish that two observations concern the same subscriber or diagnose a live network by itself.
The proposed telco foundation
Nokia and Microsoft announced a unified telco data foundation on September 17, 2026. The release describes Nokia data products combined with Microsoft Fabric, including use cases correlating subscriber, network and RF data. Claimed reductions in integration time remain the partners' statements. The relevant development here is the proposed cross-domain operating model, not a measured outcome for every operator.
Give each team a definition
Give each participating team a distinct responsibility. The RF team defines observation identity and sampling; the network team owns element references; the customer-experience team defines the subscriber-key scope it is authorized to use. A string called id in all three extracts is not a shared identity. Preserve the namespaces until an approved translation explicitly connects them.
Require the agreed fields
An exercise manifest can record those namespaces, the mapping version and the expected join cardinality. For the telecom identifier exercise manifest prepared as a Polars DataFrame, Truthound Core Python SDK 3.3.4's ColumnExistsValidator can require the agreed fields to be present. A separate reference rule can inspect the supplied mapping. Field presence is an intentionally small obligation: it should stop the exercise when a required identifier is absent, without declaring that a present identifier has the right meaning.
An existing schema alternative
Pandera 0.35.0 schemas are a direct alternative for that structural acceptance rule, including a Polars implementation. Keep the schema already used in team training if it accurately describes the same fields. Custom rules and the translation process still require tests. Neither framework converts a field called subscriber_id into an authorized, globally unique customer identity.
Teach a namespace collision
Teach the exception with a small example whose local identifiers collide across two namespaces. Ask participants to predict the join, inspect the mapping and explain the rejected match. This makes the data contract part of the collaboration rather than an invisible preprocessing step. Use synthetic or permissioned examples so the lesson does not broaden access to subscriber information.
Keep the lesson current
Publish the exercise's agreed dictionary and revision alongside its inputs. When a data product changes, the responsible team should update the lesson and rerun the acceptance checks before a demonstration is reused. A shared platform can simplify delivery; shared understanding of the identifiers is the condition that makes its cross-domain claims testable.
Evidence
- Event checked
- Subject checked
- Modified
