When Multi-Source Analysis Has to Earn Its Answer
A LightBI Beta development note on exact document reconciliation, source-bound evidence, and why multi-source analysis is not the same as putting files in one table.
Putting files together is easy
Multi-source BI often begins with a tempting shortcut: load several files, find columns with similar names, join them, and continue as if the resulting table were obviously valid.
For operational data, that shortcut can create confident-looking totals from the wrong relationship. LightBI treats multi-source analysis as a separate governed boundary.
Each source keeps its identity
In the production multi-source work, Home profiles every selected source independently. The system requires explicit source role, document identity, period, currency, and monetary-column evidence before building a combined logical dataset.
The composite identity is order-invariant and built from immutable source-local artifacts and overlays. This means the relationship does not depend on file position, filename conventions, row order, or a hidden legacy fusion table.
The verified relationship was deliberately narrow
The positive case was an exact document-identity reconciliation between two May 2026 sources with 1,500 rows each. The relationship was confirmed as one-to-one across 1,500 matching identities.
That still did not authorize every possible cross-source operation. Because the two sources had different structural grains, the artifact limited the relationship to identity reconciliation and prohibited cross-source measure joins.
Gross profit continued to use the frozen Accounting formula. Sales was materialized independently, while the relationship and scope were verified separately. The full-source result was 3,075,721,244 VND, matching the expected value exactly.
Restrictions are part of the result
A governed analysis is not only the number displayed at the end. It also includes the conditions under which that number is valid. LightBI validates source fingerprints, generations, overlays, handoff data, query-plan membership, and actual participating row counts.
When the relationship is stale, incomplete, or unsupported, the system rejects it. Unsupported source combinations do not become generic joins, and unrelated Sales or Logistics sources do not silently change an inventory snapshot metric.
What this does not claim
This milestone does not establish generic multi-source BI, universal relationship discovery, or unrestricted cross-source metric joins. It closes a specific production boundary for the existing commerce and distribution MVP.
That limitation is intentional. A narrow relationship with exact evidence is more useful than a universal join feature that cannot explain why its output should be trusted.
The development lesson
Multi-source analysis has to earn its answer through source identity, relationship evidence, scope, and execution checks. The engineering work is slower than concatenating tables, but the resulting analysis can explain not only what it calculated, but also why that calculation was allowed.
LightBI is available at lightbi.thaiduy.digital. The live app is at lightbi.thaiduy.digital/app, and the implementation is open in the LightBI GitHub repository.
Technical references
The multi-source boundary and its restrictions are documented in Phase 8.1 production multi-source closure, with the broader source and relationship contracts in the architecture overview.