Mergewall
A pull-request impact check that traces model and column changes to downstream BI assets, routes uncertainty to owners and applies team-approved merge policy.
A data-model change can pass technical tests and still break a live dashboard because the transformation repository and BI layer are validated separately. The supplied research confirms enterprise lineage products, an existing pull-request impact action and real APIs for a major BI platform. It also confirms open column-lineage substrate.
The gap is narrower than the original thesis. Sifflet already posts asset-level impact comments, though as an enterprise observability add-on rather than a self-serve blocking check. Column lineage is incomplete around dynamic SQL, templating, extracts, semantic calculations, manual uploads and stale ownership. A predicted impact therefore cannot automatically own merge authority. The claimed per-customer graph may become sticky, but incumbents can move down-market and no structural conflict is proven.
A code diff, parsed model change, column-lineage candidate, downstream asset match, runtime usage observation, owner record, impact hypothesis, owner finding, policy result, waiver, merge decision, deployment readback and business incident are separate. Mergewall should make likely blast radius reviewable before enforcement.
An analytics-engineering or data-platform lead at a mid-market company with version-controlled transformations and business-critical dashboards across one or more BI tools.
Preventing dashboard breakage requires early intervention, but false or incomplete lineage can make an automatic blocker harmful.
Lineage tooling and BI interfaces now make pre-merge impact analysis practical.
Several related signals support a transformation-to-BI dependency boundary.
The input identifies a concrete failure mode, confirmed lineage interfaces, active enterprise spend and a closest competitor that still leaves a self-serve enforcement gap.
The buyer quartet is incomplete, lineage coverage is imperfect, closest competition already reaches pull requests, blocking can disrupt teams and incumbents can package a lower tier.
Discussion
No comments yet — be the first to weigh in.