Case · ECOM Chaco
Institutional dashboards on already-unified data
At ECOM the work is not a stray spreadsheet: it is continuous analytic operations. Datasets, metrics, and publication in Apache Superset and Power BI, scoped by organization. Dashboards feed institutions other than the team that builds them.
The problem to solve
Decision-makers do not enter the source system. They query a slice: a dashboard, an embed, or an export with permissions. If that slice is poorly defined, the meeting and the procedure suffer. An indicator is not decoration: it is the number used to decide.
The work starts by agreeing what each number means — which field, which frequency, who may see it — and only then is it published. The source system is not rewritten so a KPI can be shown.
What gets published
Each agency sees only its slice. The dashboard and the report follow the same rule. Apache Superset and Power BI are the publication channel, on production SQL Server and PostgreSQL.
Refresh runs on a cluster, not a laptop. Where a map is needed, there is a map. Institutional internal use, not a course demo.
If it fails, it alerts
If a load or refresh fails, the system alerts. It does not fail silently. There is a record of who queries or exports. Governance is the slice the user sees when they open the dashboard.
This case does not list non-public system names or internal figures. The point is the class of problem — institutional data, many agencies, living legacy — and the class of solution: unify, publish, alert.
Jobs are orchestrated with queues (Celery / Redis).
What this work is not
It is not a departmental spreadsheet. It is not a rewrite of the transactional core. It is not a commercial ECOM product exposed on this page. ECOM is the employment context; the site documents the kind of production analytics, with the organization’s public URL.
Skadia, SIGAG, and the demonstration lakehouse are another context: field, offline, livestock. The method is similar (unify, model, publish, alert on failure). The environment is not: here the user is an agency; there, someone in a paddock without signal.
Recruiter questions
- Superset or Power BI?
- Both, depending on the agency and the stack it already runs. The work is the dataset, the metric, and the slice; the visualization tool is the publication channel.
- Is the source system rewritten?
- No. A query layer is exposed over SQL Server and PostgreSQL. The transactional core stays.
- Are management figures on this page?
- No. No internal numbers or non-public system names. The case is described by the problem and the design, not by a confidential dashboard.
Institutional internal use. No day-to-day screens of third parties.