DEV Community

DataCheck Research
DataCheck Research

Posted on

Publishing an auditable company-trends dataset: what we chose to expose

Open data is most useful when it is not merely downloadable, but explainable.

At DataCheck Research, we publish a small, reusable dataset of Israeli company-registration and status-change trends. The objective is not to claim that one monthly count explains the economy. It is to make a constrained question reproducible: what changed in the register, when, and according to which sources?

The modelling choice

A registration date, a dissolution event and a registry status update are not interchangeable. In particular, an update appearing in a public register may lag the legal event behind it. Treating all status changes as the same event creates a chart that looks decisive but is methodologically weak.

We therefore keep separate series for registrations and status changes, document their provenance, and label the combined series as a net-movement proxy rather than a census of operating businesses.

What a reusable public dataset needs

  1. Provenance. Every series should lead back to its source and extraction method.
  2. Scope. State what the record represents and what it does not.
  3. Stable fields. A lightweight CSV is often more durable than a dashboard-only chart.
  4. Citation. Reuse should preserve visible attribution to the original source.
  5. A human explanation. Metadata is not a substitute for a method note.

The repository includes CSV files, methodology, contribution guidance and a citation file. It is released under CC BY 4.0: reuse is welcome with visible credit and a link back to the source.

A public dataset earns trust when a skeptical reader can understand where it came from, what changed, and where its limits are.

opendata #datascience #research #dataviz

Top comments (0)