Sustainability index · Data & Automation · 2022

80k companies, one lookup tool

ARCHIVED

A third-party data provider had the company records we needed, but no official API. Access was gated behind a web interface that analysts had to click through manually. The whole update cycle slowed down because the bottleneck was human navigation.

Trigger

An analyst spent three days re-checking records that had already been gathered the previous month.

  1. 01

    Buy access, then watch the network tab

    We paid for a seat on the platform. The real work was opening the browser dev tools, recording the requests, and reconstructing the undocumented endpoints and auth flow.

  2. 02

    Automate with Python

    A Python script replayed those calls with proper session handling, pagination and retries. It ran headlessly and logged every request so failures were traceable.

  3. 03

    Structure and store

    Raw responses were normalized and loaded into PostgreSQL with source timestamps, so the index was always built from the latest data rather than a stale export.

  4. 04

    Expose to the team

    Analysts could query the structured dataset through a simple lookup interface instead of opening the original tool at all.

  • Python
  • PostgreSQL
  • Requests
  • Reverse-engineering

DAYS

of manual clicking reduced to a script that runs in minutes

What I'd do differently

The data itself is never stale — it comes straight from the provider's live API, so whatever the script pulls is current. The unfinished part is the trigger: I never solved automatic renewal of the auth token, so the run still has to be kicked off by hand with a fresh session. It saves days of clicking, but it is not yet a scheduled job. Capturing the login flow properly, instead of copying a token, is the piece I would finish next.

Share this

Next case study — 2025AI workflows