But behind the curtain was a different story: overworked teams cobbled together analytical products using decade-old tooling. They pulled data from myriad geoportals and field surveys and then used desktop GIS to build vector maps for responding organizations, including maps of closed roads, critical infrastructure, and displaced populations. Due to continuing distrust in AI algorithms and non-optimal workflows, they performed damage assessments manually. When donor appeals went out, all this work was condensed into written text and frozen into simple static PDFs each containing a few vector maps that outlined the damage, accompanied by eye-level photographs of people in need. Satellite imagery was siloed into a very small community of users.

For the humanitarian organizations we’ve talked to, such low usage of EO data is normal, familiar, and undesirable. We’ve heard this same story repeatedly over the past decade, so much so that even before this hurricane season, we started exploring how to fix it.

Fig. 1: A PDF describing Who Does What, Where (3W) during Hurricane Beryl