The Data We Have,
and Cannot Find

The American Red Cross is not short on data. It is drowning in it — and almost none of it can see the rest. Every line of service keeps its own: Disaster Response has the fires and the shelters, Recovery has the casework and the dollars, Biomedical has the blood drives, Volunteer Services has the people, Real Estate has the buildings. Each lives in its own system, behind its own login, guarded by its own gatekeeper. To assemble one honest picture of a single county, you must track down a dozen owners and a dozen exports. The data is real. The access is tribal.

These are the questions that should be a single query and are instead a multi-week project: Where are our DAT volunteers, and can they actually cover the home fires we keep responding to? How much have we invested in long-term recovery in this county over the last three years? How many nights did we shelter people here, and where? Which counties have the most need and the least Red Cross presence? Every one of these is answerable. None of them has a single place to ask.

This report inventories that data: roughly fifty Red Cross operational layers, grouped by line of service. For each it records what the layer tells us, where it lives inside the organization today, how we actually reach it, and how hard that is. Public reference data — Census, hazard, vulnerability — is real and valuable, but it is already easy to get; it appears at the back as context we layer on top, not as the headline. The headline is everything we generate ourselves and then lose track of.

Every layer in this report is hard to use for two separate reasons, and it helps to keep them apart.

The first axis is the silo: which pocket of the organization holds it. A regional GIS analyst has the Florida DAT layer; a national team owns the 800-RED-CROSS calls; one person receives the BioMed file each month; Recovery owns the relief dollars. None of these were built to be seen together, so they aren’t.

The second axis is the friction: how hard the data is to get out once you find it. That friction sorts onto a four-rung ladder — and the inventory tables that follow tag every layer with its rung.

A layer can be low on one axis and high on the other — easy to extract but trapped in one region, or organization-wide but locked behind a system with no export. The pages that follow turn horizontal so the full inventory reads across one line: what it answers, where it lives today, how we reach it, and how hard that is.

Read the friction column down the page and the pattern is impossible to miss: almost nothing the Red Cross generates itself is Open. The wall of MANUAL, NEGOTIATED, and LOCKED tags is the report’s real finding. The data exists; reaching it is the job.

Three things compound:

This is why the same map gets rebuilt every year, why two teams report two different numbers for the same county, and why a question that should take seconds takes a month. It is not a failure of any one program — each protects its data for good reasons. It is the absence of a shared place and a shared key to bring them together.

Stop Fixing the Same Errors Twice

The friction is not only about access. It is also that the same work is redone every single time. A team downloads a source, geocodes the same addresses, and re-applies the same corrections to the same known errors — misspelled county names, bad ZIP codes, Connecticut’s planning regions, territory rows that don’t match. Then it hand-joins the result to chapter, region, and division. Next month, someone downloads the same file and does all of it again. The corrections die in a spreadsheet and are never saved back.

The answer is to move the work upstream. Data should be born correct and born joinable — fixed once, where it is created, not re-rehabilitated by every team that touches it.

What “Good” Looks Like

This report does not name a platform. That choice — a warehouse, a hosted layer service, a shared reference database — belongs to the organization, and the technology matters far less than the criteria it must meet. Wherever this data lands, it should pass five tests.

The Shared Place Now Exists

The platform choice has been made the simplest way possible: a versioned repository of flat reference layers — red-cross-data — is the source of truth, and every tool reads from it or from a mirror published out of it. It already holds the county spine (one row per FIPS, 86 fields), every rolled-up grain, boundary geometry, a source registry, and provenance on every field. Here is how it answers the five tests above today.