diff --git a/docs/deployment.md b/docs/deployment.md
new file mode 100644
index 0000000..3492eb5
--- /dev/null
+++ b/docs/deployment.md
@@ -0,0 +1,54 @@
+# Deployment
+
+This page explains how to generate and upload a complete set of metadata and
+data, which conforms to the [output specification](output_structure.md), to
+Azure blob storage.
+
+## Overview
+
+The deployment process is divided into two main steps:
+
+1. **Fetching the data.** This means running the entire job for each country.
+   This step does not publish anything to Azure; instead, it generates pickled
+   data that is stored inside `$DAGSTER_HOME`.
+
+2. **Uploading the data.** This means running each of the cloud sensor assets,
+   which have the names `publish...`. There are four of these assets: one for
+   the `countries.txt` file, one for the metadata structs, one for the
+   geometries, and one for the metrics. These assets are tied to custom IO
+   managers which read the pickled data from `$DAGSTER_HOME` and upload it to
+   Azure blob storage.
+
+Both of these steps are automated using the `popgetter.run` module, which is in
+turn invoked by the `deploy.sh` script in the repository root.
+
+## Required environment variables
+
+The deployment process requires the following environment variables to be set:
+
+- `$POPGETTER_COUNTRIES`: A comma-separated list of country IDs to generate data
+  for. This list also feeds into the `countries.txt` file.
+
+- `$ENV`: Set this to `prod` to deploy to Azure. (You can set it to `dev` too,
+  but this will publish the data to a local temporary directory, so it's only
+  useful for testing the script.)
+
+- `$SAS_TOKEN`: The SAS token for the Azure blob storage account. Contact a
+  popgetter maintainer if you need this.
+
+So, a typical deployment command might look like this:
+
+```bash
+POPGETTER_COUNTRIES=bel,gb_nir ENV=prod SAS_TOKEN="..." ./deploy.sh
+```
+
+Note that the `SAS_TOKEN` value must be quoted, because it contains ampersands
+which the shell will take to mean "run this command in the background".
+
+## Where are the data stored?
+
+The data and metadata will be uploaded to the
+['popgetter' Azure storage account](https://portal.azure.com/#@turing.ac.uk/resource/subscriptions/06e7b12a-f395-4021-9fa2-5305fa01903e/resourceGroups/popgetter/providers/Microsoft.Storage/storageAccounts/popgetter/containersList)
+under the Urban Analytics Technology Platform subscription: specifically, it
+will be placed in the container named `prod`, and the directory corresponding to
+the current version of popgetter.