diff --git a/docs/adr/004_add-a-separate-data-store.md b/docs/adr/004_add-a-separate-data-store.md
new file mode 100644
index 00000000..4ce7f6ae
--- /dev/null
+++ b/docs/adr/004_add-a-separate-data-store.md
@@ -0,0 +1,69 @@
+# Add a separate data store to app
+
+Status: Draft
+
+## Context and problem statement
+
+It's likely that our web app will need a data store in addition to what Cloud Foundry provides. Anticipated uses of this data store include:
+
+- users
+  - session management
+  - metrics like used platforms, browsers, etc.
+  - preferences that can persist
+  - tailoring the experience based on user roles/attributes
+- orgs / agencies
+  - billing and agreements
+  - correlating the same person across multiple origins
+- compliance
+  - scan artifacts
+  - compliance status and ATO information
+- external connections such as GitHub tokens
+
+## Decision drivers / forces
+
+The dashboard app is a web UI for Cloud Foundry data, and the app's primary data source will be from an API.
+
+Eventually, the app might need to talk to multiple external data sources, especially while other products are consolidated into it.
+
+While we don't see an immediate need for a datastore to communicate with other clients besides our application, it is a future possibility.
+
+These factors lead us towards an API-first approach.
+
+## Considered options
+1. Use a database service with a tightly coupled API built into the dashboard application. This is similar to the approach used by [pages-core](https://github.com/cloud-gov/pages-core/blob/b3a1f6e3c55eba7555ba92007908b40b107886a0/docs/DEVELOPMENT.md#deployment)
+1. Use a database service with a separately managed API, similar to [this approach](https://cloud.gov/pages/knowledge-base/website-api/)
+1. Use a localized data source and switch to MVC (Model, View, Controller) architecture
+
+## Decision outcome
+* Chosen Option: *[tbd]*
+* *[justification. e.g., only option, which meets KO criterion decision driver | which resolves force force | ... | comes out best (see below)]*
+* *[consequences. e.g., negative impact on quality attribute, follow-up decisions required, ...]* <!-- optional -->
+
+### Success criteria
+
+- We can collect and store the data we need with ease
+- We can add more data sources with ease
+- The app has secure and compliant ways to communicate with its data sources
+
+
+## Pros and cons of the options
+
+### 1. Tightly coupled API
+
+* `+` Managing the API alongside the same application / repository means less overhead for development and deployment
+* `+` Builds for our current needs rather than theoretical futures
+* `-` The API would be in the same stack as our app, NextJS, which does not come out of the box with support for models, migrations, database-related tests, nor object relationship mapping (ORM). We would need to invest in additional work to implement these
+* `-` We risk storing and managing data that would be better off pulled into our app rather than being owned by it
+* `-` Out of the box software for an "admin" view would likely be distinct from the rest of the application and lead to two separate applications in the same repo
+
+### 2. Separately managed API
+
+* `+` Anticipates a potential future where our application pulls in data from multiple APIs / sources, and where the datastore serves multiple applications
+* `+` Allows for selection of technology / framework specifically for an API with database backend
+* `-` More overhead for our current team to support: two applications, separate authentication flow, deployment, maintenance, security scanning, etc.
+
+### 3. MVC architecture
+* `+` *[argument 1 pro]*
+* `+` *[argument 2 pro]*
+* `-` *[argument 1 con]*
+* *[...]* <!-- numbers of pros and cons can vary -->