diff --git a/docs/adr/0000-record-architecture-decisions.md b/docs/adr/0000-record-architecture-decisions.md
new file mode 100644
index 00000000..85b17402
--- /dev/null
+++ b/docs/adr/0000-record-architecture-decisions.md
@@ -0,0 +1,19 @@
+# 1. Record architecture decisions
+
+Date: 2023-11-16
+
+## Status
+
+Accepted
+
+## Context
+
+We need to record the architectural decisions made on this project.
+
+## Decision
+
+We will use Architecture Decision Records, as [described by Michael Nygard](http://thinkrelevance.com/blog/2011/11/15/documenting-architecture-decisions).
+
+## Consequences
+
+See Michael Nygard's article, linked above. For a lightweight ADR toolset, see Nat Pryce's [adr-tools](https://github.com/npryce/adr-tools).
diff --git a/docs/adr/0001-use-grafana-for-dashboards.md b/docs/adr/0001-use-grafana-for-dashboards.md
new file mode 100644
index 00000000..810d02e7
--- /dev/null
+++ b/docs/adr/0001-use-grafana-for-dashboards.md
@@ -0,0 +1,23 @@
+# 1. Record architecture decisions
+
+Date: 2023-11-16
+
+## Status
+
+Accepted
+
+## Context
+
+As a requirement for the [Metrics and Dashboards Initiative](https://docs.google.com/document/d/1VTNfY2Ezv4wxWUAFGhCeOHns285tko-AY0pX2Eo7gTc/edit#heading=h.mnb03peyeng9), we need to be able to display metrics and graphs (in particular, timeseries graphs) to a variety of people across the Bennett Institute.
+
+We carried out a couple of spikes ([1](https://github.com/opensafely-core/job-runner/issues/649), [2](https://github.com/opensafely-core/job-server/pull/3578)) to assess how many of these use cases could be fullfilled by Grafana (or other solutions) and how easy it was to setup and use.
+
+## Decision
+
+To use a self-hosted Grafana instance for the dashboards for the product, system and delivery metrics. To use a managed Digital Ocean Postgres database for the Grafana data.
+
+## Consequences
+
+We considered a few alternatives, such as Honeycomb (which we're already using for opentelemetry data) and datasette (good for visualising sqlite data). However, we felt that Grafana and its plugin ecosystem offered a good selection of features that would allow us to meet the objectives for all three areas (product, system and delivery) within a reasonable timeframe. We believe that self-hosting Grafana, backed by a managed Digital Ocean database, will be low maintenance. We may wish to consider alternatives, such as Grafana Cloud, if that turns out not to be the case.
+
+In future, we may want to use more specialised tools for the three different areas. For example, with the product dashboards, a business intellience tool, such as [Apache Superset](https://superset.apache.org/) or [Metabase](https://www.metabase.com/) may be required to allow easier editing of data visualisations by non-developers and for an improved ability to drill down into the data. We may also want to consider specialised tooling such as prometheus to improve the system dashboards.