diff --git a/docs/initiatives.rst b/docs/initiatives.rst
index d49d4d2..a63d1d0 100644
--- a/docs/initiatives.rst
+++ b/docs/initiatives.rst
@@ -9,7 +9,6 @@ Drafts
 .. toctree::
     :maxdepth: 1
 
-    registry_to_quay/index
 
 Completed review
 ----------------
@@ -26,6 +25,7 @@ Completed review
     github2fedmsg/index
     badges/index
     fcas/index
+    registry_to_quay/index
 
 Implemented
 -----------
diff --git a/docs/registry_to_quay/index.rst b/docs/registry_to_quay/index.rst
index 6d6caba..f62e609 100644
--- a/docs/registry_to_quay/index.rst
+++ b/docs/registry_to_quay/index.rst
@@ -50,16 +50,27 @@ Fedora flatpaks, Fedora container images (base, minimal, coreos) and toolbox ima
 Conclusions
 -----------
 
-This is a conclusion for investigation. It should contain output and recommendations.
+Migrating `registry.fedoraproject.org <https://registry.fedoraproject.org/>`_ to `quay.io <https://quay.io/>`_
+doesn't seem to be initiative worthy. Especially with changes happening in Fedora 39.
+There needs to be changes done to all three build pipelines we investigated, but those aren't complex.
+
+Recommendations
+---------------
+
+It will be best to first set everything on staging and then continue with production.
+Only redirect the Fedora 39+ workflow and continue with step 4 after Fedora 38 will
+be EOL.
 
 Proposed Roadmap
 ----------------
 
-* Step 1 - Migrate registry.fp.o
-* Step 2 - ???
-* Step 3 - Profit
+* Step 1 - Create namespaces corresponding to candidate-registry and registry on `quay.io <https://quay.io/>`_ (Optional: Create staging namespaces as well)
+* Step 2 - Modify configurations and scripts (see corresponding investigation document for what needs
+  to be changed)
+* Step 3 - Redirect proxies to `quay.io <https://quay.io/>`_
+* Step 4 - Decommision `candidate-registry <https://candidate-registry.fedoraproject.org/>`_ and `candidate-registry <https://candidate-registry.fedoraproject.org/>`_
 
 Estimate of work
 ----------------
 
-The estimate for this work.
+This should be handled as normal ticket. It would need 1 sysadmin for 2-4 weeks.