diff --git a/docs/howto/create_new_package.md b/docs/howto/create_new_package.md
index cd2d2643e..ddfe115e3 100644
--- a/docs/howto/create_new_package.md
+++ b/docs/howto/create_new_package.md
@@ -12,6 +12,8 @@ This will ensure that you're following latest recommendations for the package fo
 
 _Pick the directory where you'd like to create a new package. For integrations, it's: [packages/](https://github.com/elastic/integrations/tree/master/packages)._
 
+_Decide what [package type](./package_types.md) you would like to create._
+
 ### Steps
 
 1. Bootstrap new package using the TUI wizard: `elastic-package create package`.
diff --git a/docs/howto/package_types.md b/docs/howto/package_types.md
new file mode 100644
index 000000000..58ad7a700
--- /dev/null
+++ b/docs/howto/package_types.md
@@ -0,0 +1,55 @@
+## Package types
+
+There are different types of packages, for different use cases.
+
+Integration packages provide complete monitoring solutions for specific
+applications of services. An integration package can be used to configure
+multiple data streams and multiple agent inputs, it also includes dashboards
+and other assets related to the monitored service.
+For example the Apache package includes configurations to collect and store
+access and error logs and metrics. It also includes dashboards to leverage the
+collected data.
+
+Input packages are intended for generic use cases, where the user needs to
+collect custom data. They define how to configure an agent input, so the user
+can adapt it to their case. An example is the log input package, that users can
+use to collect custom logs, and provide their own processing and mapping.
+
+Content packages include assets for the Elastic Stack, but are not intended to
+define how to ingest data. They can include static data for indexes, as
+knowledge bases or sample data. Or they can contain resources that can be used
+with data collected according to certain conventions, for example dashboards for
+data collected by other means.
+
+### What package type to use?
+
+When [creating a new package](./create_new_package.md) you will need to select
+the type. If you are not sure, you can follow this section to help deciding:
+
+* Do you plan to provide a whole solution for a given service?
+  * If yes, can Fleet manage the collector agent?
+    * If yes, create an integration package.
+    * If not, create a content package that complements the method/s used for ingestion.
+  * If not, do you plan to define how to collect data?
+    * If yes, do you plan to define how to collect data in a generic way for a given protocol?
+      * If yes, create an input package.
+      * If not, create an integration package for the use case.
+    * If not, create a content package.
+
+### A note on using integration packages for everything
+
+For legacy reasons, integration packages can be used for most use cases. For
+example a package can include dashboards without including any data stream, as
+if it were a content package. Or an integration package can define a single data
+stream, as if it were an input package.
+
+Even when this is true, these cases push integration packages (and the code
+supporting them) beyond what they were designed for. By having different types
+we can focus on better supporting this specialized use cases. For example content
+packages can be allowed to have bigger sizes, and we can prepare all our infra
+and code for them. Or input packages are better designed for custom use cases,
+giving a better experience on these cases.
+
+In the future we may be limiting what can be done with integration packages,
+providing migration paths to the other types.
+