diff --git a/docs/source/glossary.rst b/docs/source/glossary.rst
index dd691b6..6717472 100644
--- a/docs/source/glossary.rst
+++ b/docs/source/glossary.rst
@@ -24,6 +24,7 @@
 
     error
     error code
+    violation
         The symbol associated with a specific :term:`check`. For example,
         pycodestyle implements :term:`check`\ s that look for whitespace
         around binary operators and will either return an error code of
diff --git a/docs/source/internal/index.rst b/docs/source/internal/index.rst
index 482b898..06fcd13 100644
--- a/docs/source/internal/index.rst
+++ b/docs/source/internal/index.rst
@@ -18,6 +18,7 @@ pull gently.
     contributing
     writing-documentation
     releases
+    start-to-finish
     checker
     cli
     formatters
diff --git a/docs/source/internal/start-to-finish.rst b/docs/source/internal/start-to-finish.rst
new file mode 100644
index 0000000..381ef91
--- /dev/null
+++ b/docs/source/internal/start-to-finish.rst
@@ -0,0 +1,129 @@
+==================================
+ What Happens When You Run Flake8
+==================================
+
+Given |Flake8| 3.0's new organization and structure, it might be a bit much
+for some people to understand what happens from when you call ``flake8`` on the
+command-line to when it completes. This section aims to give you something of
+a technical overview of what exactly happens.
+
+
+Invocation
+==========
+
+The exact way that we end up in our ``main`` function for Flake8 depends on
+how you invoke it. If you do something like:
+
+.. prompt:: bash
+
+    flake8
+
+Then your shell looks up where ``flake8`` the executable lives and executes
+it. In almost every case, this is a tiny python script generated by
+``setuptools`` using the console script entry points that |Flake8| declares
+in its :file:`setup.py`. This might look something like:
+
+.. code-block:: python
+
+    #!/path/to/python<version>
+    # EASY-INSTALL-ENTRY-SCRIPT: 'flake8==3.0.0','console_scripts','flake8'
+    __requires__ = 'flake8==3.0.0'
+    import sys
+    from pkg_resources import load_entry_point
+
+    if __name__ == '__main__':
+        sys.exit(
+            load_entry_point('flake8==3.0.0', 'console_scripts', 'flake8')()
+        )
+
+If instead you invoke it like:
+
+.. prompt:: bash
+
+    python -m flake8
+
+Then you're relying on Python to find :mod:`flake8.__main__` and run that. In
+both cases, however, you end up in :func:`flake8.main.cli.main`. This is the
+primary way that users will end up starting Flake8. This function creates an
+instance of |Application|.
+
+via Setuptools
+--------------
+
+If you're invoking |Flake8| from your ``setup.py`` then you actually end up in
+:meth:`flake8.main.setuptools_command.Flake8.run`. This then collects the
+files that are included in the package information and creates an instance of
+|Application|.
+
+via Git or Mercurial
+--------------------
+
+In both cases, they call their respective ``hook`` functions which create
+instances of |Application|.
+
+
+Application Logic
+=================
+
+When we create our |Application| instance, we record the start time and parse
+our command-line arguments so we can configure the verbosity of |Flake8|'s
+logging. For the most part, every path then calls
+:meth:`~flake8.main.application.Application.run` which in turn calls:
+
+- :meth:`~flake8.main.application.Application.initialize`
+- :meth:`~flake8.main.application.Application.run_checks`
+- :meth:`~flake8.main.application.Application.report_errors`
+- :meth:`~flake8.main.application.Application.report_benchmarks`
+
+Our Git hook, however, runs these individually.
+
+Application Initialization
+--------------------------
+
+:meth:`~flake8.main.application.Application.initialize` loads all of our
+:term:`plugin`\ s, registers the options for those plugins, parses the
+command-line arguments, makes our formatter (as selected by the user), makes
+our :class:`~flake8.style_guide.StyleGuide` and finally makes our
+:class:`file checker manager <flake8.checker.Manager>`.
+
+Running Our Checks
+------------------
+
+:meth:`~flake8.main.application.Application.run_checks` then creates an
+instance of :class:`flake8.checker.FileChecker` for each file to be checked
+after aggregating all of the files that are not excluded and match the
+provided file-patterns. Then, if we're on a system that supports
+:mod:`multiprocessing` **and** :option:`flake8 --jobs` is either ``auto`` or
+a number greater than 1, we will begin processing the files in subprocesses.
+Otherwise, we'll run the checks in parallel.
+
+After we start running the checks, we start aggregating the reported
+:term:`violation`\ s in the main process. After the checks are done running,
+we record the end time.
+
+Reporting Violations
+--------------------
+
+Next, the application takes the violations from the file checker manager, and
+feeds them through the :class:`~flake8.style_guide.StyleGuide`. This
+determines whether the particular :term:`error code` is selected or ignored
+and then appropriately sends it to the formatter (or not).
+
+Reporting Benchmarks
+--------------------
+
+Finally, if the user has asked to see benchmarks (i.e., :option:`flake8
+--benchmark`) then we print the benchmarks.
+
+
+Exiting
+=======
+
+Once :meth:`~flake8.main.application.Application.run` has finished, we then
+call :meth:`~flake8.main.application.Application.exit` which looks at how
+many errors were reported and whether the user specified :option:`flake8
+--exit-zero` and exits with the appropriate exit code.
+
+
+.. Replacements
+.. |Application| replace:: :class:`~flake8.main.application.Application`