diff --git a/lib/snowplow-emr-etl-runner/emr_job.rb b/lib/snowplow-emr-etl-runner/emr_job.rb
index 6dfe823..468cf56 100755
--- a/lib/snowplow-emr-etl-runner/emr_job.rb
+++ b/lib/snowplow-emr-etl-runner/emr_job.rb
@@ -593,7 +593,7 @@ def initialize(debug, staging, enrich, staging_stream_enrich, shred, es, archive
           archive_shredded_step = get_archive_step(csbs[:good], csbs[:archive], run_id, s3_endpoint, ": Shredded S3 -> Shredded Archive S3", encrypted)
           @jobflow.add_step(archive_shredded_step)
         elsif archive_shredded == 'recover'
-          latest_run_id = get_latest_run_id(s3, csbs[:good])
+          latest_run_id = get_latest_run_id(s3, csbs[:good], 'atomic-events')
           archive_shredded_step = get_archive_step(csbs[:good], csbs[:archive], latest_run_id, s3_endpoint, ": Shredded S3 -> S3 Shredded Archive", encrypted)
           @jobflow.add_step(archive_shredded_step)
         else    # skip
@@ -798,21 +798,26 @@ def get_rdb_loader_steps(config, targets, resolver, jar, rdbloader_steps, skip_m
       end
 
       # List bucket (enriched:good or shredded:good) and return latest run folder
-      # Assuming, there's usually just one folder
       #
       # Parameters:
       # +s3+:: AWS S3 client
       # +s3_path+:: Full S3 path to folder
-      def get_latest_run_id(s3, s3_path)
+      # +suffix+:: Suffix to check for emptiness, atomic-events in case of shredded:good
+      def get_latest_run_id(s3, s3_path, suffix = '')
         run_id_regex = /.*\/run=((\d|-)+)\/.*/
-        folders = list_object_names(s3, s3_path,
+        folder = last_object_name(s3, s3_path,
             lambda { |k| !(k =~ /\$folder\$$/) and !k[run_id_regex, 1].nil? })
-          .map { |k| k[run_id_regex, 1] }
-        if folders.empty?
+        run_id = folder[run_id_regex, 1]
+        if run_id.nil?
           logger.error "No run folders in [#{s3_path}] found"
           raise UnexpectedStateError, "No run folders in [#{s3_path}] found"
         else
-          folders.first
+          path = File.join(s3_path, "run=#{run_id}", suffix)
+          if empty?(s3, path)
+            raise NoDataToProcessError, "Cannot archive #{path}, no data found"
+          else
+            run_id
+          end
         end
       end
 
diff --git a/lib/snowplow-emr-etl-runner/s3.rb b/lib/snowplow-emr-etl-runner/s3.rb
index 32ed2dc..8042b18 100644
--- a/lib/snowplow-emr-etl-runner/s3.rb
+++ b/lib/snowplow-emr-etl-runner/s3.rb
@@ -36,15 +36,15 @@ def empty?(client, location,
         empty_impl(client, bucket, prefix, key_filter)
       end
 
-      # List all object names satisfying a key filter.
+      # Retrieve the alphabetically last object name satisfying a key filter.
       #
       # Parameters:
       # +client+:: S3 client
       # +location+:: S3 url of the folder to list the object names for
       # +key_filter+:: filter to apply on the keys
-      def list_object_names(client, location, key_filter)
+      def last_object_name(client, location, key_filter)
         bucket, prefix = parse_bucket_prefix(location)
-        list_object_names_impl(client, bucket, prefix, key_filter)
+        last_object_name_impl(client, bucket, prefix, key_filter)
       end
 
       # Extract the bucket and prefix from an S3 url.
@@ -59,18 +59,23 @@ def parse_bucket_prefix(location)
 
     private
 
-      def list_object_names_impl(client, bucket, prefix, key_filter, max_keys = 50)
+      def last_object_name_impl(client, bucket, prefix, key_filter, max_keys = 50)
         continuation_token = nil
-        filtered = []
+        last = ""
         loop do
           response = list_objects(client, bucket, prefix, max_keys, continuation_token)
-          filtered = filtered + response.contents
+          new_last = response.contents
             .select { |c| key_filter[c.key] }
             .map { |c| c.key }
+            .sort
+            .last
+          if not new_last.nil? and new_last > last
+            last = new_last
+          end
           continuation_token = response.next_continuation_token
           break unless response.is_truncated
         end
-        filtered
+        last
       end
 
       def empty_impl(client, bucket, prefix, key_filter, max_keys = 50)
diff --git a/spec/snowplow-emr-etl-runner/s3_spec.rb b/spec/snowplow-emr-etl-runner/s3_spec.rb
index 80c4bb4..d8e5de4 100644
--- a/spec/snowplow-emr-etl-runner/s3_spec.rb
+++ b/spec/snowplow-emr-etl-runner/s3_spec.rb
@@ -54,15 +54,27 @@
     end
   end
 
-  describe '#list_object_names' do
+  describe '#last_object_name' do
     it 'should take a client, a location and a filter argument' do
-      expect(subject).to respond_to(:list_object_names).with(3).argument
+      expect(subject).to respond_to(:last_object_name).with(3).argument
     end
 
     it 'should filter file names based on the filter' do
       s3.stub_responses(:list_objects_v2, { contents: [{ key: 'abc' }, { key: 'defg' }]})
-      expect(subject.list_object_names(s3, 's3://bucket/prefix', lambda { |k| k.length == 3}))
-        .to eq(['abc'])
+      expect(subject.last_object_name(s3, 's3://bucket/prefix', lambda { |k| k.length == 3}))
+        .to eq('abc')
+    end
+
+    it 'should retrieve the alphabetically file names based on the filter' do
+      s3.stub_responses(:list_objects_v2, { contents: [{ key: 'abc' }, { key: 'defg' }]})
+      expect(subject.last_object_name(s3, 's3://bucket/prefix', lambda { |k| k.length >= 0}))
+        .to eq('defg')
+    end
+
+    it 'should be the empty string if there is nothing' do
+      s3.stub_responses(:list_objects_v2, { contents: []})
+      expect(subject.last_object_name(s3, 's3://bucket/prefix', lambda { |k| k.length >= 0}))
+        .to eq('')
     end
 
   end