diff --git a/read_mattermost.py b/read_mattermost.py
index 12ea4b1..2e64457 100755
--- a/read_mattermost.py
+++ b/read_mattermost.py
@@ -7,7 +7,7 @@ import datetime
 import threading
 from time import sleep
 import json
-from typing import Optional
+from typing import Optional, Mapping, Set
 import mattermost
 import mattermost.ws
 
@@ -15,6 +15,7 @@ SERVER = "mattermost.zeus.gent"
 TEAM_NAME = "zeus"
 CHAN_NAME = "pannenkoeken"
 EMOJI_NAME = "pancakes"
+DOUBLE_EMOJI_NAME = "own_pancakes"
 TAGGERS = [
 	# Board
 	"flynn",
@@ -34,6 +35,7 @@ TOKEN = os.getenv("MM_ACCESS_TOKEN")
 USER = os.getenv("MM_USERNAME")
 PASSWORD = os.getenv("MM_PASSWORD")
 
+# Set to False to disable reacting with an emoji for the count
 CONFIRMATION_EMOJI = True
 
 
@@ -48,7 +50,7 @@ if since_arg_i:
 clean = "--clean" in sys.argv[1:]
 live = "--live" in sys.argv[1:]
 if "--no-confirm" in sys.argv[1:]:
-	CONFIRMATION_EMOJI = None
+	CONFIRMATION_EMOJI = False
 
 
 
@@ -124,12 +126,46 @@ def to_mm_timestamp(dt):
 
 
 def reaction_qualifies(reaction):
-	return reaction["emoji_name"] == EMOJI_NAME and reaction["user_id"] in tagger_ids
+	if reaction["user_id"] not in tagger_ids:
+		return 0
+	if reaction["emoji_name"] == EMOJI_NAME:
+		return 1
+	if reaction["emoji_name"] == DOUBLE_EMOJI_NAME:
+		return 2
+	return 0
 
 
-awarded = {} # awarded[awardee][post_id]: set of verifiers
-def award_if_appropriate(reaction):
-	if not reaction_qualifies(reaction):
+def post_score(awardee_id, post_id, awarder_id):
+	return max(awarded[awardee_id][post_id][awarder_id], default=0)
+
+
+def emit_change_line(post, awardee_id, awarder_id, prev_score, score):
+	if score == prev_score:
+		return
+
+	awardee = get_username(awardee_id)
+	awarder = get_username(awarder_id)
+
+	post_time = parse_mm_timestamp(post["create_at"]).isoformat(timespec="microseconds")
+
+	if score == 0:
+		message = f"{awarder} retracted their verification"
+	elif prev_score == 0:
+		message = f"{awarder} verified with score {score}"
+	else:
+		message = f"{awarder} updated their verification's score from {prev_score} to {score}"
+
+	print(f"{awardee} {post['id']} at {post_time}: {message}", flush=True)
+
+
+# awarded[awardee][post_id][verifier]: set of values
+awarded: Mapping[str, Mapping[str, Mapping[str, Set[int]]]] = \
+	defaultdict(lambda: defaultdict(lambda: defaultdict(set)))
+
+
+def process_change(reaction, action):
+	value = reaction_qualifies(reaction)
+	if value == 0:
 		return
 
 	post = get_post(reaction["post_id"])
@@ -139,48 +175,25 @@ def award_if_appropriate(reaction):
 	awardee_id = post["user_id"]
 	awarder_id = reaction["user_id"]
 
-	if awardee_id not in awarded:
-		awarded[awardee_id] = {}
-	if post["id"] not in awarded[awardee_id]:
-		awarded[awardee_id][post["id"]] = set()
-
-	if awarder_id in awarded[awardee_id][post["id"]]:
-		# We already knew that this user verified this post
-		return
-
-	awarded[awardee_id][post["id"]].add(awarder_id)
-
-	reaction_time = parse_mm_timestamp(reaction["create_at"]).isoformat(timespec="microseconds")
-	post_time = parse_mm_timestamp(reaction["create_at"]).isoformat(timespec="microseconds")
-	awardee = get_username(awardee_id)
-	awarder = get_username(awarder_id)
-	print(f"{awardee} {post['id']} at {post_time} verified by {awarder} at {reaction_time}", flush=True)
+	prev_score = post_score(awardee_id, post["id"], awarder_id)
+	action(awarded[awardee_id][post["id"]][awarder_id], value)
+	score      = post_score(awardee_id, post["id"], awarder_id)
 
+	emit_change_line(post, awardee_id, awarder_id, prev_score, score)
 	update_confirmation(post["id"])
 
 
+def award_if_appropriate(reaction):
+	process_change(
+		reaction,
+		lambda values_set, value: values_set.add(value)
+	)
+
 def retract_if_appropriate(reaction):
-	if not reaction_qualifies(reaction):
-		return
-
-	post = get_post(reaction["post_id"])
-	if parse_mm_timestamp(post["create_at"]) < SINCE:
-		return
-
-	awardee_id = post["user_id"]
-	awarder_id = reaction["user_id"]
-	awarded[awardee_id][post["id"]].discard(awarder_id)
-
-	if not awarded[awardee_id][post["id"]]:
-		del awarded[awardee_id][post["id"]]
-	if not awarded[awardee_id]:
-		del awarded[awardee_id]
-
-	awardee = get_username(awardee_id)
-	awarder = get_username(awarder_id)
-	print(f"{awardee} {post['id']} verification removed by {awarder}", flush=True)
-
-	update_confirmation(post["id"])
+	process_change(
+		reaction,
+		lambda values_set, value: values_set.discard(value)
+	)
 
 
 def get_posts_for_channel(mmapi, channel_id, since, **kwargs):
@@ -222,6 +235,19 @@ def persevere(f, backoff=1):
 		sleep(backoff)
 
 
+# awarded[awardee][post_id][verifier]: set of values
+def count_verifications(user_id):
+	return sum(
+		max(
+			(
+				max(values, default=0) for values in post_verifications.values()
+			),
+			default=0
+		)
+		for post_verifications in awarded[user_id].values()
+	)
+
+
 def update_confirmation(post_id):
 	if not CONFIRMATION_EMOJI:
 		return
@@ -229,7 +255,7 @@ def update_confirmation(post_id):
 	post = get_post(post_id, force_fetch=True)
 	remove_reactions_from_post(post)
 
-	new_count = len(awarded.get(post["user_id"], []))
+	new_count = count_verifications(post["user_id"])
 
 	if new_count > 0:
 		persevere(lambda: mm.create_reaction(our_user_id, post_id, confirmation_emoji_name(new_count)))
@@ -268,8 +294,8 @@ def handle_live():
 
 
 if clean:
-	for post in get_posts_for_channel(mm, channel, SINCE):
-		remove_reactions_from_post(post)
+	for _post in get_posts_for_channel(mm, channel, SINCE):
+		remove_reactions_from_post(_post)
 
 else:
 	# Note: skipping this step and updating an existing file would be dangerous: you would miss revocations that happened while not listening.