When a night is consistently dead, the first instinct is almost always the menu. Add a special. Cut a few items. Try a prix fixe. It feels productive because it is something you control directly, and because the kitchen is where you are most comfortable solving problems.
The trouble is that slow nights are rarely a food problem. If the food were the issue, your good nights would be slow too.
What a slow night usually means
Nine times out of ten it means one of three things. Nobody has a reason to think about you on that specific night. Or the people who would come do not know you are open. Or the occasion that fills your busy nights simply does not exist on that one.
That last one matters more than people expect. A place that runs on date nights and celebrations is going to struggle on a Monday no matter how good the special is, because nobody is celebrating anything on a Monday. You are not competing with the restaurant down the street. You are competing with somebody's couch.
"You are not trying to win Monday. You are trying to invent a reason for it."
Build the occasion, then market it
The places that fix slow nights usually do it by creating a reason that has nothing to do with the food being cheaper. Industry night for the restaurant workers who are off when you are slow. A standing event that people can plan around. A supper club. Something with a name, a time, and a reason to tell somebody else about it.
It has to be repeatable. A one off event brings a crowd once and teaches you nothing. The same thing every week for three months builds a habit, and habits are what fill rooms.
Then give it somewhere to be seen
This is where marketing for restaurants stops being about pretty photos and starts being about repetition in the right places. Local press covers recurring events far more readily than one offs. Regulars will show up if you tell them, which means email and social are doing real work rather than decoration. And a slow night with a name is something a food writer can put in a roundup.
One caution. Do not launch four things at once across four nights. Pick the worst night, build one thing, give it a full quarter, and measure it honestly. If it does not move after twelve weeks, kill it and try something else. Most operators quit at week three, which is right about when people were starting to notice.