A taqueria that normally fields thirty phone calls on a Saturday will field a hundred on Cinco de Mayo, most of them inside a four-hour window, and most of them asking the same question. The kitchen plans for the volume. The phone almost never does.
The result is predictable and mostly invisible. The dining room is full, the tickets are flying, the day looks like a success, and the calls that rang out during the peak leave no record anywhere. You cannot miss what you never saw.
The volume is not the problem. The shape is.
Daily totals are the wrong way to think about this day.
Your kitchen can produce a lot of food across twelve hours. What it cannot do is absorb three hours of compressed demand across both the dining room and the phone at the same time, with a host stand that is also managing a wait list on the sidewalk.
Look at your busiest hour, not your busiest day. If you are answering seven calls in an hour normally and twenty-five arrive between five and six, the gap is not something a person absorbs by trying harder. Whoever is holding the phone is choosing, call by call, between the person in front of them and the person on the line, and the person in front of them wins every time. That is the right call to make and it is also how the leak works.
Leak one: the wait-time question
The single most common call on this day is some version of "how long is the wait", and it is a call that produces no revenue and consumes a host who is standing in front of a line.
It is also the easiest thing to take off a person. A current quoted wait, spoken in the first ten seconds of the call, ends most of those conversations. Callers who hear ninety minutes and decide to order takeout instead have just converted themselves into a ticket. Callers who hear ninety minutes and go elsewhere were never going to wait anyway, and you learned that without spending a host's attention on it.
The thing that makes this work is accuracy. A quoted wait that is fifteen minutes optimistic is worse than no quote, because you now have people arriving on a false promise and standing in your entryway getting annoyed. Update it hourly at minimum, from whoever is actually running the door.
Leak two: the takeout order that has nowhere to go
The second leak is takeout orders taken with no capacity control behind them.
On a normal night your kitchen absorbs phone orders into the gaps. On Cinco de Mayo there are no gaps. Every phone order competes directly with a table that is already seated, and if you keep saying yes, both channels degrade at once: the dining room slows down and the takeout orders come out twenty-five minutes past their quoted time.
The fix is a cap per pickup window rather than a cap per day. Decide how many takeout orders your line can push into a fifteen-minute slot alongside dine-in tickets, and stop offering that slot when it is full. It feels like turning away money. What it actually does is protect the quoted time on the orders you did accept, which is the thing customers remember. The mechanics of holding that line are in order throttling and kitchen capacity.
Party trays are a separate queue
Large pickup orders on this day are worth chasing and they need to be handled differently from the twenty-dollar tickets. They arrive earlier in the day, they need a real conversation, and they should be capped at a much lower number per window because a fifty-person order occupies your line completely while it is being built.
Route them to a person, take them in advance, and stop accepting them at a cutoff you announce. A tray order taken at four in the afternoon for a six o'clock pickup is a decision to run your dinner service badly.
Leak three: language
If a real share of your callers speak Spanish and your phone cannot handle a Spanish order, those calls end in a hang-up and you never learn they happened. This is true on a Tuesday in February too, but on a day when your volume triples, so does the number of people who quietly go elsewhere.
Language coverage is not a holiday feature. It is a property of your line that gets stress-tested on this particular date. If you have been putting it off, the decision is laid out in bilingual Spanish AI phone ordering and the wider version in multilingual phone ordering.
Leak four: everything you configured in April and forgot
Special hours. A special menu with items that only exist on this day. A different closing time. A limited menu after nine. Each of those lives in your POS, your website, your delivery listings, and whatever answers your phone, and they do not update each other.
The specific failure to watch for is a special item that sells out at seven and keeps being offered by the phone until close. If your menu availability is not syncing live, someone has to manually 86 it in every channel, and the phone is the channel that gets forgotten. That whole category of failure is worked through in real-time 86ing and menu sync.
What to do in the two weeks before
Pull last year's call log for the day and count total versus answered by hour. That number, more than any argument, tells you whether the phone is a real leak for your restaurant or a minor one. The method is in how many calls does your restaurant miss.
Then set four things: your special hours everywhere they appear, your takeout cap per window, your large-order cutoff and who takes those calls, and the exact wait-time message with a plan for who updates it hourly. None of that is complicated. All of it is the kind of work that does not happen on May 4th.
The same structure applies to any single-day surge with a fixed date, which is why the game-day version in Super Bowl and game day phone orders reads nearly identically. If you are running a Mexican concept year-round, the everyday version of these decisions is in voice AI for Mexican restaurants and taquerias.
The measure of whether you got it right is not revenue on the day, which will be up regardless. It is your answered-call rate between five and eight, compared to a normal Saturday. If it dropped, you lost orders you will never see on a report.