The most expensive part of a point of sale is not the software or the hardware. It is the fortnight where everybody is slower, and the size of that fortnight is decided almost entirely by how the training was done.
Most restaurants train by putting someone next to the screen during a shift and hoping. Here is a better version, and it is not much more work.
Train the sequence, not the buttons
A new server does not need a tour of the system. They need to be able to do one thing: take a table from seated to paid without help.
So train the sequence, in order, and make them do all of it:
- Seat the party and open the check
- Ring in drinks
- Ring in food with courses
- Add a modifier, including an awkward one
- Add something after the order has been sent
- Void something that has not been made
- Split the check by item
- Take two cards and a cash payment on one table
- Add a tip and close it
Nine steps. Somebody who has done all nine once is roughly three times faster on their first real shift than somebody who has watched.
Do it away from service, then do it again in service
Session one, closed restaurant. Everyone works through the nine steps on their own, on a real device, with a real menu. An hour. Not a demonstration — a demonstration teaches nobody anything, because watching is not the skill.
Session two, staff meal. Order and serve staff food through the new system, with the kitchen participating. This is where you find out that the printer is not routing, that the modifier list is missing something everyone asks for, and that two of your servers have not actually understood splitting.
The second session is the one people skip and it is the one that pays.
Train the hard cases explicitly
The easy path gets learned by repetition. The hard cases get learned in front of an annoyed customer unless you teach them.
The list, for most restaurants:
- The eight-person split. Unevenly, with shared items. Genuinely fiddly everywhere
- Moving items between checks. After the fact
- A comp versus a void, and which is which. Staff who are not clear will use whichever one has fewer clicks
- A table that splits or merges mid-meal
- A card that declines. What you say and what you press
- A refund, and who is allowed to do one
- A gift card with a partial balance
- The end of the night, for whoever closes
Ten minutes each in a quiet moment, once, is worth more than any amount of on-the-job absorption.
Decide who is allowed to do what
Permissions are a training tool as much as a control.
Voids, comps, refunds, discounts and reopening a closed check should require an authorisation from whoever is accountable for them. Not because staff are dishonest, but because an unexplained comp is indistinguishable from a mistake and both need to be findable afterwards.
Set this up before training rather than after, so the boundaries are learned as part of the normal flow instead of arriving later as a restriction.
Build the two people who answer questions
In week one, most questions are best answered by a colleague standing next to the asker. Support is for things nobody on the floor knows.
So pick two people who took to it, tell them explicitly that they are the floor's first line, and make sure they are on the shifts that matter. It is the single highest-return thing you can do for the first fortnight, and it costs nothing.
Do not train on a menu you have not finished
An obvious-sounding trap that catches people constantly.
If the menu is still being built, training on it teaches staff a layout that is about to change, and it teaches them that the system is unreliable. Finish the menu, have the chef check the modifiers and a server check the ringing-in order, then train.
What to measure afterwards
Two numbers tell you whether the training worked.
Void and comp rates in week one versus week four. A high week-one rate is normal and is mostly people correcting their own ringing-in. If it is still high in week four, something specific is confusing and it is findable — look at what is being voided.
Ticket times. If the kitchen is slower than before, the floor is probably ringing in badly rather than the kitchen being slow. Coursing that is not being used is the usual cause.
Both of those are in the reports. Neither takes more than five minutes to look at.
For new starters after go-live
Write the nine steps on one page and keep it with the onboarding paperwork. Most restaurants train their launch cohort carefully and then train everybody hired afterwards by osmosis, which is how a system drifts back into being used badly.
One page, twenty minutes, on their first shift. That is the whole ongoing cost of keeping it right.