Cashier order read-back scripts - four error-prone fields (hot/iced, milk, shots, sweetness), each locked with one line

You are the cashier trainer at a coffee shop. Turn the
order read-back lines below into a verbatim confirmation
card - one standard line per error-prone field - for
the register staff to keep taped inside the counter.

Background: coffee orders carry many modifiers - hot/
iced, milk (whole/oat/skim), shot count, syrup and
sweetness level, cup size - and one miss remakes the
whole drink. The register read-back is the first gate:
three extra seconds against a remake's beans, milk,
and two minutes of rush-hour queue.

Standard confirmation lines, four fields, fixed order
by cost-of-error (hot/iced and milk mistakes scrap the
cup, so they go first):
1. Hot/iced: "Quick check - iced or hot?"
   - If the customer already said "hot latte," don't
   re-ask; just carry the word "hot" in the read-back
2. Milk: "Default milk okay, or oat?"
   - Ask only when the customer didn't mention milk;
   if they did, read it back
3. Shots: "Standard single shot - want an extra?"
   - Offer only on medium and large; never on small
4. Sweetness: "Standard sweetness okay?"
   - Ask only for drinks with syrup; our "standard"
   means 【】 pumps - answer honestly if asked

Full read-back (at the close of ordering, one breath):
"Confirming: large oat latte, hot, extra shot, half-
sweet vanilla - 【】 total. For here or to go?"
Points: carry all four fields inside the drink name;
if the customer shakes their head or changes something,
repeat only the changed item, never the whole ticket.

Two fallback moments:
- Impatient customer ("yeah, that's fine"): compressed
  version locks the two highest-risk fields only -
  "Oat milk, iced, right?" - everything else unspoken;
  the cashier decides when to switch, no time rule
- Misheard or heavy accent: "Sorry, say the cup size
  and milk one more time - those two are the costliest
  to get wrong." Speaking the reason out loud makes
  customers slow down and repeat willingly

Ticket change (already ordered and paid):
"Changing it to 【changed item】 - that's 【】 back to
you, or 【】 more. I'll reprint, this is the ticket."

Constraints:
- No "check carefully" or "stay focused" instructions -
  nothing that can't be read aloud
- Zero upselling on this card - read-back is error
  prevention; recommendations are a different card, and
  two goals on one card fail both
- The compressed version never drops milk and hot/iced;
  everything else is optional

Why it is written this way

The four fields are ordered by cost-of-error, not menu order: a wrong hot/iced or milk call scraps the cup, shot count and sweetness mean a remake, a wrong cup size means a swap. The first two lines always get spoken; the rest depend - error-prevention effort goes where the scrap cost lives, which is also why the compressed version keeps exactly those two.

The hardest execution question in read-back is "what if the customer already said it": skip it and their slip of the tongue never gets a chance to self-correct; re-ask everything and they think you weren't listening. The compromise lives in line one - read back the keyword instead of re-asking (the word "hot" carried along), which leaves the self-correction opening without spending an extra question. After the full read-back, only the changed item gets repeated, for the same reason: keep the cost of confirming low enough that people actually do it during a rush.

The misheard fallback speaks its reason out loud ("costliest to get wrong") because it translates a procedure into the customer's interest: read-back works only with customer cooperation - a rule-flavored repeat gets ignored, a "this protects you" repeat gets the customer to slow down and say it again. It is the highest-converting line on the card.

Banning upsells isn't purity: read-back aims at a confirming nod, selling aims at a bigger ticket, and both destinations in one sentence dilute each other - the read-back turns rambling, the pitch turns nagging, and accuracy drops on both. Two separate cards make each effect measurable.

Variables