Order confirmations
An order confirmation is mostly a reference number, and a reference number read as a quantity is a reference number the customer cannot repeat back to you.
What goes wrong
Order references are the worst case for general-purpose text-to-speech: long, mixed letters and digits, and meaningless as a quantity. ORD-4402-19 has no correct reading as a number, so a model that tries to find one produces something the customer writes down wrongly and then quotes back to your support line.
Delivery windows carry a second problem. "Between 9 and 12" is a range, not two separate numbers, and the pause between them is what tells the listener that.
Scripts you can use
Written to be read aloud by a machine, not by a person. The note under each one says which part a general-purpose voice model gets wrong.
Order placed
Thank you for your order with Northwind Supply. Your order reference is O R D, 4 4 0 2, 1 9. Please quote it if you contact us about this order.
Grouping matters as much as per-digit delivery. Reading twelve characters as one unbroken string is technically correct and practically useless — group them the way the reference is printed.
Delivery window
Your delivery is booked for Thursday the 12th of March, between 9 in the morning and 12 midday. Someone must be present to sign for it.
"9 to 12" needs to be heard as a span. Writing it out with "between" and "and" gives the model the structure to place the pause correctly.
Part-shipped
Two of the three items in your order have shipped. The remaining item, product code A 9 1 4, is expected within five working days.
"Two of the three" are quantities; A 914 is a code. The same sentence needs both readings, which is exactly the case a single global setting cannot handle.
Before you record it
- The reference is grouped in the audio the same way it is grouped in the confirmation email.
- Letters in a reference are spoken as letters, not folded into a word.
- Delivery windows use "between X and Y", not a bare range.
- The call states what happens next, not only what happened.
Questions
How should an order reference be read out on a phone call?
Character by character, grouped the way it is written. ORD-4402-19 is read as "O R D, four four zero two, one nine", with a pause at each hyphen. Grouping is what lets the listener write it down without asking for a repeat.
Can the same recording read some numbers as quantities and others digit by digit?
It has to. A confirmation call routinely contains both — "two of the three items" alongside "product code A 914". Handling this per field, rather than with one setting for the whole message, is the difference between a usable confirmation and a confusing one.
Generate these with IVR Voice Gen
IVR Voice Gen is text-to-speech built for phone systems: digit strings, references, dates and addresses are read the way a caller can write them down, in English, French, German, Spanish and Portuguese. It is in early access.