Start with the call, not the brand
Most briefs for phone audio arrive describing the company. Premium, friendly, innovative, trusted. All of it true, none of it useful to the person casting the voice, because the caller is not thinking about your brand. They are thinking about the reason they picked up the phone.
So start there. Are people calling because something has gone wrong? A utility fault line, an insurance claim, an IT helpdesk. Those callers are already tense, and a bright upbeat read lands badly on someone whose boiler has packed in. Are they calling to buy, to book, to enquire? That call can carry more energy. Are they internal, ringing a switchboard forty times a week? Then the priority is that the voice does not grate by the fortieth time.
Write one sentence describing the caller’s state of mind before you write anything about tone. The casting follows from it almost automatically.
The one sentence
Who is calling, and why
“Someone whose broadband has been down since this morning.” That tells a voice artist more than three paragraphs of brand values, and it tells us which of our voices to put in front of you.
Does the gender of the voice matter?
Less than people expect, and not in the way the old received wisdom claims. You will still hear that female voices are inherently more reassuring on phone systems. Treat that as folklore rather than fact. What genuinely matters is whether the voice suits the call, whether it is clear at telephone bandwidth, and whether it matches whatever else your customers hear from you.
Where gender does earn a real decision is consistency. If your advertising, your app and your explainer videos all use one voice, and the phone system uses a different one, you have quietly told the caller that the phone is a separate, lesser part of the company. Using the same artist across all of it is usually cheap, because the prompts are short, and it does more for brand recall than most of what gets spent on brand recall.
Accent: neutral, or local?
A neutral British read is the safe default for a national audience, and it is what most UK phone systems use. It carries no regional signal, which means it also carries no wrong signal.
A regional voice earns its place when the company is genuinely of that place and the caller knows it. A Yorkshire building society, a Belfast council service, a Glasgow trades firm. Then the local voice is not a gimmick, it is a fact about you, and callers hear it as confidence rather than costume.
The trap is casting a regional voice for warmth alone, on a national line, because someone in the meeting liked the sound of it. That reads as a marketing decision, and callers pick up on it.
Pace, and why studio pace is wrong
Phone prompts need to be slower than you think, and slower than the same script would be read for video. The caller is holding a handset, often somewhere noisy, frequently doing something else, and they are being asked to hold numbers in their head while a menu runs. There is no rewind.
Slower does not mean laboured. It means clean separation between the action and the number, a real beat between options, and no crowding at the end of the sentence. A good phone read sounds unhurried to you and merely normal to the caller.
Why one voice beats four
Phone systems accumulate. The main greeting was recorded when the company launched, the new department options were added by whoever was around, the out-of-hours message is somebody’s mobile memo from a snowy Tuesday, and the hold loop came with the phone system. The result is a caller journey that changes voice three times before anyone answers.
Callers notice this more than almost anything else, and what they infer is disorganisation. One voice across greeting, menu, hold, out of hours and voicemail is the single cheapest improvement available to most phone systems, because the recording itself is minutes of work. The cost is in deciding, not in recording.
It also protects you later. If the same artist holds a retainer for your prompts, a single menu change next year matches the original tone, microphone and processing. Without that, one new option means re-recording the set or living with an audible seam.
Audit it yourself
Ring your own number
From a mobile, from outside the building, at half past six in the evening. Then ring it at nine on a Monday. Most people discover at least one message they did not know existed.
What contact centres learn the hard way
In a contact centre the phone audio is not decoration, it is the first stage of the process. Every second a caller spends confused at the menu is a second of handling time, and a proportion of confused callers simply press zero, or press whatever they pressed last time, and arrive at the wrong team. That mis-route costs two agents instead of one, and it usually starts with a menu that was written to describe the org chart rather than to describe what callers want.
The menus that work are written from the caller’s vocabulary. Customers do not ring wanting “customer operations”, they ring because a payment failed or a delivery has not turned up. Naming the option after the problem, in the words the customer would use, does more for routing accuracy than any amount of clever call flow design behind it.
The same applies to queue audio. A queue message that repeats every twenty seconds with the same brisk apology becomes an irritant by the third loop, and callers arrive at the agent already annoyed. Longer gaps, more variation in the lines, and a voice that sounds calm rather than relentlessly cheerful all take heat out of the call before it is answered.
Write for the caller
Name the problem, not the team
“If your delivery has not arrived, press two” routes better than “for customer operations, press two”, because the caller does not have your org chart.
What customer feedback actually flags
Phone audio rarely gets praised. What it gets is complaints, and the complaints are consistent enough to be worth designing around. When companies read back their own call feedback, the same things come up.
- Too many options, held too long. By option five the caller has forgotten option two, and there is no way to look back.
- The number before the action. “Press three for accounts” makes the caller hold a number while waiting to find out what it is for. Action first, number second, every time.
- A different voice at every stage. Greeting, menu, hold and voicemail recorded by four different people over six years reads to callers as a company that does not talk to itself.
- Volume jumps. A loud greeting into a quiet menu into loud hold music has people moving the handset away from their ear.
- Out-of-date information. A Christmas message still running in February does more damage than a plain one, because it says nobody is looking.
- No route to a person. Callers who cannot find a way out of the menu do not give up, they find another number, often the sales line.
None of these are voice problems strictly speaking. They are script and maintenance problems. But they are all fixed in the same session as the recording, which is why it is worth treating the prompt set as one piece of work rather than a file someone uploads.
Listen to demos the right way
The commonest mistake in choosing a phone voice is auditioning it on good speakers. A demo that sounds rich and warm on studio monitors or decent headphones can turn thin and sibilant once it has been squeezed through 8 kHz telephony, because everything above about 4 kHz is simply gone. Sibilance that was pleasant detail becomes a hiss. Low warmth that carried the read disappears entirely.
- Play the demo down an actual phone line before you commit, not through your laptop.
- Ask for your own script, not a showreel. A showreel is edited from an artist’s best moments. Your menu is not.
- Listen for clarity on numbers, because “press five” and “press nine” is where a phone read either works or fails.
- Play it to someone who does not work with you, and see if they can repeat the options back.
- Check the read at the end of a long menu, where energy usually sags.
We record custom auditions against your actual script rather than sending showreels, precisely because this is where the decision is really made. There is no charge for that and no commitment attached to it.
Once you have chosen
Get the whole set recorded at once, including the messages you think you will never need. Out of hours, bank holiday, system fault, department moved, call volume high. Recording ten extra prompts in the same session costs very little. Recording them separately in eighteen months costs a re-book and a voice that no longer quite matches.
When you are ready, our IVR voice recording service delivers prompts in the format your phone system actually needs, named to your admin tool’s convention, with a full-quality master kept for reuse on web and video later.
