The 60-second answer
Design a phone menu from the reasons people call, not from your org chart. Three to five options, one level deep, business name in under five seconds, an explicit route to a person, and no dead ends. Most small-business menus fail because nobody listed the actual reasons first.
The mistake almost everyone makes
Ask a business to design its auto-attendant and it will produce a menu that mirrors the org chart. Press 1 for Sales, 2 for Service, 3 for Accounts, 4 for Administration.
Now consider the caller. They do not know your departments. They have a reason for calling — a broken furnace, an invoice they do not recognise, a question about whether you cover their town — and they are being asked to translate that reason into your internal structure while a recording plays. Every mismatch produces either a hang-up or a call landing in the wrong place, which costs an internal transfer and a repeated explanation.
The fix is to invert the design. Spend an hour listing the actual reasons people call, in the words they would use. Most businesses find that four or five reasons cover the overwhelming majority of calls, which is convenient, because four or five is roughly the number of options a caller can hold in their head.
A menu built from reasons routes calls correctly. A menu built from departments routes calls to whoever answers first and makes the caller explain twice.
Six rules that do most of the work
- Name the business in under five seconds. Then stop. Callers dialled you deliberately; they do not need the tagline, the hours, the website address and the mission statement before they can act.
- Three to five options at the top. Beyond five, retention collapses and callers press whatever they heard last.
- One level deep. Sub-menus are where callers give up. If you think you need a second level, you probably need to remove options rather than nest them.
- Say the reason before the number. “For service on an existing job, press 2” beats “Press 2 for service”, because the caller is listening for their situation and only then needs the digit.
- Always offer a person, explicitly. And route silence to that person too.
- No dead ends. Every path terminates in either a human, a voicemail box someone genuinely monitors, or a callback commitment. A path that loops back to the main menu and eventually disconnects is a lost customer with a bad story.
Why menu selections work reliably now
Worth a brief technical note, because older systems earned a poor reputation here. Keypad presses used to be carried as audio tones, which compression could distort badly enough that a system would miss or double a digit. The modern approach carries digits as named telephony events in their own payload rather than as audio [1], so the receiving end gets an unambiguous “the caller pressed 2”. Combined with standard call signalling for the transfer itself [2], a well-built menu today does not lose presses. If yours does, that is a configuration problem worth raising, not something to design around.
Accessibility is not optional decoration
This is the section most menu-design advice omits, and it carries real obligations in Canada.
Federally regulated organizations operate under the Accessible Canada Act, a framework built around identifying, removing and preventing barriers — including barriers in communication [3]. More broadly, discrimination in the provision of services customarily available to the general public is prohibited under federal human rights law [4], and the means by which a service is offered forms part of that service. A phone system is, for many businesses, the primary channel through which service is requested.
What this means in design terms is concrete:
- Never make speech the only input. A caller with a speech disability, a strong accent the system mishandles, or a noisy environment must be able to use the keypad.
- Allow adequate response time. Timeouts that assume a fast keypress exclude people who need longer.
- Speak clearly and unhurriedly. A rushed recording is harder for everyone and disproportionately harder for callers with hearing loss.
- Keep the route to a human short. This is the universal accommodation: whatever the menu failed to anticipate, a person can handle.
- Do not require a smartphone or an app as the only alternative path.
Language
For businesses in New Brunswick, and for federally regulated entities operating under the federal official languages framework [5], bilingual handling is frequently an expectation rather than a differentiator. The design implication is simple: offer the language choice first, before anything else, as a single keypress, and then run the entire flow in the chosen language rather than switching back partway through. A bilingual greeting followed by an English-only menu is worse than an English-only system, because it makes a promise it does not keep.
Hours, holidays and the after-hours path
An auto-attendant with one greeting is a system that lies to callers every evening and all weekend. At minimum you want three states: open, closed, and holiday — with the holiday list actually populated for the year rather than left empty. The closed-hours message should do three things: say you are closed, say when you reopen in plain terms, and offer a route that produces a callback. “Please call back during business hours” followed by a disconnect is the worst available outcome, because the caller has already decided to spend money and you have asked them to try again later.
The privacy angle nobody considers
A menu selection is information about the caller. “Press 3 for collections” or a menu that separates new patients from existing ones records something about a person's circumstances, and where it is logged against their number it becomes personal information governed by the fair information principles [6] — including limits on retention and use. This is not a reason to avoid useful menus. It is a reason to keep menu labels neutral where the category is sensitive, and to know how long selection logs are kept.
Start from your own call log
Most Canadian establishments are small [7], with fewer real departments than their menu implies. Before designing anything, take two weeks of calls and tally the reasons. The exercise usually reveals that one option handles more than half of all calls — in which case it should be option 1 and it should also be where silence routes — and that at least one existing option receives almost nothing and can be deleted outright.
Call-flow and auto-attendant design is included in every MapleReceptionist band, along with business-hours and after-hours routing, an extension per person, Canadian-hosted voicemail with voicemail-to-email, and roughly 30 to 60 minutes of administrative changes each month — which is what it takes to keep a holiday list current. Pricing is $99 CAD per month up to 10 extensions, $219 up to 25, $399 up to 50, $699 up to 100, plus HST.
Bottom line
Build the menu from reasons, keep it to five options and one level, get to the point in five seconds, always offer a person, and populate the holiday list. Then check it against real call data after a month and delete whatever nobody chose. The technology stopped being the limitation years ago [1]; what remains is a design problem, and it is solvable in an afternoon.