Nobody designed the touch-tone menu to irritate customers. It was a sensible answer to a hard constraint. You had more callers than people to answer them, and no way to learn what a caller wanted until someone picked up and asked. So the menu sorted callers into buckets before a human minute got spent on any of them.
Everything the menu does follows from that. The options mirror your departments, because departments are how human attention was allocated. The levels deepen as the org chart deepens. The caller does the sorting, because at the moment they call, the sorting is the expensive part.
The constraint is the part that changed. Answering is no longer scarce or scheduled. The menu is still sitting on the line, though, and so is every cost it was built to justify. This is about who pays those costs now, what a real IVR replacement changes, and which parts of the old menu are still worth keeping.
What the phone menu was actually built for
Strip away the hold music and a phone menu does one job: pre-classification. It turns an unknown caller into a known category before a person is committed to the call. That was valuable when the person was the bottleneck, and it is why the design has survived decades almost unchanged.
You can see the logic in how menus get built. Look at the order of your own options. If they follow your departments rather than how often each one is actually chosen, the menu is a map of your internal structure. When a department splits, the menu grows a level. Each of those choices is correct if the job is to protect scarce human attention.
A phone menu is an org chart read out loud to someone who has never seen it.
The unpaid work you hand the caller
Before a caller is allowed to say what is wrong, the menu asks them to do a translation job. Is a duplicate invoice billing, or accounts? Is a part that arrived broken a warranty issue, a returns issue, or technical support? You know the answer because you know your own structure. The caller is guessing, with a one-keypress commitment and a penalty for getting it wrong.
That is real work, and it has been handed to the person with the least information about how your company is organised. Each added level means holding a longer option list in memory and mapping a messy problem onto tidy categories, under time pressure, while a recording keeps talking.
The business then pays for that work twice. Once in the caller who gives up in the middle. Once in the caller who guesses wrong and lands in the wrong queue, where a person has to greet, listen, apologise and transfer, after which the receiving agent starts over.
Two bills for the same menu
A caller who abandons costs you the call. A caller who mis-routes costs you two rounds of live handling before the real conversation even starts.
The second bill is worth pricing, because the arithmetic is easy to check against your own line. Count the calls transferred because they arrived in the wrong place, and multiply by the extra live minutes each one burns. Assume an illustrative 200 mis-routes a day, and substitute your own count, at two extra minutes each: that is 400 live minutes a day, at your own fully loaded cost per live minute, spent undoing the classification you asked the caller to perform.
Where the call really dies
Ask your phone system what the menu costs and it probably cannot tell you, because the menu sits upstream of most of what it measures. Calls answered, average handle time, first-contact resolution: all of them start counting after routing has already happened.
It is worse than invisible. A caller who abandons inside the menu often never becomes a record anywhere, so the loss registers as an absence in a report nobody reads. Where a transfer is counted as a new answered call, a mis-route registers as two of them.
If your reporting counts a transfer as another answered call, a mis-route looks like productivity.
To count this properly you have to compare what the caller said in their first sentence with where the call ended up, and that comparison does not work on a sample. On our own published figure, manual QA reaches 3 to 5% of calls; Locator reviews and scores 100% of them, at EUR 0.10 a minute excluding VAT. Menu abandons leave a thin trace or none at all, but every mis-route is recorded.
What an IVR replacement actually looks like
The alternative is not a better menu. It is a different order of operations. The caller says what they want in their own words in the first sentence, and routing becomes a consequence of understanding rather than a precondition for it.
One sentence holds more than a chain of keypresses can. An opening line like "tracking says my order was delivered and it is not here" says what the problem is, in the caller's own words. A keypress says only which department they guessed.
Routing should be something the system concludes, not something the caller is interrogated for.
For that inversion to hold up on a live line, several unglamorous things have to be true.
- The reply has to be immediate. Benerra's agents answer in about 0.3 seconds. Let the pause stretch and the caller starts talking over the agent, and turn-taking collapses. Latency is not a comfort metric here; it decides whether the conversation works.
- Interruptions have to be handled. People cut in mid-sentence and add details late, and the agent has to stop mid-word and listen. A touch-tone menu tolerates interruption in one form only, a digit.
- Understanding has to be confirmed out loud. The agent reads a delivery address back before committing to it. It also detects voicemail, so talk time is not spent on an answering machine.
- The audio has to stay where your data has to stay. Speech recognition, synthesis and the language model run on Benerra's own servers, so call data stays in-region and does not depend on third-party APIs.
The second change is the one that shows up in your budget. Bene Hotline, our inbound reception and routing agent, is EUR 0.20 a minute excluding VAT. Billing follows talk time, and there is no subscription: the length of the conversation and the cost of the call are the same number.
What to keep from your old menu
Menus are not always the wrong tool. In a few situations a fixed, deterministic message beats a conversation, and pretending otherwise would cost you something real.
- Tiny, stable option sets. Two destinations that have not changed in years: one level, one keypress, and nothing a conversation would usefully add. A menu's cost sits mostly in its depth, and at one level there is very little of it.
- Announcements you are required by law to make. Where your regulator asks for a fixed notice, whether that is a recording notice, a rate disclosure or a regulated warning, a recording is the simple way to deliver the same wording on every call.
- After-hours emergency routing. A single hard-wired path to an on-call number is worth having precisely because it depends on understanding nothing. Keep it as a fallback, not as the front door.
Keep two more things: the existing keypress path on the line behind the agent, for callers on a bad line or who would rather not talk to software, and an obvious route to a person, stated in the greeting rather than buried a few levels down.
Keep the announcement, drop the interrogation
The parts of a menu worth keeping are the parts that broadcast something fixed. The part worth removing is the part that asks the caller to classify their own problem.
Migrating without breaking your numbers
The usual failure is a rip-and-replace that changes routing and measurement in the same week, after which nobody can say which change caused what. A safer sequence keeps your destinations and your reporting stable while the front of the call changes.
Put speech analytics on the line before you touch routing, and compare the caller's opening sentence with the extension the call ended on. This is stage one of our recommended rollout order anyway.
Change the front of the call before you change what sits behind it. Bene Hotline handles inbound reception and routing, and your existing destinations are a separate decision you can take later. Agree the cut-over and the fallback path with whoever runs your telephony before you start, rather than assuming either.
After hours, and peak overflow. There the honest comparison is not a menu, it is an unanswered phone.
Remove one level of depth at a time, in the branch where the agent has proved it routes correctly. Depth is the expensive part.
Containment, the share of calls the agent finished alone, is a vendor's metric. Mis-routes, transfers and repeat calls are the ones your customers feel.
The menu was a correct answer to a shortage of human attention. Before you renew that platform for another year, check whether the shortage still exists on your line, and what you are paying to keep protecting it. Per-minute pricing is public on the homepage, and the public cost calculator takes your call volume, region and product.