Latest / The Operator's Manual with Fexingo: Daily Decisions, Tactical Wins, and Business Execution / How One Bank Cut Call Center Escalations 30 Percent With One IVR Change
Transcript
- Lucas: If you've ever called a bank and spent three minutes fighting a voice menu only to end up shouting 'representative' at the phone, you already understand today's episode. We're talking about one specific change a regional bank made to its IVR — the interactive voice response tree — that cut escalations to a live agent by thirty percent. Luna: Thirty percent, just from changing the menu. So the fix was not adding a better virtual assistant or hiring more people. It was purely about how the options were structured. Lucas: Exactly. And the bank in question is a mid-sized regional player in the Southeast — I'm not naming them because they never publicized this internally. But one of their operations analysts, a guy named Derek, shared the data in a practitioner forum I follow. The original menu had nine options on the first layer, plus sub-menus that could go three or four levels deep. Luna: Nine options — that already violates the rule of seven plus or minus two, right? Cognitive load theory says people struggle to hold more than about seven items in working memory at once. Lucas: That's exactly the problem. And Derek told me that when they mapped the call-flow data, they found that sixty-two percent of callers who reached the third sub-menu simply hung up and called back — which of course reset the whole process. The escalation rate wasn't driven by complex issues. It was driven by people getting lost. Luna: So what was the single change? Did they flatten the menu entirely? Lucas: They did something counterintuitive. Instead of giving callers more paths to self-service, they reduced the first-layer options from nine to four. They collapsed account services, lost cards, fraud alerts, and general inquiries into just two branches: 'Something wrong with my account' and 'Something I want to do.' And they added a fifth option: 'Talk to someone now' — which sounds like it would increase escalations, but actually did the opposite. Luna: Interesting. So the 'talk to someone' button acted as a pressure valve. People who would have been frustrated by a long menu just pressed it early, but they were a small group. The bigger effect came from the people who actually stayed in the new tree. Lucas: Right. The new tree had only two levels for most requests. For example, under 'Something wrong with my account,' the next prompt was: 'Lost or stolen card, press 1. Unauthorized charge, press 2. Other problem, press 3.' That's it. No fourth level. Derek's team measured the average time to completion for routine tasks like reporting a lost card dropped from two minutes and forty-five seconds to forty-eight seconds. Luna: And the escalation rate went from twenty-two percent of all calls to just over fifteen percent. That thirty-percent reduction. Did they track customer satisfaction alongside it? Lucas: They did. Post-call survey scores improved by twelve percentage points. The really interesting finding, though, was that customers who used the 'Talk to someone now' button actually reported higher satisfaction than customers who used the old menu — even though they ended up talking to a person. Derek suspects it's because the decision to escalate felt like their choice, not a failure of the system. Luna: That's a classic behavioral economics insight — giving people an explicit escape hatch reduces frustration even if they don't use it. The mere option makes the process feel less like a trap. Lucas: By the way, this is exactly the kind of thing we dig into on this show — one specific operational lever that someone pulled and the numbers that resulted. We deliberately don't run ads on these conversations. If you want to support that choice, the link is buy me a coffee dot com slash fexingo. No pressure, just a way to keep it ad-free if it matters to you. Luna: Yeah, and honestly the fact that we can talk about a thirty-percent reduction in escalations without having to pitch a product in the middle is part of why this format works. Appreciate anyone who chips in. Lucas: Back to Derek's data. The other thing they did was remove the 'main menu' option during sub-menus. In the old system, if you pressed nine at any point, you'd go back to the beginning. That sounds helpful, but it actually created a loop: people would accidentally press nine, get dumped back at the top, and restart the whole thing. Derek's team found that ten percent of all calls involved at least one unintended return to the main menu. Luna: So by removing the ability to go back, they actually made navigation easier. People had to go forward, but the forward path was so short that it didn't matter. Lucas: Exactly. They replaced the 'main menu' option with a simple 'go back one step' option. That alone cut the average number of keypresses per call from eleven to six. And the error rate — where a caller selected the wrong option and had to start over — dropped by forty percent. Luna: Were there any unintended consequences? Like, did more complex issues get mishandled because the menu was too simple? Lucas: That was the big fear. But Derek's team tracked the outcome of every call that did get escalated. They found no increase in repeat calls for the same issue, and no increase in average handle time for the agents who took those calls. In other words, the calls that required an agent were still the same kind of calls — they just weren't being buried under a pile of navigation failures. Luna: So the thirty-percent reduction came almost entirely from people who had been escalating because the menu frustrated them, not because their problem was genuinely complex. Lucas: That's the headline. Derek's rough estimate was that about one in five escalations in the old system were 'false positives' — calls that could have been handled by self-service if the customer had been able to find the right option. By simplifying the tree, they captured most of that lost traffic. The cost savings were significant: they calculated roughly three hundred thousand dollars a year in reduced agent time. Luna: For a mid-sized bank, that's real money. And the change itself took about two weeks to implement, because they didn't touch the backend — they just reconfigured the voice prompts and the menu logic. Lucas: Right. And this is the kind of operational win that often gets overlooked because it's not flashy. It's not AI. It's not a chatbot. It's just understanding how your customers actually behave and removing the friction that makes them give up. The same principle applies to website navigation, to in-store signage, to any interaction where a user has to make a choice. Luna: It's also a reminder that more options is not better. The bank's original menu was designed by someone who wanted to be thorough — every possible reason for calling got its own button. But that thoroughness actually made the system worse. Less was more. Lucas: And Derek told me the hardest part was convincing the product team to let go of some of those options. They had stakeholders who insisted that 'report a lost debit card' and 'report a lost credit card' had to be separate buttons, because they were different departments internally. Derek had to show them that callers didn't care about the internal org chart — they just wanted to stop the bleeding. Luna: So the organizational friction was actually harder than the technical change. That's a lesson in itself: process improvements often require someone to give up a little control. Lucas: Exactly. And once the data came in, the product team was convinced. The satisfaction scores and the cost savings made the case. But it took a pilot — they ran the new menu on just ten percent of incoming calls for two weeks — to gather the evidence. Luna: That pilot approach is smart. Low risk, quick results. Any team could do the same with their own phone system or web form. Lucas: So the takeaway from this episode is pretty direct: if you manage any customer-facing flow — IVR, website, app, even a physical checkout — look at where people are dropping out or escalating. Then ask yourself whether the structure of the choices is helping or hurting. And be willing to cut options, even if it feels like you're losing control. Luna: One number to remember: sixty-two percent of callers who hit the third sub-menu hung up and called back. That's a diagnostic signal. If you see that pattern, your menu is too deep. Lucas: And if you want to dig deeper into the data, Derek published a longer case study on his personal site. We'll link to it in the show notes. For now, that's today's tactical win. We'll be back tomorrow with another one. Luna: See you then.