Years ago, I took my London team to Amazon’s Customer Care Center in Edinburgh.
We were there for training, and part of that training was sitting beside the representatives while they took live calls. Some of the calls were about our product, the Amazon Appstore, while others were about retail orders.
One call I will never forget. A woman phoned in about an item that had arrived broken, and she came in armed for bear. Not quite yelling, but intense, loaded, and ready for the fight she was certain was coming.
The representative I was paired with took the customer’s information, pulled up the order, and issued the refund while the woman was still venting. Then she waited, calmly, for the storm to finish. And when it did, she said, in a quaint Scottish accent I can still hear: “Miss, thank you for the information, and I am sorry you have had this issue. I have already issued your refund. You do not need to return the item. Would you like me to order another one for you, or is there anything else I can help with?”
You could almost hear the wind leave the caller’s sails. A sigh, and then gratitude. The fight she had prepared for never happened.
I have admired the poise of that room ever since. The representative absorbed a storm without matching its volume, acted before she spoke, and let the action do the arguing. The quietest person in that exchange held all the power.
But the individual representative is not the whole story. The refund was already processed before the customer finished talking, and that wasn’t just because of the representative’s kindness. It was part of the company’s design.
Not every company is designed like this.
Recently, I spent the better part of two weeks trying to cancel a business phone service.
The website would not let me. Cancellation required a phone call; the phone call required a hold; and the hold led to a person whose job was to offer me discounts I had not asked for before finally, reluctantly, opening a ticket. Even after the ticket was filed, the billing page still showed the renewal charge, scheduled and ready, contradicting the cancellation banner two inches above it.
None of that is an accident. Somebody designed that flow. They looked at the churn numbers, understood exactly what a phone call plus a hold plus a retention script does to cancellation rates, and approved it.
The two experiences I am sharing are not simply different levels of customer service. They are different beliefs about customers, rendered as workflows.
The phone service company believes that the customer is an adversary to be outlasted. Every barrier is a bet that you will give up before they do, and if you are too frustrated to continue the fight, they take that as a win.
The Amazon/Edinburgh model, by contrast, believes the customer is telling the truth. The refund-first design is that belief, operationalized.
Amazon does have (and had) real safeguards against genuine bad actors who were gaming refunds, and those safeguards ran separately, but the default stance was to trust the customer. The design choice was “set the default to trust, and handle the exceptions as exceptions.” When companies set the default to suspicion, every honest customer pays the cost.
This is a leadership decision, not a ground-level customer service quirk.
I know something about where the customer service friction road ends, because I worked at a company that had already driven it.
Years before I joined, AOL was caught making it extraordinarily difficult for customers to cancel their dial-up service. The FTC acted, gave them a fine, and sent down a mandatory training for every single employee on why and how not to do that to customers.
This training even reached me. I joined years after the decision, and I still had to sit through the training built to atone for it.
Think about the accountability arithmetic there. The executives who chose friction were optimizing a retention metric, and it worked, on the spreadsheet, for a while. By the time the consequences arrived, they had moved on. The fine landed on the company. The training landed on people like me, who had nothing to do with it.
The accountability arrived, but at the wrong address. The lesson was absorbed by everyone except the people who needed it.
I think about that whenever I see a company quietly install a cancellation maze. The leadership implementing it knows their decision isn’t harmless; they just know the consequences are deferred. Their retention rates will get an artificial boost, and the brand damage or regulatory action will be someone else’s problem.
There is a second cost of friction that companies also need to take seriously.
On that same Edinburgh trip, we listened to an Appstore call from a customer who had received more than twenty confirmation emails for a single app purchase and was, understandably, worried he had been charged twenty times. He had not been, and the representative confirmed it, but the emails never made that clear.
I took that call back to our teams in London, and when we dug in, we found a genuine edge case in the purchase flow: when the connection to the bank was failing intermittently, retries could each trigger a confirmation email while the charge itself sat unresolved. It was not in our use cases or test plans, and I doubt any amount of internal review would have found it because nothing was technically “broken” from our side of the glass.
The customer’s confusion served as the bug report.
That is what a call center actually is, and what high-friction customer service flows are missing: a fine-tuned sensor array. It is the only instrument a company has that measures reality as customers experience it rather than as systems report it. A company that makes customers fight to reach a human is not just spending goodwill; it is blinding itself to the signals it most needs. Edge cases and confusions that do not show up on dashboards show up in customer service calls. If the customers carrying that signal give up on hold, the company will never receive it.
Friction does not just cost you customers. It costs you information that could have made the product better. A company that makes customers fight to be heard loses its most valuable source of feedback.
And if that mechanism sounds familiar, it should. It is the same filter that operates inside organizations, one layer down. A meeting culture that rewards volume does exactly the same thing to its quietest people, and pays the same price: the edge cases, the early warnings, the truths that do not survive the barrier. Friction and loudness are one filter wearing two uniforms. Organizations that build either one are choosing comfort over signal, and the bill arrives later, itemized as surprises.
One rule I hold personally is to never be critical of the person in the call center, not even the person at the phone company that made me fight for two weeks. They did not write the policy. They were handed a script and a metric, and the friction I am angry about lives several layers above their desk. I make that distinction explicit to them every time.
I only get frustrated with them if they start personally defending the policy. That is the moment they stop executing a decision and start owning it.
Criticize the rule, never the person handed the rule. It is a small ethic, but I have found it maps onto almost everything inside organizations too. The person absorbing the complaint is almost never the person who caused it, and treating them as if they were is both unfair and useless. Accountability lives with whoever wrote the rule. Aim there.
The question I would leave with you is this, and I do not think it resolves cleanly:
Some friction is legitimate. Real bad actors exist, and safeguards against them are responsible engineering. So how does a leader tell the difference between friction that protects the business from genuine abuse and friction that protects a metric from genuine customers?
The uncomfortable answer is that the spreadsheet looks identical either way. Churn improves. Losses drop. The dashboard cannot tell you which kind of friction you built.
The only reliable test I know is the one the woman in Edinburgh passed without thinking: would the person in front of you recognize the policy as fair if you explained it to them plainly? The refund-first design survives that explanation. The cancellation maze does not, which is why nobody ever explains it plainly to anyone, including themselves.
That test cannot be run from the executive floor. You have to sit next to the person taking the call.
