Support is a profit center: the evidence

Every support leader has sat through the budget meeting where their team is the line someone wants to shrink. Support gets filed under cost of goods sold, measured in tickets-per-hour, and treated as a tax on the inconvenience of having customers. The framing is tidy. On the evidence, it is also wrong.

The trouble with cost-center thinking is that it only counts one side of the ledger. It sees the salaries and the software licences. It does not see the revenue that walks out the door when the queue is slow, the answer is wrong, or someone has to ask three times. That revenue is enormous, and for once we can put a figure on it.

The bill for bad service is already being paid

That is the Qualtrics XM Institute's estimate of global consumer sales put at risk every year by poor experiences — and it climbed by roughly $119 billion year over year. "At risk" is not the same as "lost," but the mechanism behind it is not mysterious. When people have a bad experience, they spend less.

Just over half of consumers say they will cut or stop spending with a brand after a negative experience. Notice what that does to the cost-center math: the money your team protects never appears as a line item, because retained revenue looks identical to revenue that was never threatened. Support's wins are invisible by construction. That is the whole trap.

Retention is the quiet compounding machine

Churn is where service quality turns into money most directly, because keeping a customer you already have costs far less than winning one you don't. Zendesk's 2025 research found that the companies it calls CX "Trendsetters" — the ones investing seriously in service — reported markedly stronger loyalty than their peers.

A 22% higher retention rate is not a rounding error. Compounded across a book of business, it is the gap between growing and treading water.

You don't need to delight anyone

Here the data does something genuinely useful: it kills an expensive myth. The goal is not to wow people. Qualtrics found that simply moving a customer from a one- or two-star experience up to a merely-fine three-star one made them meaningfully more likely to come back.

Competence, delivered reliably, beats heroics. The same skepticism applies to the recovery playbook. The much-loved "service recovery paradox" — the idea that a great save leaves customers happier than if nothing had gone wrong — mostly doesn't hold up: a meta-analysis in the Journal of Service Research found it lifts satisfaction but has no significant effect on repurchase intent, word of mouth, or corporate image. And when Uber studied 1.5 million riders, monetary make-goods beat apologies, and repeated apologies sometimes did worse than none at all. Fix the problem; don't perform the fix.

People will literally pay more

If retention feels abstract, here is the most concrete finding in the file. When an airline replied to an unhappy customer's tweet, that customer was willing to pay the airline more afterward.

Up to $20 more when the reply landed within six minutes, and only $2.33 when it took over an hour. Same customer, same airline, different response time — and a measurable swing in what they would pay. Speed and attention are not soft skills. They carry a price.

What to do with this on Monday

None of this means support should chase quotas or start upselling in the middle of an outage. It means the framing is backwards. A team that resolves problems the first time, quickly, without making people repeat themselves is not spending money — it is defending a revenue base the P&L is too blunt to show.

So bring numbers to the budget meeting. Put your first-contact resolution and response times next to the retention curve for the accounts your team touches. Measure the cost of customer effort, not just the cost of headcount. The evidence tying service quality to money is no longer thin — and "cost center" is mostly a decision about what you would rather not measure.

For the metric that predicts loyalty best, see the Customer Effort Score reference; for how your numbers compare, start with the support benchmarks.