What does your customer service really cost? Calculating it per query

Ask any business owner what their customer service costs and you'll almost always get the same answer: number of agents times salary. That figure isn't wrong, but it's the least useful number you can have about your support operation. It tells you what you spend, not what you get back, and certainly not where the money actually goes.

The number that does steer decisions is cost per customer query. It's the only measure that lets you do two things a payroll line never can: identify which types of questions cost your business disproportionately, and calculate whether an investment in automation, self-service or an extra agent pays for itself.

This article explains how to calculate that cost, including the items almost everyone forgets, and what to do with the outcome. All figures below are worked examples to demonstrate the method, not industry benchmarks. The whole point is that you replace them with your own.

The calculation most companies make looks like this: take total monthly staff costs for the service team, divide by the number of queries handled that same month. A team of three costing 12,000 euros a month and handling 2,000 queries lands at six euros per query.

That outcome is almost always too low, for three reasons. First, it uses gross salary rather than full employer costs. Social contributions, pension, holiday pay, insurance and workplace costs add substantially to the gross figure. Include them and the staffing line rises considerably without a single extra query arriving.

Second, it treats all hours as productive hours. Someone on a 40-hour contract does not spend 40 hours on customer queries. Meetings, training, sick leave, holidays, admin and the waiting time between queries all come off. Counting those hours as if they were spent on queries artificially lowers your cost per query.

Third, it ignores everything that isn't salary. Software, telephony, shipping costs on goodwill deliveries, return costs you absorb to keep a customer happy: they sit on other budget lines and disappear from view, even though they are genuinely costs caused by customer contact.

The fairest way to determine cost starts with an hourly rate that holds up. Take the full employer cost of an agent per year and divide it not by contract hours, but by the hours actually spent on customer queries.

A worked example. Say an agent costs you 55,000 euros a year including all charges. On paper that's roughly 2,080 contract hours. Subtract holidays, public holidays, sick leave, training and internal meetings and many teams are left with something around 1,500 hours. Of those, a portion still goes to work that isn't a customer query: reporting, projects, updating content. Land on 1,300 productive hours and your real hourly rate is over 42 euros instead of the 26 euros contract hours would suggest.

That gap is the heart of the matter. Anyone calculating with contract hours believes a ten-minute query costs a little over four euros. In reality those same ten minutes cost over seven. At two thousand queries a month that's a sixty thousand euro annual difference in your picture of reality, while nothing about the operation has changed.

The point isn't that the higher number is depressing. The point is that the lower number leads to wrong decisions. Every judgement about automating, outsourcing or hiring gets made against a cost that isn't real.

Once you have a reliable hourly rate, the next step gets more interesting than the first: distinguishing between types of queries. An average cost across all queries hides exactly the information you want to act on.

In almost every webshop the largest volume consists of short, repeated questions: where is my order, can I exchange this, when will it arrive. Individually they're cheap, but their volume often makes them the biggest share of total costs. Alongside them sits a smaller category of complex cases: an escalating complaint, a delivery gone wrong, a B2B customer with a bespoke agreement. Those cost a multiple each, but their low count means they add up to less than people assume.

That distinction fully determines your strategy. For a team processing mostly repeat volume, the win lies in removing the question: better carrier integration, clear status information, automated answers to the most common questions. For a team handling mostly complex cases, the win lies in context: making sure the agent sees the whole picture on one screen, so twenty minutes don't disappear into searching.

Invest in the wrong one and it won't work. Automating complex cases yields little because the volume is too low. Extra staff on repeat volume doesn't solve the problem, because the questions keep coming.

Part of the real cost of customer contact sits nowhere near the service team budget. These are downstream costs: spending that exists only because there was a customer query.

The best known is goodwill. A customer calls about a late parcel, and to make it right the agent sends a replacement or offers a discount. Those costs land with logistics or against margin, not with customer service. Yet they were caused by the contact moment and belong in the cost per query.

Then there are returns that stem from uncertainty. A customer who calls with doubts and isn't helped well returns more often. That too is a downstream cost. Also counting: telephony costs, software licences, time from other departments consumed by service requests, and the indirect cost of an agent who can't get to sales during a peak.

You don't need to account for this to the last cent. But without an estimate of these items you're looking at a cost that covers only the visible half. In most companies doing this exercise for the first time, the full picture sits noticeably higher than the payroll figure they started with.

Once cost per query is right, the conversation changes. An investment in software or automation is no longer a matter of instinct, but a calculation you can actually make: how many queries need to disappear or be handled faster for this spend to pay for itself?

Say you land on nine euros per query and a third of your volume consists of status questions you could largely remove by making order status automatically available. At two thousand queries a month that's six hundred queries. If two thirds of those fall away, you're talking about four hundred queries times nine euros, every month. Against that outcome any investment decision can be made rationally.

Equally important is what the calculation tells you about where not to invest. If complex cases are only a small share of your volume, building automation for them is pointless, however technically interesting. And if your cost is driven mainly by reopened queries, opening another channel is the last thing that helps: the first answer needs to get better, not the offering bigger.

Ultimately the calculation isn't an accounting exercise but a prioritisation tool. It tells you which of your customer queries cost your business the most, and that is exactly the list you should start with.

What you shouldn't do is use the outcome as an argument to thin out service. Getting cheaper by being less reachable works on paper and costs you customers in practice. Cost per query exists to make expensive questions disappear, not customers.

There's one item that appears in no calculation and is often worth the most money: the query never asked. Once you know cost per query, you can also calculate what it's worth to make a question disappear entirely.

That's a fundamentally different kind of saving from working faster. Handling a query ten percent faster yields ten percent of that one query. Removing a query structurally yields the full cost, every single time, without anyone having to work on it. For high-volume questions the difference between those two approaches is enormous.

The practical application is a simple list. Take your ten most common questions, put cost and monthly volume next to each, and assess per question whether it's avoidable. Avoidable here means: the customer could have known or found the answer themselves if the information had been in the right place.

What usually happens next is sobering. The most expensive category rarely turns out to be a complex problem and is usually a matter of information sitting just off the right place: a delivery window that only appears in the confirmation email, a return condition three clicks deep, a product detail missing from the page and therefore asked about by phone. Those aren't service problems but information problems, and they're cheap to fix once you can express them in euros.

Cost per query is a steering tool for the organisation, not a performance figure for individual agents. The moment someone is assessed on the cost of their conversations, behaviour changes immediately and not for the better: conversations get cut short, investigation gets avoided and difficult questions get passed on.

You only see the effect a month later, in reopened queries and complaints that have grown heavier. The costs haven't gone away, they've moved to a moment where they're more expensive. A conversation that ran five minutes longer because the agent genuinely investigated is almost always cheaper than the three follow-up contacts that would otherwise have happened.

So use the calculation to change processes rather than to compare people. The question it answers is which type of customer query is too expensive for the value it delivers, not who works slowly.