Most conversations about reducing support costs begin and end with the rate per hour. It is the number on the quote, so it is the number people argue about. It is also the smallest lever available and the one most likely to damage the service. A team that costs less per hour and resolves less per contact is not cheaper. It is slower, and the cost moves somewhere you are not measuring: repeat contacts, refunds, churn, and the time your own people spend cleaning up.
The larger savings are structural. They come from the number of contacts you handle, the skill level you pay for on each one, where the work is done, and the channel it arrives on. This article works through those levers in the order they usually pay off, then covers the cuts that look like savings and are not.
Start with the shape of the cost, not the rate
Support cost is a product of three things: how many contacts you handle, how long each one takes, and what you pay for the time. Almost every cost programme attacks the third factor first because it is the easiest to see. But the first factor is the one with the most room in it, and it is the only one that improves the customer's experience as it improves the cost. Fewer contacts means fewer customers who had a reason to contact you.
So before you ask any provider for a lower rate, get the contact data. Categorise a month of contacts by reason, count them, and rank the reasons. If you cannot do that today, that is the first job. Our call centre analytics service exists for exactly this, and it is where we start when a client asks us how to spend less.
The categorisation does not need to be sophisticated. A dozen reason codes, applied consistently by agents at the end of every contact, is enough to see the shape. What matters is that the codes describe why the customer contacted you, not what the agent did about it. A code that says refund processed tells you nothing about cost. A code that says delivery notification not received tells you exactly where to look.
The cheapest contact is the one your product made unnecessary.
Lever one: remove contacts at source
A contact that never happens costs nothing and irritates nobody. Once the drivers are ranked, take the top three and ask what upstream change would make each one disappear. The answers are usually unglamorous: a confusing checkout step, a delivery notification that arrives after the parcel, a billing description nobody recognises on their statement, a password reset that fails on mobile.
None of these are support problems. They are product, logistics and billing problems that land in support because that is where the phone rings. The support leader's job here is to own the data and take it to the people who own the fix. A monthly review where the top contact drivers are presented to product and operations, with a named owner for each, does more for cost than any staffing change. It also changes how the support function is seen inside the business: from a cost that absorbs problems to the team that finds them.
- Where-is-my-order contacts point at tracking and notification gaps
- How-do-I contacts point at onboarding and documentation gaps
- Billing-query contacts point at statement wording and invoice timing
- Repeat contacts on one issue point at a first-contact resolution problem, not a volume problem
Lever two: tier the contacts you keep
Paying a technically skilled rate for password resets is the most common way to overspend. Once the avoidable contacts are gone, what remains splits into routine and complex. Routine contacts follow a documented path and need consistency. Complex contacts need judgement, product depth, or the authority to make an exception.
Staff each at the appropriate level and route accordingly. A first line trained on the documented paths handles the routine volume; a smaller, more experienced second line takes what the first line cannot close. The saving comes from matching skill to work, and it only holds if the escalation rule is precise. If first-line agents are unsure what to pass up, they either hold contacts too long or pass up everything, and both outcomes eat the saving.
Tiering also protects quality. Your most capable people stop being interrupted by resets and start spending their time on the contacts where their skill changes the outcome. That is better for the customer with the hard problem and better for the agent who wanted to solve it.
Write the escalation rule as a short list of triggers rather than a description of judgement. A confirmed defect, a request outside the refund policy, a customer on a contract with a priority commitment, a complaint about an agent. Anything on the list moves up; anything not on the list stays at first line until the list is updated. Review the list monthly against what was actually escalated.
Lever three: choose location by contact type, not by preference
Where the work is done is the lever most buyers think of as the whole decision. It is one lever among several, and it works best when it is applied per contact type rather than to the whole operation.
Fully documented, rarely ambiguous work runs well offshore at the lowest cost per seat, because it does not depend on a same-day answer from your team. Work that needs a decision from your side, or that touches a customer who expects a shared time zone and shared context, degrades across a large time gap and belongs nearshore or onshore. Splitting by contact type usually beats choosing one location for everything, and it lets you keep the sensitive or judgement-heavy work close while moving the volume.
The mistake is choosing the location first and forcing every contact type into it. Do the categorisation first, then place each category where it will be handled well. The same categorisation you did for lever one gives you the answer here.
Lever four: shift channel where customers accept it
Voice is the most expensive channel because one agent handles one conversation at a time. Chat and email allow concurrency, so the same agent hours cover more contacts. The saving is real, but it has a condition: customers have to be willing to use the channel for that contact type.
Order status, address changes, simple account updates and receipt requests move to chat or email without complaint. A customer whose payment has failed twice, or whose service is down, wants a voice. Move the contact types customers are happy to handle in writing, and leave voice for what genuinely needs it. The channel decision is made per contact type, like the location decision, and for the same reason.
Self-service belongs in the same lever. A clear order tracking page, a returns portal that works, and a help centre article that answers the top questions absorb the contacts customers would rather not make anyway. Treat self-service as a channel with its own quality standard, and check that it actually closes the contact rather than sending the customer to the phone with a worse mood.
Do not remove the phone number. Hiding it reduces contacts and increases churn. The customers who most need to call are the ones you can least afford to lose, and a customer who cannot reach you does not become cheaper. They become someone else's customer.
Lever five: fix repeat contacts
A repeat contact is the same customer, the same issue, a second time. Every one of them is a cost you have already paid once. Repeat rate is worth tracking on its own because it exposes problems that average handle time hides: a first line that closes tickets without resolving them, a knowledge base with a wrong answer in it, a handoff between tiers where context is lost.
The fix is usually procedural. Give agents the authority to resolve the common exceptions without an escalation, write the resolution steps down where they can be found, and review a sample of repeat contacts every week to find the pattern. This lever is smaller than the first two, but it compounds, because every repeat you remove also removes the frustration that drives the next one.
The false economies
Some cuts look like savings on a spreadsheet and cost more within a quarter.
Cutting quality assurance saves a small amount and removes the only mechanism that tells you when the service is drifting. By the time the drift shows up in complaints or churn, the saving is gone several times over.
Understaffing to hit a cost target produces queues, abandonment and repeat contacts. The customers who abandoned call back, so the volume did not go away. It moved to a worse moment and arrived angrier.
Targeting handle time directly teaches agents to end calls rather than resolve issues. Handle time is a useful diagnostic and a terrible target. Resolution on first contact is the measure that lowers cost; handle time follows it when the process is right.
Cutting training to get agents on the phones faster produces agents who escalate everything, and the saving reappears as second-line cost and customer effort.
A sequence that works
If you are starting a cost programme, the order matters. Get the contact driver data. Remove the top avoidable drivers with the teams that own them. Tier what remains and write the escalation rule. Place each tier where it will be handled well. Move the written-friendly contact types to chat and email. Track repeat rate and first-contact resolution as the primary measures, with handle time and rate per hour as secondary. Then, and only then, negotiate the rate, because by that point you know exactly what you are buying and how much of it you need.
If you want a diagnosis before you commit to a staffing change, our customer experience consulting service looks at where the cost is actually coming from. And if you are comparing quotes, what outsourced customer service costs explains the pricing models you will be shown and what each one rewards.