Ask a company that outsourced support what actually went wrong and you rarely hear "the agents were bad." You hear that nobody could see anything. Deloitte's global outsourcing survey puts a number on it: the most-cited challenge is data visibility, inadequate tracking and reporting, named by 55% of respondents. Poor vendor performance during service transitions sits at 46%.
Both of those are downstream of one question that gets answered in the first week of a contract and then never revisited. Which system does the outsourced team actually work in - yours, or theirs?
Most buyers treat that as an implementation detail and let the provider decide. It is a commercial decision. It sets who holds your ticket history, how quickly you can see a problem forming, what your reporting is made of, and what it costs you to change your mind in two years.
The tooling question is an ownership question
Vendor comparison pages will tell you which help desk is best. That is the wrong axis. Zendesk, Freshdesk, Intercom, Jira Service Management and HubSpot all resolve tickets competently. None of them is going to lose you a customer because of a missing feature.
What changes the outcome is whose account the work happens in. Every ticket your outsourced team touches produces four things you will want later: the conversation history, the saved replies the team wrote to handle repeat issues, the tags and categories that make any reporting possible, and the internal notes explaining why an unusual case was handled the way it was. That is the raw material of your support operation. If it accumulates inside a platform your provider owns, you are renting your own operating history.
This is not a bad-faith scenario. Very few providers are planning to hold anything hostage. It is simply that the default arrangement - the vendor working in the vendor's system, because that is how the vendor is already set up - quietly makes your support function non-portable. You find out at exactly the wrong moment: when you want to renegotiate, add a second provider for a new region, or bring part of the work back in-house.
Three help desk models, and what each one really costs
Whatever the proposal calls it, there are only three arrangements.
| Model | How it works | What you gain | What it costs you |
|---|---|---|---|
| Your stack | You own the help desk instance. The provider's agents are named users inside it with a scoped role. | Full ticket history, native reporting, same-day visibility, nothing to migrate if you switch. | You pay per agent seat and you administer roles, permissions and offboarding. |
| Vendor stack | The provider works in its own help desk and sends you reports, sometimes a read-only portal. | No license line on your side, faster to start, the provider brings mature workflows. | History and saved replies live in their account. Reporting is whatever they choose to send. Leaving means rebuilding from an export you have to negotiate for. |
| Hybrid | The provider works in its own system, synced into yours through an integration or an interface. | Workable when a provider serves many clients from one platform and cannot leave it. | Two systems to reconcile. Sync gaps become invisible failures. You usually end up paying for the licenses anyway. |
The seat math nobody puts in the proposal
The usual objection to the first model is cost, so here are the real 2026 numbers. Zendesk lists Support Team at $19 per agent per month, Suite Team at $55, Suite Professional at $115 and Suite Enterprise at $169, on annual billing, with bundles such as the AI copilot adding roughly $50 per agent per month on top. Freshdesk lists Growth at $19, Pro at $55 and Enterprise at $89, with its AI copilot at $29 per agent per month.
Run it for a six-agent outsourced pod on a mid tier: six seats at $55 is $330 a month, or $3,960 a year. Push it to a professional tier with one add-on and you are nearer $12,000 a year.
Now put that next to the service. An outsourced support retainer for a team that size typically runs $6,000 to $12,000 a month. The license line is roughly 3% to 5% of what you are already paying for the people. Teams argue about it because it arrives as a new charge on their own card while the retainer is one they have already accepted. Measured against what it protects, it is the cheapest term in the deal.
One thing worth checking before you budget the sticker price: most help desks price light, contributor or collaborator seats well below a full agent seat for people who only add internal notes, and several cap administrators separately from agents. If your provider's team lead is the only one who needs configuration rights, you are not buying ten full seats.
When the vendor's own platform is genuinely the right call
There are real cases for it. Be honest about which one you are in.
You have no help desk at all and no plan to build one. If support is a shared inbox today and the provider brings a functioning ticketing setup, taking theirs is a genuine upgrade rather than a concession. You have a volume that is small and temporary - a seasonal spike, a launch window, a single campaign - and the work will not outlive the contract. Or the work is genuinely commoditized, such as after-hours triage that only routes and escalates, where the history has little value to you because nothing is resolved in it.
What makes those cases safe is not the model. It is writing the exit terms at signature rather than discovering them at the end: the export format, who pays for it, how long it takes, and whether it includes internal notes and saved replies or only the customer-facing messages. If you already suspect you will want to move later, read our guide on how to switch customer support providers safely before you sign, not after.
What access to grant on day one
Deciding to run the work in your instance raises the obvious next question: what do you actually hand over? Six rules cover it.
1. Named accounts, never a shared login. One account per agent. A shared login destroys your audit trail and makes offboarding impossible to verify.
2. The lowest role that does the job. Agent, not administrator. Most providers never need to change your automations, triggers or business rules.
3. Scoped visibility. Restrict the team to the queues, groups or brands they are contracted to cover. If they handle first line for one product, they do not need the enterprise sales queue.
4. Keep the administrator on your side. Grant configuration rights to one named lead, ideally yours. If the provider's team lead needs them, that is a specific decision you make, not a default.
5. Offboarding in the contract. Agree that the provider notifies you within one business day when anyone leaves the pod, and that you deactivate the seat the same day. Otherwise a former agent keeps read access to your customer conversations for months.
6. Nothing that touches money or production. Billing consoles, payment data and production systems stay out of scope unless the work genuinely requires them, and then only with a separate, logged permission.
Treat the request itself as information. A provider that asks for owner-level access to the whole instance before the first ticket, or that resists named accounts because one shared login is simpler to manage, has just told you how it runs every other client account too.
You cannot measure what lives in someone else's account
Go back to that 55%. The most common complaint about outsourcing is not that the work is bad, it is that the client cannot see it. That is not a reporting-template problem and no dashboard fixes it. It is an account-ownership problem.
When the tickets sit in the provider's system, your monthly review is a document the provider wrote about itself. You can only ask the questions the report anticipated. Response time looks fine, so you accept it, without being able to check whether the queue was reordered to make it look fine. Resolution rate climbs, and you cannot see that repeat contacts climbed with it.
When the work runs in your instance, the review changes shape. You open the ticket, read the actual thread, and check the claim yourself. You can pull any two metrics together instead of the pair the provider chose to print. That is the precondition for everything in measuring outsourced support quality - the scorecard is only as honest as your access to the data underneath it.
There is a second effect that is easy to miss. When a client can read raw tickets, the provider knows it. Reporting stops being a performance and becomes a summary of something you can both see. That changes the monthly conversation more than any contract clause.
How IMMIDO works inside the client's own tools
We run first-line support as a business-to-business operation. The cases come from other companies and their staff rather than from consumers, and the work is checking cases and transactions and handling incidents as they come in. We do that inside Jira and the client's own help desk. The client's stack is the working surface, and we adapt to it.
That choice costs us something. Every new client means learning another configuration, another set of automations, another way of naming things, which is part of why bringing a new client to full coverage takes two to four weeks rather than two days. A provider that moved everyone into one system of its own would onboard faster and carry less overhead.
What it buys the client is the part that matters after year one. The ticket history stays in their account. Their reporting runs on their data and their tickets, not on a parallel system we control. If they ever want to move the work, to another provider or back in-house, there is nothing to extract and nothing to negotiate for. Our commercial model is a fixed monthly retainer, typically $6,000 to $12,000 a month for a team of six to ten agents, and because the work happens in the client's own tools, the help desk licenses sit on the client's account rather than ours.
It is worth saying plainly: this is not generosity. A client who can leave easily and does not is a better reference than a client who stays because leaving is expensive. You can see the rest of what we run on our services page.
Decide by where the data lands
You are not choosing a help desk. You are choosing where your support history accumulates for the next few years, and every other question follows from that. Three questions settle it before you sign.
If we ended this contract in eighteen months, what would we have to ask the provider for? If the honest answer includes ticket history, saved replies or your knowledge base, you are in the vendor-stack model whatever the proposal says.
Can I read a raw ticket thread today, on my own, without asking anyone? If not, your visibility depends on someone else's willingness, which is precisely the thing 55% of outsourcing buyers complain about.
Who administers the account the work happens in? If it is not someone at your company, you have outsourced more than the support.
None of this makes the vendor-stack model wrong. It makes it a decision instead of a default. Choose it deliberately, with the exit written down, or choose your own stack and pay the seat cost, which for most teams is a few percent of the retainer and the cheapest control you will buy all year.
If you are scoping an outsourced support team and want to know what it looks like when the work runs entirely inside your own tools, that is worth a conversation. Book a call and get a quote → We will map the access scope for your help desk, tell you which seats you actually need, and show you what your reporting looks like when the data never leaves your account.