First Contact Resolution for an IT Service Desk: What 90% Really Measures

TL;DR: First contact resolution on an IT service desk, the "90% at first contact" in a tender, is not one number but at least three, depending on what counts as contact, resolution and an eligible ticket. Before you tender, fix the definition, the denominator and the audit right, and pair the figure with reopen rate and answer time. A desk can only resolve at first contact what it is allowed and equipped to touch.

Picture a Hong Kong-headquartered group of a few hundred staff, with offices in Singapore and mainland China, putting its whole IT service desk out to tender. The RFP is thorough, but one line carries most of the weight: "first-line resolution of 90% or above, 24×7, bilingual". Three bidders reply, and all three say yes. The head of IT now holds three signed confirmations and no idea whether they mean the same thing.

That group is an illustrative composite, not a named client, and this guide is deliberately not a story about it. It is a working guide to the number itself: what first contact resolution measures, how it gets gamed, what actually moves it, which metrics must sit beside it, how to write it into a contract, and what a monthly report should show. If you are comparing outsourced service desks, the goal is that by the end you can tell an honest 90% from a convenient one.

Why does an RFP ask for 90% first-line resolution, and why does every bidder say yes?

Buyers write a first contact resolution target into a tender for good reasons. It is easy to say in one line. It sounds like a measure of competence: the person who picks up the phone fixes the problem, and nobody has to wait. It also feels like a proxy for cost, because every ticket that bounces between tiers consumes more minutes of expensive engineers and more minutes of your own staff's patience.

Bidders say yes for a different reason. The phrase has no single owner. ITIL does not prescribe one definition, tools report it in different ways, and a sales team answering a tender will naturally pick the interpretation under which they can comply. A bidder who answers "yes" is usually not lying. They are answering a question that is a little different from the one you thought you asked.

So the three yeses can differ in at least four places: what counts as a "contact", what counts as "resolved", which tickets are in the measurement at all, and who decides all of that. The rest of this guide takes those four in turn. In this illustrative group, the buyer's first useful move is not to pick the bidder with the biggest number. It is to ask each bidder to write down the sentence that defines the number.

What does first contact resolution actually measure?

At the simplest level it is a fraction: tickets resolved at first contact, divided by tickets that could have been. Every word in that sentence is a decision. There are three common definitions in circulation, and they are different enough that the same desk, handling the same month of tickets, can produce noticeably different results under each.

Definition A: resolved during the first interaction

The strictest reading. The user contacts the desk, the agent resolves the issue in that conversation, with no transfer, no callback and no "we will get back to you". A password reset done on the call counts. A problem that needs the agent to go away and research does not, even if the agent solves it twenty minutes later.

This is the definition closest to what end users imagine when they hear the phrase, and it is the hardest to meet. It is also the one that best reflects the user's experience, because the user's clock stops during the conversation, not at some later point.

Definition B: resolved by the first line, at any time

This is better described as first-line or first-level resolution. The ticket is resolved by the first line without being passed to a second or third tier, but there may be several touches: a call, a follow-up message, a remote session the next morning. It measures how much of the workload the first line can carry, not how quickly the user is freed.

This is a legitimate measure and often the right one for cost, but it is not the same as "at first contact". A bidder who reports it under the name of the stricter metric will look better than they are.

Definition C: resolved without escalation to a higher tier or third party

The broadest reading. Anything that never left the first-line team counts, including tickets where the agent called a senior colleague on the same floor for help, or where the first line waited three days for a user to respond. It tells you very little about speed and a fair amount about team structure: it rewards a first line that is staffed with senior people and labelled as first line.

Which tickets count?

Even with one definition, the population matters. Incidents only, or service requests as well? Are password resets and account unlocks in the count, which are quick and numerous, or are they excluded as "not real incidents"? Are how-to questions included? Are tickets raised by the monitoring system, not by a person, included? Each choice moves the number, and none of them is wrong in principle. They are wrong only when they are unstated.

What is the denominator?

This is where the largest differences tend to hide. One option is every contact the desk handled. Another is "FCR-eligible" contacts, meaning those the provider has judged capable of being resolved at first contact. The second option shrinks the denominator by removing the hard tickets, and a smaller denominator pushes the percentage up without a single extra ticket being solved. The buyer should treat any denominator other than "all contacts in scope" as something that needs an explicit, written eligibility list.

Who classifies?

The last question is who decides what was resolved, what was eligible and what was a contact. If the provider's agents classify their own tickets, in the provider's system, with no sample audited by anyone else, then the number reflects the provider's discretion as much as the provider's performance. That is not a criticism of any particular provider. It is a structural fact, and the contract needs to deal with it.

As an illustration, imagine the same month in the composite group's desk. Under definition A, one figure comes out. Under definition B, a higher one. Under definition C, higher still. Add an eligibility filter on the denominator, and the number rises again. No figures are given here on purpose: any number would be invented for the composite and carry no market meaning. The point is only that none of the three yeses in the tender can be compared until all four choices are written down.

How does first contact resolution get gamed?

Most gaming is not fraud. It is a measurement system doing what it was told, in the way a busy team under a target will naturally drift toward. A buyer who knows the patterns can ask the right questions during evaluation.

• Reclassification into "not eligible" — A hard ticket is tagged as a project, a change or a "complex incident" at the moment of logging, so it never enters the denominator. The percentage improves; nothing else does.
• Premature closure and reopen — The ticket is closed after the first reply, on the theory that the user will reopen it if the problem persists. If the reopen is logged as a new ticket, the first one counts as a success and the second one is a fresh attempt at a fresh success.
• Splitting one problem into several tickets — A single failure, such as a laptop that cannot join the network and cannot reach email, becomes three tickets, two of which are quick. Volume goes up and so does the share that resolves quickly.
• "Resolved" meaning "workaround given" — The agent supplies a temporary fix, the ticket is closed as resolved, and the underlying cause is left for no one. The user is satisfied for a day.
• Excluding channels — Phone calls are measured; chat, email and walk-up requests are not, or the reverse. Whichever channel is easiest to resolve on is included.
• Counting the first line's own escalations as first-line — The first-line agent asks a senior colleague in the same team for help. The ticket never changes owner in the system, so it stays "first line", even though the answer came from someone more senior.

None of these needs bad faith. All of them are visible if you ask for the right evidence, which is why the later sections of this guide spend more time on audit and reporting than on the target itself.

What actually raises first contact resolution?

A target does not resolve tickets; people with tools and permissions do. The factors that move first contact resolution are mostly structural, and a buyer can ask about each of them before signing.

A knowledge base and scripts the first line can execute

A first-line agent can only resolve what they know how to resolve. That means a knowledge base that is maintained, searchable and written as steps, not as essays, and scripts or runbooks the agent can actually execute with the access they have. The measure to ask for is not the size of the knowledge base but how many articles were added or corrected last month and which ticket categories prompted them.

First-line administrative rights

This is the single most common reason a first line cannot close a ticket. If the agent cannot reset a password, re-enrol multi-factor authentication, assign a licence, change a mailbox permission or take remote control of a device with the user's consent, then those tickets must go to another tier, and no amount of agent skill will change that. The question for the bidder is not "how many agents do you have" but "which of these actions can your first line perform in my tenant, on day one, with what approval and what audit trail?"

Language match

A desk that handles Cantonese, Mandarin and English on the same shift will resolve more at first contact than a desk that passes a Cantonese caller to a call-back queue. For a Hong Kong group with offices in Singapore and mainland China, language is not a courtesy. It decides whether the agent understands the symptom the first time. Ask which languages are staffed on which shifts, not which languages the company "supports".

Endpoint visibility and remote-support tooling

An agent who can see the device's state, disk, patch level, security status and recent changes, will diagnose in minutes what would otherwise take a series of questions to the user. An agent who can take a consent-based remote session, with the user watching, will fix in one conversation what would otherwise need a visit or a long phone walkthrough. This is the area where the tooling matters most, and it is covered in more detail in the bridge section below.

ITSM integration, so the agent sees history

If the agent can see that the same user raised the same fault twice last month, the conversation changes: they skip the basics and go to the cause. That requires the desk to work in, or integrate with, a proper ITSM platform. When a provider works in its own system and your internal team works in yours, history is invisible to one side of the conversation.

Which metrics should sit beside first contact resolution?

First contact resolution on its own is dangerous, because it can be improved by making everything else worse. It needs companions that push in the opposite direction.

• Reopen rate — This is the honesty check. A desk that closes quickly and has a high reopen rate has not resolved anything; it has moved the work. Ask for reopen rate in the same report, over the same period, under a stated reopen window.
• Answer time and abandon rate — A desk can post a good resolution figure by letting hard calls abandon. Answer time tells you how long people waited, and abandon rate tells you how many gave up. One data point from Brocent, to be clear about what it is: its 24×7 Multi-Lingual Help Desk answers 90% of calls within 40 seconds. That is an answer-time figure, not a resolution figure, and the two should never be confused.
• Time to resolve by priority — Averages hide the long tail. A report that shows resolution time by priority level shows whether urgent faults are being treated differently. The priority ladder itself is covered in our guide to P1 to P4 priority levels.
• Customer satisfaction per ticket — A short question after closure, linked to the ticket, so that a poor score can be traced to a specific resolution. Satisfaction at ticket level is much harder to flatter than a monthly average.
• Escalation rate — The complement of first-line resolution, but worth reporting separately by category. It shows which kinds of issue the first line cannot carry, which is exactly where the next knowledge article or permission change belongs.
• Backlog age — How old is the oldest open ticket, and how many are older than the agreed threshold? A resolution percentage computed only on closed tickets says nothing about the ones that are not closed.

Together these metrics make a number like 90% harder to manipulate, because improving one at the expense of another becomes visible in the same table.

How do you write first contact resolution into the contract?

A target in a tender is a hope. A clause in a contract is an obligation, and the clause is only as good as its definitions. In evaluating bidders, ask for draft wording on each of the following, because a provider's willingness to commit to the wording tells you more than the number does.

• Definition clause — State which of the three definitions applies, in a sentence. If the provider proposes a different one, the proposal is information.
• Eligibility list agreed up front — A list of ticket categories in scope and, separately, any categories excluded, with the reason. Changes to the list require agreement, not notice.
• Denominator — All contacts in scope on all channels, unless the eligibility list says otherwise.
• Measurement source — Is the number calculated from the provider's ITSM or from yours? If the provider's, does the client have read access to the raw ticket data? Calculation from data you can inspect is worth far more than a figure in a PDF.
• Audit right over a ticket sample — The client may pull a random sample of closed tickets each month and test them against the definition. Without the right to audit, the definition is decoration.
• Reopen window — A ticket counts as resolved at first contact only if it is not reopened within a stated period, for example five business days. This single clause removes most of the incentive for premature closure.
• Ramp-up period — In the first 90 days, the desk is learning your environment, your applications and your people. A reasonable contract allows for a ramp-up, with the target reported but not enforced, then enforced after.
• What a service credit should and should not be attached to — It is tempting to attach a credit to the headline percentage. That is also where the incentive to game is strongest. Credits tend to work better on measures that are hard to bend, such as answer time, and to be considered carefully for a number that depends on classification. We have written separately about how SLA service credits and performance bonds are enforced, so this guide does not repeat that.

A useful test during the tender: send the same draft clause to all three bidders and compare not only the answers but how long each takes to reply and how much they want to change.

What should the monthly report show?

If the contract defines the number, the report proves it. A report that shows only the headline percentage should be treated as a warning sign. A report that lets you recompute the percentage yourself is a sign of confidence. At minimum, ask for:

• Volume by channel and by language — Phone, chat, email, portal and any walk-up or onsite request, split by language. This tells you whether the measured population is the whole population.
• First contact resolution under the agreed definition — With the excluded categories listed beside it and the number of tickets in each, so that exclusions are visible, not buried.
• Reopen rate — For the same period and the same window as in the contract.
• Answer time — Including abandon rate. If a call-answer figure is quoted, it is labelled as an answer-time figure.
• Top ten ticket categories — This is the most practical page in the report, because it shows where the next improvement should come from: a knowledge article, a permission, a self-service flow, or a permanent fix.
• Knowledge-base articles added or revised — Evidence that the desk is learning your environment, not just answering questions in it.
• A sample of audited tickets — A short list of tickets, chosen by the client or by a stated random method, with their history, so the numbers can be tied to real conversations.

A report like this takes work to produce. That is the point: a provider who produces it routinely has built the habit of measuring honestly.

Which definition should you buy? A side-by-side comparison

What each definition counts, what it hides, and when it is the right one

• A. Resolved in the first interaction — Counts tickets closed in one conversation with no transfer or callback. It says little about how much of the workload the first line can carry across several touches. Buy it when user experience and minutes lost are your main concern, and when a good share of your tickets are routine and can be fixed on the spot.
• B. Resolved by the first line, at any time — Counts tickets the first line closes however many touches it takes. It hides slow resolution, because a ticket that takes a week still counts, and it hides rework between contacts. Buy it when your main concern is cost and the load placed on engineers in higher tiers, and pair it with time to resolve and reopen rate.
• C. Resolved without escalation to a higher tier or third party — Counts anything that never left the first-line team. It hides senior help inside the first line, long waits for the user, and the difference between a team with real expertise and a team that is merely labelled first line. Buy it only if you also get a clear description of how the first line is staffed, and treat it as the weakest of the three on its own.

If the three bidders in the composite tender report under three different definitions, the honest comparison is to ask each to restate its result under A and B on a common set of ticket categories, with the denominator stated.

Can a service desk resolve at first contact only what it is allowed to touch?

Yes, and this is the bridge between the metric and the way the service is built. A desk can only resolve at first contact what it is allowed and equipped to touch. If administrative rights, device visibility and ticket history live in three different places with three different owners, the first line will be competent and still unable to close.

Inside a per-user managed IT plan, those pieces are designed together. Brocent's 24×7 Multi-Lingual Help Desk is an ITIL-based Global Service Desk operating from centres in China, Hong Kong and Malaysia, handling approximately 15,000 IT incidents and service requests a year. It works in Mandarin, Cantonese and English, with Tier 1 to 4 engineers, and 150+ help-desk staff with certifications across 70+ disciplines. It covers incident and problem management, change management, onboarding and offboarding, remote workstation management for Mac and Windows, vendor liaison and monthly SLA metric reporting, and it integrates with ServiceNow, ManageEngine ServiceDesk Plus and Jira. If a fault cannot be resolved remotely, it escalates to on-site dispatch. Outsourced and white-label desks are also available.

Brocent does not publish a first contact resolution percentage, and this guide will not invent one. Its published service-level figure is the answer-time one: 90% of calls answered within 40 seconds. If you ask what a realistic first contact resolution rate looks like, our own professional view is that it depends entirely on the definition and the ticket mix, which is why the clause matters more than the number.

On the endpoint side, BCS Beam is one signed agent that gives consent-first, fully audited remote support: screen assistance, background maintenance, in-desktop chat and one-click time-limited invitations. Every session is visible in the system tray and tied to its ticket. The same agent runs a continuous read-only security and health audit covering disk encryption, antivirus, firewall and patch status, so the agent on the phone can see the state of the device before asking a single question. A connection audit ledger records who connected to which device, when, in what mode and for which ticket, and the service is hosted on Brocent's own infrastructure in Hong Kong. That ledger is also what makes the audit right in your contract practical.

Put together, the desk, the endpoint agent, the administrative rights and the monthly report are one engine rather than four purchases. That is the structural reason a managed arrangement can commit to measurable resolution at the first line, and why an outsourced desk bolted onto an environment it cannot touch cannot.

Frequently asked questions

What is a realistic first contact resolution rate?

There is no honest answer without the definition. The same desk can report quite different figures for the same month depending on whether it counts only the first interaction, anything the first line closes, or anything that was not escalated, and whether password resets and how-to questions are in the denominator. Brocent publishes no first contact resolution figure. Our own professional view is that you should ask each bidder to state the number under a named definition and an agreed ticket list, then judge it alongside reopen rate.

Does a password reset count?

It should count if it is within the agreed ticket scope, and the contract should say so. Password resets and account unlocks are quick, frequent and almost always resolved on the first contact, so they pull the percentage upward. That is not wrong, but it means a figure that includes them is not comparable with one that excludes them. Ask for the result with and without them, and agree which one the contract uses.

How does language affect first contact resolution?

A great deal, particularly for a group operating across Hong Kong, Singapore and mainland China. If the agent cannot understand the symptom as the user describes it, the ticket goes to a callback queue or to another agent, and the contact is no longer a first contact. Ask which languages are staffed on which shifts. Brocent's help desk works in Mandarin, Cantonese and English, but the question to put to any bidder is who is on shift at the hour your users actually call.

Should first contact resolution carry a service credit?

Cautiously. A credit attached to a percentage that depends on classification gives the provider the strongest incentive to classify generously. If you do attach one, tie it to the agreed definition, a fixed eligibility list, a reopen window and your right to audit a sample. Measures that are harder to bend, such as answer time, often make safer credit triggers. The mechanics of credits and performance bonds are covered in the related article on SLA enforcement.

How is first contact resolution audited?

By sampling. Each month, the client draws a random set of closed tickets, or a set chosen by an agreed method, and checks each against the contract definition: was it resolved in the first interaction, was it reopened inside the window, was it correctly classified. This needs read access to ticket data and, for remote sessions, a record of who connected and why. A connection audit ledger tied to tickets helps, because it shows what actually happened on the device.

What is the difference between first contact resolution and first-level resolution?

First contact resolution is about the moment: the issue is solved in the first interaction. First-level resolution is about the tier: the issue is solved by the first line, however many interactions that takes. They answer different questions, one about user experience and one about how much work the first line can carry, and a tender that uses the terms interchangeably is leaving room for the numbers to differ. Name which one you mean.

Does 24/7 coverage change first contact resolution?

It changes what you can reasonably compare. A desk that is staffed around the clock but only has senior engineers during office hours will resolve differently at 3 a.m. than at 11 a.m. Ask for the figure by shift or time band, and for the same language and administrative-rights arrangements out of hours. Round-the-clock coverage widens the population of contacts; it does not by itself improve the share resolved.

Can first contact resolution be measured on chat tickets?

Yes, and it should be, if chat is a channel your users will use. Chat makes the definition harder, because a conversation can pause for hours and still be one ticket. Decide in the contract whether a paused chat that resumes counts as one contact or two, and make sure chat is in the denominator. Excluding the channel that is easiest to resolve on, or the hardest, is one of the quietest ways a figure drifts.

What to do next

If you are drafting a tender, the practical sequence is short. Write the definition into the RFP, list the eligible ticket categories, name the denominator, require the report layout above, and reserve the right to audit a sample. Then compare bidders on how they respond to the wording, not on the number they quote.

If you would rather have the desk, the endpoint agent, the administrative rights and the monthly report arrive as one service, that is what the per-user managed IT plan is for, and the 24×7 desk and BCS Beam are included in it or available as add-ons to it. You can see how it is priced on the pricing page, or talk to us with your current ticket categories and we will go through the definition question with you before anyone mentions a percentage.