What is call center? Inbound vs outbound, structures, KPIs, and staffing

What is a call center?

It is the organized handling of telephone conversations for a business, including the people, routing rules, records, and controls that support each call. A call center may receive customer requests, place planned calls, or do both. The direction of the call changes the operating problem.

Inbound and outbound teams are sometimes given the same agent targets, dashboards, and staffing ratios. That obscures an important difference. An inbound queue must respond when customers decide to call. An outbound team can usually release work from a planned list, subject to customer contact rules and campaign priorities. Plan each model around that difference.

Start with how demand reaches the team

Inbound demand is customer initiated. It can rise when an order fails, a service is unavailable, or customers need urgent help. The operation needs enough available agents, clear routing, and an escalation path at the moment the call arrives. Waiting time, abandonment, and resolution quality all matter because the customer has already chosen to seek help.

Outbound demand is planned work. A team may call to follow up on a service request, confirm an appointment, collect survey responses, remind a customer about a payment, or qualify a prospect. Managers decide which records to release, when to call, and what outcome the agent must record. The work still needs capacity, but it does not create the same always-open queue as inbound support.

A blended team handles both directions. Blending can use otherwise idle time, but it can also damage inbound access if outbound work is released without a queue-protection rule. Reserve the required inbound coverage first. Release outbound work only when the team has a defined trigger for doing so.

Compare the operating models before setting targets

An inbound vs outbound call center comparison should cover the flow of work, not just whether an agent answers or dials a phone number. Each model needs a different definition of a completed interaction.

Inbound work depends on access and ownership

An inbound call begins with the customer's reason for contact. A short menu, a routing rule, or an available agent sends the call toward an owner. The agent then needs enough context to diagnose the issue, record the result, and create follow-up work when the issue cannot be closed on the call.

The question is whether the caller can reach an appropriate owner in time. Managers should watch interval demand, queue time, agent availability, transfers, escalations, and repeat contact. A short handle time does not help if it creates unnecessary transfers or callbacks.

Outbound work depends on a controlled work release

An outbound call starts with a business purpose and a record that is eligible to contact. The team needs an approved list, contact-window rules, a call purpose, outcome codes, and a next-step owner. A call may result in a completed task, an appointment, a follow-up date, an opt-out, or no connection at all.

The question is whether planned effort produces a valid next step. Dials show activity, but they do not prove that the list was suitable, the customer was reached, or the promised follow-up happened. Track outcomes by record as well as by agent.

Blended teams need a priority rule

When the same people handle both models, define what interrupts outbound work. For example, a queue threshold, service-level alert, or supervisor instruction can return agents to inbound coverage. The rule should state who makes the decision and how the team records the interruption. This is a recommendation, not a universal configuration: the right trigger depends on customer urgency and the business's service promise.

Match the metric to the operating decision

Call center metrics should support a decision, rather than reward the lowest handle time or highest dial count. Inbound measures explain access and resolution pressure. Outbound measures explain contactability, progress, and completion. Quality and schedule controls apply to both.

For inbound work, service level and abandonment show whether demand and coverage are in balance. First-contact resolution and transfer patterns show whether the caller reached the right person and whether that person had the information to act.

For outbound work, contact rate, completed work, outcome quality, and follow-up completion are usually more meaningful than a raw activity count. A campaign can produce many dials and still fail because the list is weak, contact timing is wrong, or agents cannot complete the required next step.

Plan inbound coverage with Erlang C

This is a simplified illustrative approximation, not a production workforce-management forecast. Erlang C is commonly used for inbound queues because it models random arrivals and the probability that a caller waits. It is not a staffing promise.

Use one planning interval at a time, such as 30 or 60 minutes. Gather these five inputs:

1. Expected inbound calls during the interval
2. Average handle time per call in seconds, including wrap-up work
3. Interval length in seconds. A 30-minute interval is 1,800 seconds.
4. Answer-time target in seconds. For example, 20 seconds.
5. Service-level target as a decimal. For example, 80% becomes 0.80.

Start with the workload calculation:

Workload in Erlangs = expected calls × average handle time in seconds ÷ seconds in the interval

For example, 100 expected calls with a 300-second average handle time in a 30-minute interval creates 16.67 Erlangs of workload:

100 × 300 ÷ 1,800 = 16.67 Erlangs

Next, enter the expected calls, handle time, answer-time target, and service-level target into an Erlang C calculator. Test whole agent counts starting above the workload. In this example, start at 17 agents, then test 18, 19, and so on. Select the first agent count that meets the answer-time and service-level target.

The calculator uses the standard Erlang C waiting calculation in the background. It estimates the chance that a caller waits and the share of calls answered within the chosen target. You do not need to calculate factorials or waiting probabilities by hand to use the staffing method.

The selected agent count is the productive-seat estimate for that interval. It does not yet include breaks, training, meetings, absence, or other scheduled unavailability.

Erlang C assumes a stable queue, random arrivals, identical agent capability, and callers who wait. It does not fully model abandonment, sudden spikes, callbacks, mixed skills, or forecast error. A production workforce-management process should test its forecasts against actual interval results and use more suitable models where those limits matter.

Plan outbound capacity with a workload equation

Outbound scheduling should start with a controllable work unit. That may be completed conversations, completed reminders, qualified contacts, or closed follow-up tasks. This equation is a capacity approximation, not Erlang C:

Productive outbound agents needed =
round up((planned completed work units × seconds per completed work unit)
÷ (seconds available per agent in the interval × productive-time factor))

• Planned completed work units: the conversations, reminders, or follow-up tasks the team expects to complete in the interval.
• Seconds per completed work unit: talk time plus required after-call work.
• Productive-time factor: the share of the interval assigned to this work, written as a decimal. For example, 75% becomes 0.75.

If planning begins with dials, estimate completed work first:

Planned completed work units = planned dials × expected connect or completion rate

This is an illustrative formula. The connect or completion rate must come from the team's own recent, compliant campaign data rather than a borrowed industry ratio. Contact restrictions, list quality, customer preferences, and campaign purpose can change the result substantially.

For either model, convert productive seats into scheduled seats after estimating shrinkage:

Scheduled agents needed = round up(productive agents needed ÷ (1 - shrinkage rate))

Here, shrinkage is the planned share of paid time that is unavailable for live work. It may include breaks, training, meetings, coaching, or approved leave. Do not treat the output as a fixed staffing ratio. Use it as a starting point to test against actual queue or campaign performance.

Put Udesk AI Call Center capabilities at the right control points

At Udesk, we bring call-center, ticketing, live-chat, and AI-agent components into one customer-service environment. Our AI Call Center includes AI VoiceBot support, conversation summaries, unified monitoring, and unified reporting. These capabilities work best when they are connected to clear operating controls rather than treated as general automation claims.

For inbound calls, our AI voice capabilities can handle approved routine questions before handing a conversation to a human agent. Conversation summaries can give the next owner a record of the request, while ticketing keeps follow-up visible when work continues after the call. Your team still needs to define the questions that automation may handle, the handoff trigger, and the person accountable for exceptions.

For outbound work, Udesk supports structured follow-up through call records, outcome fields, and reporting views that show what was attempted and what next action remains. Our platform also supports voice and text bots, customer tags, interaction history, and reporting components. These capabilities can give teams a more complete view of campaign work, but they do not replace approved contact policies or supervisor review.

Udesk integrates with more than 30 mainstream channels and provides a unified service view. This can help when a customer calls after a prior message or when an outbound call needs a documented follow-up. It does not change the staffing rule. Measure inbound queue demand and outbound workload separately before allocating people.

Use two planning views in the weekly review

An inbound review asks whether customers had access to help. Compare interval forecast and actual call volume, answer time, abandonment, handle time, transfers, and repeat contacts. When results change, inspect the call reason and routing path before adding a universal agent target.

An outbound review asks whether planned work reached a valid outcome. Compare planned and completed records, connection results, disposition quality, after-call work, follow-up completion, and any permitted business outcome. If productivity falls, check list quality, contact timing, system access, and task complexity before demanding more dials.

The two views can sit in one management report, but they need separate definitions. This lets a blended operation protect customers who need immediate help while using planned outbound time responsibly.

Set the operating rules before adding more seats

The best staffing decision begins with a clear work-entry rule. For inbound calls, define how customers reach the right queue, what wait time the business accepts, and who owns escalations. For outbound calls, define the eligible work list, permitted contact process, required outcome fields, and follow-up owner.

Then use the correct planning method. Erlang C can provide an inbound queue approximation when its assumptions are understood. A workload equation can estimate outbound productive capacity when the work unit and completion assumptions are visible. Neither approach replaces measured local data, quality review, or accountable process ownership.

FAQ

Q: Is a call center always inbound or outbound?

A: No. A call center may be inbound, outbound, or blended. A blended team should still plan and report the two directions separately because customer-initiated queues and planned outreach create different workload risks.

Q: Why is Erlang C more suitable for inbound call center staffing?

A: Erlang C estimates waiting in a queue with random incoming calls. Outbound work is usually released from a planned list, so a workload-capacity calculation is usually a more suitable starting point.

Q: Which call center metrics should a blended team keep separate?

A: Keep inbound access and resolution measures, such as service level, abandonment, transfer rate, and first-contact resolution, separate from outbound measures such as connection results, completed work, valid dispositions, and follow-up completion.

Q: When should a team use more than a simplified staffing formula?

A: Use production workforce-management forecasting or simulation when demand is volatile, callers abandon often, agents have different skills, callback behavior is important, or a missed service target has material customer or business risk.