The true biaya contact center of one inbound call is fully loaded operating cost divided by resolved contacts, not the telephony tariff. Add labor, licenses, training, quality, and channel fees, then divide by volume. That unit cost is what automation should beat.
Why the telco rate is not the real number
Many teams still quote the operator’s per-minute rate as if it were the cost of a call. It is one line on an invoice. An inbound call also pulls agent pay, benefits, training, supervisors, desktop licenses, recording, and either floor space or a BPO fee.
After-call work is paid time. If an agent spends two minutes writing notes after a five-minute conversation, you paid for seven minutes, not five. An honest unit cost uses occupancy time, not talk time alone.
Channels in Indonesia are not interchangeable. Voice, WhatsApp, email, and social each have different handle times and fee structures. Blending them into one average without a per-channel unit cost turns automation decisions into guesswork.
WhatsApp is a dominant business channel in Indonesia. Since 1 July 2025, Meta has charged WhatsApp template messages per message, not per 24-hour conversation. Notification and follow-up templates belong in the model, separate from human reply sessions.
An auditable formula
Start with the monthly operating cost that belongs to customer service. Include frontline pay and benefits, supervisors, workforce management, and quality assurance. Add platform seats, telephony or SIP, numbers, and channel fees.
Divide that total by contacts actually resolved in the same period. Do not divide by offered calls if many are abandoned or repeated. The same customer returning the same day is one problem, not two successes.
Write the formula on a sheet internal audit can open. Cost per contact equals fully loaded monthly cost divided by resolved contacts. Split it by channel so voice economics do not hide messaging economics.
Voice cost per contact is almost always higher than messaging, because handle time is longer and the work is real-time. That is not a reason to kill voice. It is a reason to move simple questions to a cheaper channel.
Line items teams forget
Labor and shrinkage
Base salary is only the start. Benefits, overtime, night shifts, and non-productive time — breaks, briefings, training, absence, idle — all sit in the numerator. Shrinkage means one FTE is not one productive hour on the phone.
Ignore shrinkage and cost per contact looks cheaper than it is. That becomes a problem the moment you promise savings to the board. Count paid hours first, then hours that actually handled a customer.
Indonesian labor rules also limit how far you can squeeze cost through working hours. Unplanned overtime wrecks unit cost faster than a channel tariff. A brittle roster looks cheap on paper until peak season arrives.
Licenses, recording, and quality
Seat licenses, recording, quality management, and the knowledge base rarely appear in a “cost per call” slide. Without them, a regulated operation should not run. Banks and multi-finance firms need an audit trail. That is a compliance cost, not an optional extra.
Indonesia’s UU PDP (Law 27/2022) has been fully enforced since 2024. Recordings, transcripts, and chat history are personal data. Storing, restricting, and deleting them has a cost. Leave that out, and your unit cost is incomplete.
Digital channel fees
Do not mix WhatsApp template fees with agent labor. An automated message and a human conversation have different economics. Mass templates that produce no resolution only move waste from payroll to a Meta invoice.
Keep three buckets: mandatory utility messages, human-handled service conversations, and automation that finishes without escalation. Only the third bucket may be claimed as savings.
AI contact center, omnichannel service and compliance notes for Indonesian service leaders.
Every Monday, five minutes.
How to calculate without inventing industry benchmarks
Do not paste an unverified global benchmark into a board pack. Take your payroll file, platform invoices, and system volume. Those three sources are enough for an honest first version.
The steps are plain. Take fully loaded cost for a closed month. Group it into labor, technology, channels, and operating overhead. Pull volume by channel from the platform, not from a supervisor’s estimate.
Compute unit cost by channel, then a blended figure. Repeat across three months so seasonality does not mislead. If WhatsApp volume rises while voice falls, blended cost per contact can drop with no agent efficiency gain.
That is mix shift, not an operations win. Report both.
What the number is for
Unit cost answers three decisions. Which channel is too expensive for a given question type. Where AI and self-service should be installed first. Whether insourcing, BPO, or a hybrid model fits the load.
Deflection is only worth money if the deflected contact is actually finished. A callback means you paid twice. Tie cost per contact to first contact resolution. Never read unit cost alone.
ConverseAIght is built for Indonesian omnichannel operations, with voice, WhatsApp, and AI agents on one case timeline. When context moves channel, the agent does not re-ask data already captured. That removes hidden handle time that rarely shows up in a weekly report.
Meta’s WhatsApp Business Summit Indonesia on 11 August 2026 reported that 71 percent of Indonesians message businesses daily, up from 60 percent in 2024 and 66 percent in 2025. Digital volume will keep stressing your formula. Build per-channel unit cost before that volume distorts the company average.
Classic errors in contact center finance
Dividing cost by headcount, not contacts
Cost per agent is an HR metric, not a service-design metric. Customers do not buy FTEs. They buy resolution. If volume rises faster than headcount, cost per agent can rise while cost per contact falls.
Ignoring repeat contacts
The ticket is closed; the customer comes back. If the report uses closed tickets as the denominator, unit cost looks healthy while customers are frustrated. Count unique issues finished, not CRM status changes.
Comparing a BPO invoice to an internal cost that is not fully loaded
BPO invoices often already include occupancy and shrinkage. Internal cost often does not. Raw comparisons mislead. Align definitions first, then compare vendor and in-house.
Frequently asked questions
What is the real cost of one inbound call?
Fully loaded operating cost for the month divided by voice contacts actually resolved. The operator’s per-minute tariff is not a substitute for that formula.
Is WhatsApp cheaper than phone?
Often, because handle time is shorter. Templates have been priced per message since 1 July 2025, so calculate separately and compare for the same question type.
Should office overhead go into the formula?
Include overhead that would disappear if the operation stopped or moved. Exclude holding-company costs that have nothing to do with customer service.
What about seasonal volume?
Use a multi-month average and show a range, not a single point. Ecommerce campaign peaks and finance close periods change the denominator.
Does automation always lower cost per contact?
Only if the contact finishes without bouncing to a human. Deflection that creates a callback raises unit cost rather than lowering it.
How often should we recalculate?
Every operating month, and whenever channel tariffs, shift mix, or the staffing model change.
Book a demo and see your own contact flows running on ConverseAIght
ConverseAIght
