Backlog flow model

Back-office staffing — workforce planning for case processing and admin

Back-office operations — claims, complaints, applications, letters, and case processing — cannot be staffed with Erlang C. They accumulate rather than queue. They have SLAs measured in days, not seconds. They require a backlog flow model. Here's how to build one.

What makes back-office staffing different

Contact centre (voice/chat)

• Contacts arrive in real time and must be answered immediately
• Service level = % answered in X seconds
• Queue length changes by the second
• Erlang C models the queue dynamics
• Understaffing = long waits + abandoned contacts

Back-office (cases/processing)

• Work arrives in batches and accumulates overnight
• SLA = % processed within X days
• Backlog changes by the day (or shift)
• Backlog flow model: inflow vs. throughput
• Understaffing = growing backlog + SLA breach

The backlog flow model

Back-office staffing is calculated with a flow equation, not a queue formula:

End backlog = Start backlog + Cases arriving − Cases processed

Cases processed per agent per day = productive hours × (3600 ÷ AHT seconds)

To calculate the FTE needed to clear a backlog within a target timeframe:

Define: current backlog size (cases), target clearance days, daily inflow, AHT (seconds per case), and working hours per agent per day.

Throughput needed = (current backlog ÷ target days) + daily inflow. This is the cases per day you need to process.

Agents needed = throughput ÷ (productive hours per day × 3600 ÷ AHT). Round up to the nearest whole agent.

Apply shrinkage: scheduled FTE = seated agents ÷ (1 − shrinkage rate). Use the shrinkage calculator to include breaks, meetings, and absence.

Worked example: complaints team

Scenario inputs

Current backlog200 cases

Target clearance5 business days

Daily new cases80

AHT per case25 minutes (1,500s)

Productive hours per agent per day5.5 hours (shrinkage accounted for)

Calculation

Throughput needed = (200 ÷ 5) + 80 = 40 + 80 = 120 cases/day

Cases per agent per day = 5.5h × 60 ÷ 25min = 5.5 × 2.4 = 13.2 cases

Agents needed = 120 ÷ 13.2 = 9.1 → 10 agents

At 10 agents, the 200-case backlog clears in 5 days while absorbing the 80 daily inflow. Apply shrinkage separately if the 5.5h already doesn't account for all shrinkage.

Common back-office work types and SLAs

Work typeTypical SLAModel

Complaints and dispute resolution5 days acknowledgement / 8 weeks resolutionBacklog flow + SLA breach tracking

Insurance claims processing5–10 business daysBacklog flow model

Mortgage and loan applications10–15 business daysBacklog flow model

Data Subject Access Requests (DSAR)30 calendar days (GDPR)Backlog flow + regulatory deadline tracker

Medical records and coding5–15 business daysBacklog flow model

Payroll and HR processingPayroll cycle (weekly/monthly)Throughput planning

Email and written enquiries1–3 business daysBacklog flow model

Order fulfilment and back-end processing24–72 hoursThroughput planning

Backlog vs. throughput planning

Two different planning questions have different answers:

Backlog clearing (SLA-driven)

“How many agents do I need to clear my current backlog to zero (or to SLA threshold) within N days?” Throughput must exceed inflow by the clearance rate. FTE is higher than steady-state.

Steady-state throughput

“How many agents do I need to keep up with the incoming volume without the backlog growing?” Throughput equals inflow. FTE is lower — but backlog never clears.

Use the email and ticket calculator to model both scenarios.

Frequently asked questions

Why can't I use Erlang C for back-office staffing?

Erlang C models an M/M/N queue: contacts arrive randomly, agents take one call at a time, and the queue is served in real time. Back-office work does not arrive in a real-time queue — it accumulates in a backlog. A case received at 10:00 is not abandoned if not answered within 20 seconds. It sits in the queue until an agent processes it. Erlang C has no concept of backlog, no concept of overnight accumulation, and no concept of an SLA measured in days rather than seconds. The correct model for back-office is a backlog flow equation.

What is the backlog flow model?

The backlog flow model computes: end-of-day backlog = start-of-day backlog + cases arriving today − cases processed today. Cases processed today = agents × productive hours per agent × (3600 ÷ AHT seconds). To hold the backlog constant, throughput must equal inflow. To clear a backlog within a target number of days, throughput must exceed inflow by enough to reduce the existing backlog to zero or to the SLA threshold.

How do I measure AHT for back-office work?

For back-office case handling, AHT (more accurately called Case Handle Time or Cycle Time) is the time from starting to process a case to completing it — including reading, analysis, decision, and any documentation. It differs from contact-centre AHT in that there is no 'hold time' or 'after-call work' distinction — all time is active case processing time. Measure it by tracking the time cases are assigned to an agent versus the time they are marked complete, averaged across a representative sample.

What is a realistic SLA for back-office processing?

SLAs for back-office operations are measured in days, not seconds. Common targets: complaint acknowledgement within 5 business days; case resolution within 8 weeks (regulatory); insurance claims processed within 5–10 business days; mortgage applications processed within 10–15 business days. The SLA determines how much backlog you can hold before a breach — and therefore how much buffer throughput capacity you need to maintain.