How do I set up workforce management software for forecasting staffing needs and tracking SLAs in a high-volume support center?

Here’s a practical way to set up workforce management (WFM) software for a high-volume support center so it can forecast staffing needs and track SLAs accurately.

1) Define the operational model first

Before configuring the tool, document these basics:

• Channels: voice, email, chat, social, tickets, back-office
• SLAs / targets:
  • Voice: answer speed, abandon rate, service level
  • Email/tickets: first response time, resolution time
  • Chat: response time, concurrency rules
• Hours of operation
• Shrinkage assumptions:
  • breaks, meetings, training, PTO, coaching, attrition, system outages
• Skill groups / queues
• Priority rules:
  • VIP, escalations, product lines, language skills
• Routing logic:
  • single-skill vs. multi-skill, overflow rules, callbacks

If this isn’t clear, the software will forecast “staffing” but not the right staffing.

2) Connect your data sources

WFM needs historical and real-time data from:

• ACD / telephony system for voice:
  • interval contacts offered, answered, abandoned, handle time, ASA, occupancy
• Ticketing / CRM for non-voice:
  • arrival volume, backlog, handle time, aging, response times
• HR / scheduling system:
  • agent availability, PTO, training, shift patterns
• QA / case management if relevant
• Real-time feeds:
  • current queue depth, staffing, adherence, occupancy

Make sure data is at a consistent interval, usually 15 minutes for voice and hourly/daily for digital channels.

3) Clean and standardize historical data

Forecasting quality depends heavily on data cleanliness.

• Remove outliers from outages, major incidents, holidays, launches
• Mark and separate special events rather than deleting them
• Standardize time zones and interval boundaries
• Ensure handle times are measured consistently
• Separate contact types if they behave differently
• Normalize for seasonality:
  • day of week, month, holiday effects, payday, campaigns

A good practice is to keep:

• raw data
• adjusted/clean data
• exception log

4) Build forecasting models by queue and interval

Forecasting usually happens in three parts:

A. Volume forecast

Predict how many contacts you’ll receive by queue and interval.

Use:

• historical seasonality
• trend
• promotions/campaigns
• holiday calendar
• known events
• regression factors if the platform supports them

B. AHT / handle time forecast

Predict average handle time by queue, contact type, and agent group.

Include:

• talk time
• hold time
• wrap time
• transfer patterns
• complexity changes over time

C. Concurrency / arrival pattern assumptions

For chat or digital channels:

• average concurrent chats per agent
• response-time targets
• backlog behavior

The tool should then convert volume + AHT + target SLA into required staffing.

5) Configure SLA logic correctly

Define SLA metrics precisely in the software:

• Voice: “80% answered within 20 seconds”
• Chat: “90% first response within 30 seconds”
• Email: “95% responded within 4 hours”
• Tickets: “resolved within 24 hours” or by priority tier

Also define:

• abandonment rules
• voicemail/callback handling
• after-hours rules
• queue priority weighting
• whether SLA excludes certain contact types

Avoid ambiguity: two teams often say “SLA” but mean different things.

6) Set shrinkage and occupancy assumptions

This is where many forecasts fail.

Shrinkage

Estimate total planned and unplanned shrinkage:

• PTO
• breaks/lunches
• training/coaching
• team meetings
• system downtime
• absenteeism
• quality calibration
• admin work

A high-volume center often starts with:

• planned shrinkage: 20–30%
• add unplanned shrinkage on top, based on history

Occupancy

Define target agent utilization:

• voice: often 80–85% depending on complexity
• chat: lower if concurrency is high or cognitive load is heavy
• digital backlog work: based on throughput and SLA targets

Use realistic occupancy; if it’s too high, schedules will look efficient but performance will degrade.

7) Translate forecast into staffing requirements

Your WFM software should calculate required agents using Erlang or interval-based models for voice, and workload models for digital channels.

For each interval, the system should consider:

• forecasted contacts
• AHT
• target service level
• abandon tolerance
• shrinkage
• occupancy target
• skill group constraints

Then it outputs:

• required paid staff
• required on-queue staff
• staffing gap/surplus by interval
• daily and weekly FTE needs

8) Build schedules based on forecasted demand

Once the staffing requirement is set:

• Define shift templates
• Add breaks, lunches, meetings
• Align shifts to peak intervals
• Use split shifts or staggered start times if needed
• Apply labor rules and union constraints
• Optimize for adherence to demand curve, not just headcount

Validate that schedules cover:

• peak intervals
• channel-specific spikes
• coverage for meetings/coaching
• skill coverage for each queue

9) Set up real-time monitoring and alerts

Forecasting is only useful if it’s connected to live operations.

Track:

• service level by interval
• queue depth
• ASA / response time
• abandon rate
• occupancy
• adherence
• shrinkage variance
• backlog aging
• schedule conformance

Configure alerts for:

• SLA at risk
• staffing below threshold
• unexpected spikes
• agent lateness / absences
• queue imbalance

This lets supervisors intervene early with:

• overtime
• voluntary time off changes
• reassignments
• skill-based routing shifts
• callback pushes

10) Create a daily operating rhythm

A strong WFM setup needs regular review:

Daily

• yesterday’s forecast vs actual
• interval-by-interval SLA performance
• staffing adherence review
• exceptions and incident notes
• intraday reforecast

Weekly

• forecast accuracy review
• shrinkage review
• schedule efficiency
• trend changes by queue
• staffing recommendations

Monthly

• model tuning
• seasonal recalibration
• target review
• staffing plan updates

11) Measure forecast accuracy and SLA performance

Common KPIs to configure in dashboards:

Forecasting

• volume forecast accuracy
• AHT forecast accuracy
• interval accuracy
• MAPE or bias
• service level forecast accuracy

Staffing

• required vs scheduled vs actual
• shrinkage realized vs planned
• adherence
• occupancy

SLA / CX

• service level
• abandon rate
• first response time
• resolution time
• backlog age
• customer satisfaction if available

12) Pilot before full rollout

Don’t launch across all queues at once.

Pilot with:

• one or two major queues
• one voice + one digital channel
• one site or one team

Test:

• data quality
• forecast logic
• schedule generation
• intraday adjustments
• SLA reporting

Then refine before expanding.

13) Common mistakes to avoid

• Using bad historical data
• Ignoring shrinkage
• Forecasting at too coarse an interval
• Treating all contacts as equal
• Failing to separate new vs known issues
• Over-optimizing schedules and under-covering peaks
• Not updating forecasts intraday
• Not validating SLA definitions with operations leaders

Recommended setup sequence

A simple implementation order:

1. Define SLA and workforce rules
2. Integrate historical and real-time data
3. Clean and standardize data
4. Build volume and AHT forecasts
5. Apply shrinkage and occupancy assumptions
6. Generate staffing requirements
7. Create schedules
8. Set real-time monitoring
9. Review forecast accuracy
10. Tune monthly

If you want, I can also give you:

• a sample WFM configuration checklist
• a dashboard layout for SLA and staffing
• or a step-by-step setup for a specific tool like NICE, Verint, Genesys, or Calabrio.