headphones, headset, music, black, speaker, electronic, gadget, headphones, headphones, headset, headset, headset, headset, headset. Call center workforce management software: what a US buyer should check first
Photo by StockSnap on Pixabay

Industry

Call center workforce management software: what a US buyer should check first

How-to for US call center WFM software buyers: check forecasting error by interval, editable shrinkage, multi-site wage rules, vendor security, and staffing ratios.

What to take away

  • Forecasting accuracy is the first check. Ask for MAPE by 15, 30, and 60 minute interval, not a single daily number.
  • Shrinkage assumptions must be editable per site and per queue. Defaults rarely fit Phoenix, Atlanta, and Dallas at once.
  • Multi-site wage handling requires state overtime rules and, for Canadian sites, Ontario scheduling law.
  • Those inputs set the agent-to-supervisor ratio and cost per contact that the Phoenix, Atlanta and Dallas page compares.
  • Security review of the WFM vendor is part of the purchase. Do it before signing.

Forecasting accuracy: demand a backtest

Ask the vendor to backtest on your historical data, not a clean demo set. A useful tool reports forecast error at 15, 30, and 60 minute intervals. Mean absolute percentage error (MAPE) under 5% for stable queues is a common target. No vendor should promise that without seeing your data.

Forecast error: ask vs. red flag

What to ask for

MAPE by interval
15/30/60 min error
Bias
Over/under by hour
Volume drivers
Bill cycle, marketing, weather

Red flag

MAPE by interval
Only daily accuracy
Bias
Always understaffs peaks
Volume drivers
"AI handles it"

A vendor that cannot show interval-level forecast error from your data is not ready for a serious evaluation.

Forecasting accuracy backtest

MetricWhat to ask forRed flag
MAPE by interval15, 30, 60 minute errorOnly daily accuracy
BiasOver or under forecast by hourAlways understaffs peaks
Volume driversBill cycle, marketing, weather"AI handles it"

Shrinkage assumptions you can edit

Shrinkage is the gap between paid hours and productive agent hours. Typical parts: breaks, lunch, coaching, team meetings, system downtime, unplanned absence. Ask the vendor to show default shrinkage percentages per site. A 30% default may fit one site but not another. If you cannot override by queue, the tool will misstate required staff.

Multi-site wage handling and scheduling rules

WFM software that schedules across US states must handle different overtime rules, meal break penalties, and predictive scheduling laws. For Canadian sites, Ontario's Employment Standards Act, 2000 sets scheduling, hours of work, and rest period rules. Test each site separately. A tool that works in Texas may fail in Oregon or Ontario.

Example: staffing a 120-agent Phoenix queue

Staffing a Phoenix queue

  1. Pull 13 weeks of contact volume by 30-minute interval.
  2. Run the vendor forecast and compare MAPE to your own moving average.
  3. Apply site shrinkage28% for Phoenix, 33% for Atlanta, 31% for Dallas.
  4. Convert to required agents using Erlang C or simulation.
  5. Add supervisor coverageone supervisor per 15 agents on the floor.

For the full worked example of agent-to-supervisor and shrinkage math, see how Phoenix, Atlanta and Dallas differ. That page also lists BLS wage inputs and a span of control table.

Wage inputs and labor supply

Use external wage data to check your cost per agent. The BLS Occupational Outlook Handbook reports median pay and employment outlook for customer service representatives. If your forecast assumes 24/7 coverage, check the labor supply in each metro. Phoenix, Atlanta, and Dallas have different competition for the same agents. Treat that as a floor, not a ceiling.

Vendor security and data handling

WFM tools hold agent schedules, wage rates, and personal data. Ask for a security questionnaire aligned to the NIST Cybersecurity Framework. Check audit logging, role-based access, and data retention. A vendor that cannot show these controls creates risk for every site you schedule.

What the contract must cover

  • Data ownershipyour interval-level volume and agent data must remain yours and be exportable.
  • Forecast refreshvendor must retrain or recalibrate after any major volume shift.
  • Shrinkage audityou need the right to run your own shrinkage study each quarter.
  • Site-level wage rulesvendor must update tax and overtime tables when state or provincial law changes.
  • Exit planyou must be able to extract schedules, forecasts, and historical data in a standard format.

Common questions

What is a good forecasting accuracy target?
For stable queues, MAPE under 5% at 30-minute intervals is a common benchmark. For volatile queues, 10% may be realistic.
How do I check shrinkage assumptions?
Ask the vendor to show default shrinkage by site and whether you can override by queue. Compare defaults to your own timekeeping data.
Does WFM software handle multi-state wage rules?
Most tools handle basic overtime but few handle predictive scheduling penalties. You must test each state and province.
Why link staffing ratios to forecasting inputs?
Agent-to-supervisor ratios and shrinkage rates are inputs to the staffing plan. Bad forecasts make those ratios meaningless.

More in Industry

Latest from Market Desk