If Black Friday is a few weeks away and your hiring classes are already full or closed, WFM still has real levers. You cannot add trained agents in three weeks, but you can protect the agents you have, move offline work, shape demand, tighten the forecast and agree in advance what you will do when service level drops.
The US National Retail Federation (NRF) forecast, published on 15 October 2026, projects holiday sales for 1 November to 31 December to grow about 4.2% over 2025, to about $1.013 trillion. It expects retailers to add roughly 325,000 to 375,000 seasonal workers. The figure excludes car dealers, petrol stations and restaurants. For contact centres that serve retail and ecommerce, whether in the US or offshore, that means another strong peak, and the part of the plan that still moves in November is operations, not hiring.
This guide is for planners and team leads. It covers what peak demand does to headcount, which levers still work, what to agree before the peak starts and what to do on the day.
What a peak does to headcount
Here is a standard Erlang C example. It is illustrative, with no shrinkage and no abandonment, at an 80/20 service level target.
| Scenario | Calls per half hour | AHT | Agents on the phones needed |
|---|---|---|---|
| Normal day | 300 | 300 seconds | 57 |
| Peak: 20% more volume | 360 | 300 seconds | 67 |
| Peak: 20% more volume, 10% longer calls | 360 | 330 seconds | 73 |
| Peak: 40% more volume, 10% longer calls | 420 | 330 seconds | 85 |
Two things matter here:
- Volume and handle time compound. A 20% volume increase needs about 18% more agents. Add 10% longer calls, which is common in peak because queries are about orders, returns and delivery problems and agents are tired, and the need rises by about 28%.
- Small shortfalls are expensive. At the 360 calls, 330 second case, 73 agents deliver about 80%. With 70 agents service falls to about 59%. With 68 it falls to about 35%. Three agents short is not three points. It is the difference between a good and a failed interval.
This is why the work in the last weeks is about protecting every available agent-hour. For why the curve is so steep, see Why Adding Agents Stops Helping, and you can run your own numbers in the free Erlang C Calculator.
Levers that still work
1. Protect shrinkage. This is the cheapest capacity you have. Freeze non-essential training, coaching and team meetings for the peak window, and move them to before or after. Review planned leave rules and approved leave dates. Planned shrinkage of 19% that drops to 12% gives you about 9% more productive hours without a single hire. See how to calculate shrinkage in a BPO for the planned and unplanned split.
2. Plan unplanned absence for the bad days, not the average. Peak weeks and winter sickness tend to bunch. Plan an absence buffer by day of week. See Planning for Average Absence Leaves 4 in 10 Intervals Short.
3. Use flexible shifts and voluntary overtime. Ask for volunteers early, for specific intervals, and pay attention to fatigue. Overtime has a ceiling. A Verint planning example describes 80 agents each working 4 extra hours a week, which adds 320 hours against 800 needed, and the modelled service level falls from 80/20 to roughly 60/20. Overtime tops up a gap. It does not close a large one.
4. Shift offline work. Email backlogs, projects and QA can wait. Pre-agree which work stops at which trigger (see the plan below).
5. Tighten the forecast. Forecast the peak at 15 or 30 minute intervals, using last year's peak shape, not a monthly average. Add known drivers such as promotions, shipping cut-offs and marketing sends. Under-forecasting costs far more than over-forecasting. See The Forecast Error Asymmetry.
6. Use callbacks carefully. A callback offer when the wait is long can protect abandonment, but it does not remove the workload, it moves it. Plan the callback workload into later intervals. See Callbacks Move Workload, They Do Not Remove It.
7. Cross-skill where you can. A pooled team is more efficient than separate small teams. Check which agents can safely take a second queue for the peak. See The Multi-Skill Pooling Effect.
8. Fix the top call drivers. A short FAQ for the top five reasons people call (delivery status, returns, promotions, payment, account access) shortens hold time and can reduce handle time. Look at the first week of your peak data, then adjust the IVR message and the chat or knowledge base answers.
Agree the real-time plan now
The worst time to decide what to cancel is in the middle of a bad interval. One planning guide suggests a one page plan with thresholds. Here is an example you can adapt:
| Service level vs goal | Action |
|---|---|
| About 5% below goal | Pause or reschedule non-critical offline work |
| About 10% below goal | Cancel all offline work, offer voluntary overtime or an early return from breaks where policy allows |
| About 20% below goal | Bring team leaders, coaches and QA onto the phones, trigger overflow or callback offers |
Write down who has the authority to trigger each step. For how to run the day, see Intraday Management: How to Recover Service Level in Real Time and Why Your Intraday Reforecast Trigger Is Firing Too Late.
Look after the people
Peak weeks push occupancy up and tire agents. Plan breaks and protect them. Offer shift swaps through a simple process, and thank people. The occupancy cost shows up later as attrition. See The Occupancy and Retention Link and Agents Want Schedule Flexibility.
A week by week checklist
Now, three weeks out
- Freeze non-essential training, meetings and leave approvals for the peak window.
- Rebuild the interval forecast from last year's peak, adjusted for this year's growth and campaigns.
- Calculate the gap in agent hours by interval, with shrinkage added.
- Ask for overtime and flex volunteers.
Two weeks out
- Agree the one page real-time plan and who can trigger each step.
- Check cross-skilling and routing for overflow.
- Update the IVR and knowledge base for the top call drivers.
- Plan callback workload into later intervals.
Peak week
- Review actual vs forecast every half hour and re-forecast early.
- Follow the thresholds. Do not wait to be sure.
- Watch occupancy and breaks, and protect them.
After the peak
- Record actual volume, AHT and shrinkage by interval.
- Keep the forecast, the plan and the lessons for next year.
- Plan the return to normal headcount, especially for any temporary capacity.
Next year, start earlier. Capacity for a fall peak depends on lead times such as about 90 days for hiring, plus training and supervised practice, so decisions are due by late spring. See Peak Season Staffing: Why September Is Already Late and Seasonal Hiring Start Date.
Want a quick view of where your planning is weakest before the peak? Try the free WFM Health Check.
FAQ
How many agents do I need for Black Friday? It depends on your volume, handle time and service level target. In the example above, a 20% volume rise with 10% longer calls needed about 28% more agents on the phones, before shrinkage. Use an Erlang C calculation by interval and then add shrinkage.
What can WFM do when hiring is already closed? Protect shrinkage, use flexible shifts and voluntary overtime, move offline work, tighten the interval forecast, plan callbacks and cross-skill agents.
Does overtime solve a staffing gap? It helps with small gaps. For large gaps it falls short, because a few extra hours across the team adds only a fraction of the hours you need, and it tires people.
Why do small staffing shortfalls hurt so much at peak? Because queues grow quickly when agents run close to full occupancy. In the example, being three agents short dropped service level from about 80% to about 59%.
When should planning for next year's peak start? Late spring for a fall peak, since hiring, training and any outsourcing decisions need several months of lead time.
What should the real-time plan include? Thresholds for service level below goal, the actions at each threshold, and the person who can trigger them. Agree it before the peak.
Sources: NRF: forecast for 2026 holiday sales (15 October 2026), Peopleware: Planning for peak season and Verint: Contact center capacity planning for fall peak season. The staffing examples use the standard Erlang C formula and invented figures calculated by the WFM Helper team.