Skip to content
Knoativ

Contact center operations

Call Center Staffing for Peak Season: A Planning Playbook

Forecast, staff and automate your contact center for the holidays, Ramadan and every peak.

Knoativ TeamRetail CX Research12 min read

Shopper on a phone call pushing a full cart through a festive supermarket

Call center staffing is matching the number of agents on the phones, interval by interval, to the contacts you expect, so customers are answered within your service level target. For a retailer, the hard part is the peak: forecasting it by half-hour, staffing it with the right mix of people and automation, and protecting quality when volume doubles.

Picture 9:30 p.m. on the Monday after Thanksgiving. The promotion went out at 6 p.m., and 140 callers are waiting, half of them asking whether their order will arrive in time. Your agents are good; there are just not enough of them, and the October hires are still learning the returns policy. This playbook is about making that night go differently.

Why does call center staffing break during peak season?

Staffing breaks at peak because volume, timing and complexity all change at once, and the plan was built from an average. More orders produce more questions, the questions arrive in sharper bursts, and the new hires answering them are the least experienced people in the building.

The scale is real. The National Retail Federation forecasts that US holiday retail sales in November and December 2026 will grow 4.2% to $1.013 trillion, and expects retailers to hire 325,000 to 375,000 seasonal workers. Every one of those orders can produce a "where is it?" or "can I change it?" contact, and some of those seasonal hires will be on your phones.

Three things make it harder than the volume suggests:

  • Bursts, not averages. A promotion email or a delivery delay can triple calls within an hour. A daily forecast hides that.
  • Deadline anxiety. Gifts must arrive before the holiday. Customers who would wait patiently in March call twice in December.
  • A different mix. Order status, delivery changes and promo problems grow faster than total volume.

How do you do call center forecasting for a peak?

Call center forecasting for a peak means predicting contacts by day, by half-hour interval, by channel and by contact reason, using last year's peak as the base and this year's order forecast as the driver. Forecast contacts per order, not just contacts, so the plan moves when the sales forecast moves.

  1. Pull last year's peak data by half-hour, channel, language and contact reason. Include the week after the peak, when returns and exchanges arrive.
  2. Calculate contacts per 100 orders by reason. This ratio is more stable than raw volume and links your plan to the commercial forecast.
  3. Apply this year's order forecast by day, then spread it across intervals using last year's intraday pattern.
  4. Add known events: promotion launches, delivery cut-off dates, carrier capacity limits, a new returns policy.
  5. Subtract what you will prevent, but only once the fix is live and measured.
  6. Build a high and a low case. Staff to the base case, and know in advance what you will do if you hit the high one.

During the peak, compare forecast and actual by interval every day and adjust the next day's plan.

What does an Erlang calculator do? A worked Erlang C example

An Erlang calculator turns three inputs (calls per interval, average handle time and a service level target) into the number of agents you need on the phones. Most use the Erlang C formula, which estimates the probability a caller has to wait, given the traffic and the number of agents.

As Call Centre Helper explains, an Erlang calculator tells you the number of staff you need for a given number of calls to meet a given service level, and some versions also take shrinkage into account. Here is the logic, step by step, so you can sense-check any tool you use.

The inputs (illustrative)

  • Calls per interval: 120 calls per 30 minutes on a normal evening; 300 per 30 minutes at the holiday peak.
  • Average handle time (AHT): 6 minutes, including after-call work.
  • Service level target: 80% of calls answered within 20 seconds.
  • Shrinkage: 30%, the share of paid time agents are not available to take calls (breaks, training, meetings, absence).

The steps

  1. Traffic intensity. Multiply calls by handle time and divide by the interval length. At peak: 300 calls × 6 minutes ÷ 30 minutes = 60 Erlangs. In plain terms, 60 agents would be busy every minute of that half-hour just to keep up.
  2. Agents on the phones. Erlang C adds agents one at a time above 60 and calculates the service level for each, until it reaches 80% in 20 seconds. Here that happens at 67 agents.
  3. Shrinkage. Divide by (1 − shrinkage): 67 ÷ 0.7 = 95.7, so roughly 96 agents scheduled for that interval.
Scenario (illustrative)Calls per 30 minTraffic (Erlangs)Agents on phonesService level reachedOccupancyScheduled with 30% shrinkage
Normal evening120242981.6%83%42
Holiday peak300606780.9%90%96
Holiday peak, 40% of calls resolved by AI first180364282.4%86%60

Illustrative: all rows assume 6-minute handle time and an 80/20 target, computed with the standard Erlang C formula.

What the example teaches

Being a few agents short hurts more than you expect. At peak, 67 agents reach about 81% of calls in 20 seconds, with an average wait of roughly 14 seconds. With 64 agents, just three fewer, the service level falls to about 59% and the average wait more than triples, to roughly 46 seconds. Lines grow faster than headcount shrinks.

Watch occupancy. At 67 agents, people are busy 90% of the time, which is hard to sustain for weeks. Call Centre Helper's worked Erlang C example caps occupancy at 85%; applying that cap here means 71 agents on the phones and about 102 scheduled.

Know the limits. Erlang C assumes nobody hangs up and all agents work at the same speed, so treat the result as a floor and check it against last year's actuals.

TipIf AI resolves the simple calls first, the calls left for people are harder and often longer. Re-measure handle time for the remaining calls before you plug them into the calculator, rather than keeping last year's average.

Which call center metrics matter during peaks?

During a peak, watch the call center metrics that show whether customers are getting through and getting answers: service level and abandonment by interval, repeat contacts, and forecast accuracy. Average handle time matters, but pushing it down at peak usually creates callbacks.

MetricWhat it tells you at peakCheck it
Service level by intervalWhether customers are answered within target in the busy half-hours, not just on averageEvery 30 minutes
Abandonment rateCallers giving up; often the first sign the plan is shortEvery 30 minutes
Contacts per 100 orders by reasonWhich reasons are growing faster than ordersDaily
Repeat contact rateCustomers calling back on the same issue; rising repeats mean answers are not fixing anythingDaily
Average handle timeWhether new hires and new policies are slowing callsDaily, by tenure
Verified resolution of AI-handled contactsWhether automation is resolving, not just deflectingDaily, by contact reason

Workforce management call center basics: shifts and intraday changes

Workforce management (WFM) turns the forecast into schedules and adjusts them through the day. At peak it comes down to shifts that follow the demand curve, a plan for the day you are wrong, and someone with authority to move people in real time.

The workforce management call center basics are simple to state and hard to keep under pressure:

  • Shape shifts to the curve. Split shifts and part-time evening shifts cover a 6 p.m. to 11 p.m. surge better than adding full days.
  • Move training and meetings out of the peak. Shrinkage you can schedule, schedule for the quiet weeks before.
  • Have an intraday playbook. If abandonment passes a threshold, who gets pulled from email to phones, who cancels a break, who approves overtime? Write it down before the peak.
  • Protect the experienced people. Route complaints to your best agents; give new hires, or AI, the simpler contacts.

The Peak Readiness Timeline: eight weeks to any peak

The Peak Readiness Timeline starts eight weeks before the busiest day. That leaves time to forecast, fix the data, set up proactive messages and test, rather than hiring in a panic during the peak. Count back from Black Friday, the last holiday shipping date or Eid.

WhenWhat to do
8 weeks outPull last year's peak data by interval and reason. Build the forecast in contacts per 100 orders. Agree delivery cut-offs with logistics.
7 weeksDecide which contact types AI will resolve, assist or route to people. Confirm it can read live order, carrier, stock and loyalty data. Book temporary staff now if you still need them.
6 weeksWrite proactive messages for each order stage. Publish peak store hours and the holiday returns and exchange policy.
5 weeksRun the Erlang C numbers by interval and build shifts. Write the intraday playbook and escalation rules.
4 weeksTrain new hires on the top ten contact reasons first. Load test phone lines, chat and tracking pages.
3 weeksGo live with proactive messages and AI coverage on a slice of traffic, so both are stable before the rush.
2 weeksAnnounce last order dates for on-time delivery everywhere. Review the first weeks daily and fix the top three issues.
Peak weekDaily forecast vs actual, intraday adjustments, a senior lead on call each evening. Send delay alerts before customers notice.
2 weeks afterPost-peak review: forecast vs actual, top contact reasons, cost per resolved contact, and three changes for next time, each with an owner.

Many peak order status calls are preventable. Our guide to WISMO and where-is-my-order calls has the order-update plan; and if stock is out of sync between web and store, our guide to retail inventory management covers the canceled orders that drive the angriest peak calls.

Peak seasons worldwide: holidays, then Ramadan and Eid

Every market has its own peaks. The playbook stays the same; the shape of the curve changes.

In the US and Europe, the peak runs from Black Friday and Cyber Monday through the last shipping date before the holidays, then flips to returns and exchanges in early January. Grocery retailers see their own spikes around Thanksgiving and holiday meals, a pattern familiar to anyone running grocery, supermarket and department store service.

The GCC example: Ramadan and Eid

In the GCC, Ramadan moves demand later into the night. Visa's analysis of the Ramadan night-time economy found that in Saudi Arabia 35% of Visa transactions happened between 10 p.m. and 4 a.m. during Ramadan, up from 17.6% in a normal month. Orders placed after midnight produce questions at hours most teams do not staff, while agents who are fasting have their own iftar and family time to protect.

Two lessons travel well. Split shifts around the dip in demand (here, the iftar hour) and staff the late window for exceptions, with AI handling order status. And serve every language as first class: many GCC customers mix Arabic and English in one sentence, and a transfer should keep the customer's language.

Plan with a buffer. Ramadan 2027 is expected to begin around February 8, with Eid al-Fitr around March 10, as Gulf News reports; exact dates depend on the moon sighting.

Staffing vs automation: what does peak coverage cost?

Temporary staff add capacity, but they cost money before they take a single call, take weeks to train and leave as they become good. Automating routine contacts reduces how many extra people you need, and the capacity stays for the next peak.

An illustrative comparison. Every figure is an assumption; replace them with your own.

  • The peak adds 2,240 contacts a day for 35 days: 78,400 extra contacts.
  • A temporary agent handles 40 contacts per shift, works five days a week and needs 10 paid training shifts.
  • A temporary agent shift costs $120 all-in, including agency fees and supervision.
  • In the automation scenario, AI resolves half of the extra contacts at an assumed $1 per resolved contact. Ask each vendor how they price.
IllustrativeTemporary staff onlyAI for routine contacts, plus fewer temporary staff
Extra contacts handled by people78,40039,200
Agent shifts needed per day5628
Temporary hires (five-day rota)7940
Paid shifts, including training2,7501,380
Staffing cost$330,000$165,600
AI cost$0$39,200
Total peak cost$330,000$204,800

In this example, automation cuts peak cost by about 38%. And 79 temporary agents still need to be found, trained and supervised late at night; some will leave in week two.

The biggest lever is not in the table: every contact prevented by a good delivery message costs nothing to answer.

Plan your next peak before it plans you. WiseRep AI resolves order status, delivery changes and stock questions on voice and chat at any hour, sends proactive updates and hands complaints to your team with full context. Book a peak readiness call and we will plan staffing and automation together on your own contact data. Book a peak readiness call.

Getting call center staffing right for the next peak

A calm peak is decided weeks before it starts: a forecast by interval, an honest Erlang C check, shifts shaped to the curve, proactive messages that remove calls, and AI for the routine contacts.

Prevention and automation work best on the same data: the live order, the carrier status, store stock and the customer's history. That is how WiseRep AI is built, as part of Knoativ's approach to AI customer service for retail, and our assessment, build and operate services can run the planning with your team. To decide which calls AI should own, see our guide to the AI call center for retail.

If your next peak is weeks away, book a peak readiness call and bring last year's peak contact data. We will plan staffing and automation with you, interval by interval, and show which calls to prevent, automate and staff.

Frequently asked questions

How do you calculate call center staffing?

Forecast calls per half-hour interval, multiply by average handle time to get traffic in Erlangs, then use the Erlang C formula to find the agents needed to hit your service level target, such as 80% of calls in 20 seconds. Divide by one minus shrinkage to get scheduled staff, and check that occupancy stays at a sustainable level, often 85% or less.

What is an Erlang calculator?

An Erlang calculator is a tool that tells you how many agents you need for a given number of calls, handle time and service level target. Most are based on the Erlang C formula. It is a useful floor, but it assumes no callers hang up and all agents work at the same speed, so check the result against your own historical data.

Which call center metrics matter most during peak season?

Service level and abandonment by half-hour interval show whether customers are getting through. Repeat contact rate shows whether answers actually fix anything. Forecast accuracy and contacts per 100 orders by reason show where the plan is wrong. Track verified resolution for AI-handled contacts too, so automation is measured on outcomes rather than deflection.

When should retailers start call center forecasting for a peak?

Start about eight weeks before the busiest day. That gives time to build a forecast by interval from last year's peak, agree delivery cut-offs, write proactive messages, decide which contacts AI will handle, hire and train any temporary staff and test everything before volume rises. Starting later usually means less training and a weaker forecast.

Is automation cheaper than temporary staff for peak season?

Often, once you count recruitment, paid training, supervision and attrition. In an illustrative example, automating half of the extra contacts cut peak cost by about 38%. The bigger benefit is consistency: AI answers routine questions the same way at any hour and is ready for the next peak, while temporary staff leave once the season ends.

Knoativ Team

Retail CX Research

The Knoativ editorial team covers identity-led retail strategy, customer data platforms, and the technology behind unified commerce experiences.

Follow on LinkedIn

Filed under

  • Contact center operations
  • AI Customer Service
  • WiseRep AI
  • Peak season
  • Workforce planning

Ready to Know Every Customer?