Why you should schedule staff past official closing time
The worst-staffed interval of the day
Look at the SL pattern of almost any contact centre that closes at a fixed time and you’ll see the same curve. Service holds steady through the day, dips briefly around lunch, recovers, and then quietly collapses in the final 20–30 minutes before close. It’s the worst-staffed interval of the day, hiding in plain sight, and almost no operation schedules to fix it. The customers who phone at 7:50pm wait twice as long, abandon more often, and rate the operation more harshly than the customers who phone at 7:20pm — despite the contact reason being the same. The fix is structural and cheap; the reason it isn’t already done is mostly cultural.
Why the trough exists
Three forces converge to empty the floor before close.
Last-call avoidance. Agents who are scheduled to finish at 8pm will, if left to their own devices, decline to start a new contact at 7:50pm because they don’t want a 10-minute call running over their shift end. The behaviour is rational; the consequence is that the operation has nominal capacity in the 7:50–8:00 interval but no-one is taking work.
Wrap-up and admin compression. Agents push their day’s outstanding ACW, system notes, and case updates into the final 15 minutes. The operation reads as “staffed” on the WFM screen; the ACD reads as “available” on no-one.
Personal admin and goodbye logistics. The team goodbye, the locker, the bus, the train. None of this is unreasonable; all of it competes with the queue for the agent’s attention in the final ten minutes.
The four-minute conversation a planner has with a frustrated team leader at 8:05pm is usually about the four minutes of SL collapse the agents themselves didn’t see. The dashboard shows full headcount, full adherence, and service level on the floor.
What the trough actually costs
The cost shows up in four places, none of which the operation usually links to the closing hour.
The first is service level on the day. A 10-point SL drop across a 30-minute interval is a 5-point drop across the closing hour, which is a 1-point drop across the operational day on a typical operation. That’s the difference between hitting target and missing it for some operations.
The second is abandonment. Customers who reach the queue at 7:50 face a multi-minute wait, the “we close at 8pm” message, and a real decision about whether to keep waiting. Abandonment in the closing 30 minutes is reliably 2–4x the daytime rate.
The third is the next morning’s opening. The customers who abandoned at 7:50 phone back at 9:01 the following day. The operation’s morning queue is built partly from the previous evening’s failure to clear it. A small under-staffed closing layer creates a small over-staffed opening problem twelve hours later.
The fourth is CSAT and complaint volume. Contacts in the closing trough rate the operation more harshly than equivalent contacts at other times, and they are over-represented in complaint files. The contact you take well at 7:55 builds a customer; the contact you take badly at 7:55 produces a complaint.
The scheduling pattern that fixes it
The fix is not subtle. A small layer of agents — typically 20–30% of the closing-hour headcount — is rostered to a shift that ends 15 or 20 minutes past official closing time. Their explicit job in the final 30 minutes is to take new contacts; the rest of the team is free to wrap up and leave at 8pm sharp. The longer-shift layer becomes the closing skeleton crew, paid through 8:20, and the operation actually delivers service to the customers who arrive in the final interval.
The numbers work out small. A 70-agent floor closing at 8pm needs roughly 6–10 agents on the extended layer. Each works an extra 20 paid minutes; the total daily cost is about 2.5 agent-hours. The recovered service level alone usually justifies it, and the abandonment, next-morning rebound, and CSAT improvements often double the value.
Some operations build the layer as a rotating commitment (each agent takes the late shift roughly one day in seven). Some build it as a self-bid pool (agents who want the slightly later end pick it themselves and get a small premium). Some make it part of the contract for a small dedicated “close team.” All three work; the operation should pick the version that fits the workforce.
The rules that stop the layer drifting into unpaid overtime
The risk with any extended-shift layer is cultural drift: the 20 paid extra minutes quietly become 35 unpaid extra minutes, and the operation reverts to the original problem with a different mask. Three rules keep the layer honest.
First, the layer is staffed by name on the rota and paid as scheduled time, not as overtime. If the operation can’t fund it as standing time, it shouldn’t be doing it on overtime either.
Second, the closing-layer agents have explicit permission to refuse a contact that will obviously over-run their finish time. The contact gets pushed to the answer-machine, the call-back log, or rolls into the next morning. The discipline is to staff the layer; the staffing answer is not “each call must be completed regardless.”
Third, the operation tracks how often the layer over-runs. If it’s consistently over-running by 10+ minutes, the layer is too small. Resize it, don’t squeeze the people who are already on it. See the cost of perfect adherence for why squeezing the close ends up costing more than it saves.
What about IVR closure messages?
The other route some operations take is to close the IVR before the floor closes — routing the final 15 minutes of arriving calls to voicemail or a call-back form. This works for some customer bases and not others. It works when the customer is willing to accept a callback (utilities, banking with established customers, B2B). It fails when the customer expects to speak to someone (claims, sales, urgent service). The honest test is whether the daytime callback completion rate is high (say above 70%); if it is, IVR-managed closing is reasonable. If it isn’t, the customer simply abandons and phones a competitor, and the operation has dressed up its failure as a feature.
What about operations that don’t have a fixed closing time?
24/7 operations or operations that taper down to a single skeleton crew at low-traffic hours don’t have this exact problem — but they have an adjacent one: the handover interval, where shifts change and the floor briefly under-staffs while the new shift logs in and the old shift logs off. The same principle applies. A small overlap layer — 10–15 minutes where both shifts are paid — protects service across the handover and is almost always cheaper than the SL miss it prevents.
Conclusion
The end-of-day SL collapse is one of the most consistent patterns in contact-centre data and one of the least-addressed scheduling problems. Agents going home on time is reasonable; the floor empty 15 minutes before close is not. A small rostered layer that runs 15–20 minutes past official close fixes the trough at a daily cost that almost always pays back in recovered service, lower abandonment, lighter morning queues, and better CSAT in the contacts that would otherwise become complaints. The change is mechanical, the cost is small, and almost no operation does it. The first one in any sector to take it seriously usually finds it’s the cheapest service-level lift on the menu.
Pair this with fixed vs flex breaks, adherence and conformance, and the cost of perfect adherence.