You can’t meet every service level during a spike in volume, so choose which ones you’ll protect and tell customers about the rest. Teams that decide this in advance look calm. Teams that decide it in the middle of a queue look like they were caught out, even when they worked just as hard.
Three months into the pandemic, almost every support operation has had that experience. Contacts rose, staffing moved (I described my own version of that in reducing a support team by 60%), and the targets you set a year ago stopped describing reality. The useful question is no longer how to hit last year’s numbers. It’s how to run honest service levels during a spike, while conditions are abnormal, and how to tell customers what that means for them.
Which service levels should you protect first?
The ones where a customer’s service is down or their money or safety is at stake. Everything else can slow down without anyone being harmed. That sounds obvious until you’re staring at a queue and the loudest caller sounds the most urgent. Loud and urgent aren’t the same.
You already have the tool for this if you’ve built a severity and customer-tier matrix, like the one in my tiered SLA worked example. Each cell says how fast you respond and how fast you resolve, for a given severity and a given kind of customer. In a normal month those targets are your promise. In an abnormal one they become a ranking: what you protect first, what you let stretch, and what you pause.
Write the reduced version down. A one-page table with the normal target, the temporary target and the trigger for going back is enough. Get your own management to agree it, and share it with the people on the phones. When an agent can say “this is a severity-one, I’ll escalate it now, and this one is a general question, I’ll give you a realistic time,” the queue becomes manageable. When they have to improvise the priorities on every call, quality varies and customers notice.
If you’ve never rerun your staffing at reduced headcount, my 2009 notes on help desk pandemic planning walk through it. Remember that the targets between your teams are connected. If Tier 2 is working remotely with a longer response time, Tier 1’s resolve target has to move with it. A customer’s clock keeps running across every handoff, so a stretched escalation path stretches the whole experience. Adjust the whole chain together, not one link.
What do you tell customers when you’re slower?
The truth, early, and in the places they’ll actually see it. Put a short, plain message at the top of your phone menu, in your email auto-response, on your support page and in the first line of any ticket acknowledgement. Say what’s different, how long it might take, and what you’d like them to do if their issue is urgent.
Avoid the temptation to cushion it with vague reassurance. “We’re experiencing higher than normal volumes” tells a customer nothing they can use. “Email replies are currently taking up to two working days. If your service is down, call and say so at the first prompt” tells them what to do. Replace the guess with a number whenever you can, and update it when it changes.
Look at how some large providers handled the policy side. On March 13, the FCC chairman asked broadband and telephone providers to sign a pledge, and The Next Web reported that AT&T, Verizon and others agreed for the next 60 days not to cut off customers who couldn’t pay, to waive late fees and to open Wi-Fi hotspots. Whatever you think of the policy, notice how it works as customer communication. It answers the question customers were already asking, before they had to call and ask it. That’s the best outcome a support team can get: a contact that never happens.
You can apply the same logic at a smaller scale. List the ten questions your agents are being asked most this month. For each, ask whether an earlier message from you could have answered it. Billing date changes, delivery delays, how to reach someone, whether a service is affected. Every question you answer in advance is a contact you don’t have to handle.
How do you stay on top of a queue that keeps changing?
Watch fewer numbers, more often, and pick them the way I described in choosing the right KPI from the customer’s side. Daily, I’d look at the share of contacts answered within target, by channel, the age of the oldest open escalation, and the number of staff available against the number scheduled. The third one is the most useful in a disrupted month, because it tells you tomorrow’s problem today. That’s the idea behind leading indicators: a number that moves before the customer feels it.
Don’t change the definition of a measure to make a hard month look better. A response-time figure that means something different in May than it did in February is noise, and you’ll need those numbers later to show what this period cost and what you did about it. Report the abnormal numbers against the normal target and the temporary one. Both tell a true story.
Split your contacts by reason, if you aren’t already. A spike driven by one cause, such as a policy change or a delivery problem, behaves differently from a general rise across everything. The first can often be fixed with a message or a knowledge base article. The second needs people. If you can’t tell which you have, you’ll staff for the wrong problem.
Can self-service and chat take the pressure off?
Some of it, if you point them at the right work. A chatbot or a well-organised help page is good at repetitive, factual questions with a known answer: opening hours, order status, how to reset a password. It’s poor at anything emotional or ambiguous. I covered the realistic upside in how chatbots improve customer experience, and the principle holds in a crisis: automate the question, not the relationship.
Don’t launch something new in the middle of a surge and expect it to work on day one. A rushed bot that sends customers in circles produces angrier contacts than the ones it was meant to prevent. If you already have self-service, improve the pages people actually visit. If you don’t, start with the ten most common questions and write the answers in plain language. A good page costs a few hours and can remove hundreds of contacts.
Make it easy to reach a person, too. The fastest way to turn a minor irritation into a complaint is to put a wall between a customer and a human when they have a real problem. If the issue is serious enough for someone to pick up the phone, the self-service route shouldn’t stand in the way.
What should you decide now, before the next spike?
Three things. First, the trigger: what number tells you it’s time to move to the reduced service levels, and who has the authority to call it. A threshold such as a limit on the oldest waiting contact turns an argument into a reading from a gauge. Second, the owner for customer messages. Someone has to update the phone greeting and the support page, and it shouldn’t take a committee. Third, the way back: when do you return to normal targets, and how will customers hear about it?
Then review what happened, while it’s fresh. Which targets did we protect, and did that choice hold up? What did customers ask for that we hadn’t anticipated? Which agents carried too much of the load? Put the answers into the plan. A review that produces a list nobody owns is theatre, and a review that gives someone a job and a date is a plan.
For comparison against what other support centres are doing, the HDI community is a useful reference. And if you want the wider picture of how a contact centre fits into a crisis, my earlier piece on whether a call centre can help in a pandemic covers the customer side.
Build a small habit around updates. Whoever owns customer messages should review the wording every working day while conditions are unstable, even if the answer is that nothing has changed. A page that says “updated today” earns more trust than one that was right two weeks ago and hasn’t moved since. It also gives your agents one source of truth to point customers to, so the story told on the phone matches the story on the website.
Finally, keep a record of the exceptions you grant. A refund outside policy, a deadline extended, a fee waived. Individually they feel reasonable. Collectively they’re a policy, and you need to know what it is before somebody else reads it back to you. Review the list weekly and decide which ones to formalise and which to end.
Customers can forgive slow. What they don’t forgive is finding out you knew and didn’t say.
Hutch Morzaria is a Director-level CX and Support Leadership professional with 19 years of experience building global support organizations across SaaS, Fintech, and enterprise technology. He has hired dozens of support and CX leaders across his career and holds ITIL Expert certification across V3 and V4.
