A schedule that matches this morning’s forecast can still fail by lunchtime, because the forecast was written before the news landed. Intraday staffing is the discipline of re-reading the day while it’s happening and moving people before the queue tells you you’re late. The clearest public example I know of is what a British bank’s phone lines did in September 2007. I don’t think most contact centres are ready for a day like that. Their forecasts are tidy, their rosters are tidy, and neither one has a plan for the hour when the news arrives.
If you read my post on mind the gap in call centre scheduling, you’ll remember the gap row: required agents minus agents on shift, hour by hour. That row is only as good as the required figure above it. Change the figure and every number underneath is wrong, even though nobody touched the roster. This post is about what happens when something outside your control rewrites that figure halfway through the day. It’s also about the boring preparation that decides whether you cope.
What did Northern Rock’s phones actually show?
In September 2007 the British bank Northern Rock had a run on its deposits. Queues formed outside branches for several days, the phone lines jammed, and the website struggled. On the Monday evening the Chancellor announced that the government would guarantee existing deposits, and the House of Commons Treasury Committee’s report, The run on the Rock, walks through how the support operation, the run and the guarantee followed one another. One statement changed why people were calling. That’s the bit a scheduler should notice.
By Tuesday 18 September the bank could say so with numbers. According to a CNBC report of its statement, its call centres had taken 3,472 calls by 10:30 BST that morning. That was less than 10% of the volume at the same time the previous day. And 80% of those calls were customers asking to reinvest their money. Read those figures the way a scheduler would. Volume fell to under a tenth of the Monday level overnight, and the reason for calling flipped from “get my money out” to “put it back.”
Same phone number, same customers, completely different call. The people rostered for Monday were wrong for Tuesday in both headcount and skill. I don’t have Northern Rock’s rosters and I’m not going to guess how they staffed it. What the public numbers show is enough for the lesson. A day-ahead schedule is a bet on how the day will go, and some days the bet misses by an order of magnitude in one direction and then the other.
Why does a good schedule fail intraday?
An Erlang C schedule, however carefully built, is a set of promises about the hours ahead, and it quietly assumes three things. It assumes the arrival pattern is the usual one, because you forecast from last month’s Tuesdays. It assumes the call mix is the usual one, because your Tier 1 and Tier 2 split came from a normal week. And it assumes handle time is the usual one. A government announcement or a press story can break all three before the first break. An anxious caller talks longer, so handle time rises exactly when arrivals do, and the required-agents figure rises twice.
That’s different from a spike you can see coming. When I wrote about staffing for a product launch, at least the date was on the calendar and you could argue about how big the day would be. A bank run, a frozen website or a headline gives you no date. I used to think the fix was a better forecast. It isn’t. Nobody forecasts the unforecastable, and the time spent trying would be better spent planning for being wrong quickly.
This matters right now because the news has been moving customers’ behaviour for weeks. On 3 October President Bush signed the Emergency Economic Stabilization Act, which temporarily raised the standard FDIC deposit insurance limit from $100,000 to $250,000 per depositor. Two days later, on the evening of 5 October, Chancellor Merkel told German savers their deposits were safe. Each statement was meant to calm people, and each one gave financial services contact centres a reason to expect an unusual day. I have no call volumes from any bank for those weeks, and I’d be inventing them if I gave you a number. An announcement made on a Sunday evening is a staffing event on Monday morning.
How do you re-forecast during the day?
You don’t need a new model. You need a habit and a trigger. The habit is comparing actual arrivals to forecast every interval, fifteen or thirty minutes, so the gap row becomes a live number instead of a report somebody reads after the shift. The trigger is a number you write down before the day starts. For example, if arrivals run 25% above forecast for two intervals in a row, the duty manager re-forecasts the remaining hours. At 11am, with the board going red, nobody agrees what “busy” means. On a quiet Thursday afternoon they will.
Here’s a small worked example. Say your forecast for 10am to 11am is 60 calls and by 10:30 you’ve had 55. You’re on pace for roughly 110, nearly double. Waiting for the hour to end tells you nothing you can use, because by then the hour’s gone. Re-forecasting at 10:30 gives you time to pull people off breaks, cancel training and call the on-call pool while there’s still a morning left to protect. Thirty minutes of warning is the whole difference between a bad hour and a bad day.
Count the calls and read them too. Sample ten. If most are about the same thing, you’ve found the cause, and a message on the website or the IVR can take more load off the queue than extra seats can. Then make sure someone can act on what you’ve found. A person has to be able to say “move email people to phones” without a committee, and if that takes three approvals the spike has passed before the decision lands. An intraday plan can fail on exactly that point, with the right data on the screen and nobody in the room allowed to use it.
Where does the extra capacity come from?
Every contact centre has the same short list, and the order matters because speed and cost run in opposite directions. Moving people between channels is fastest and free, but it’s limited to the channels you already staff. Email and back-office work can wait an hour, and a caller on hold can’t. Cancelling planned activity comes next: training, coaching and team meetings go first, so decide in advance which ones are movable rather than negotiating on the day. Then the flexible pool of part-timers, split shifts and agents who’ve agreed to be called, which only works if you asked them before the day and wrote the answer down. Notice that those first three cost almost nothing. They’re also the ones teams skip.
After that it gets expensive. Overtime is quick to arrange and costly in fatigue, so use it to cover a peak, not to run a week. Cross-trained staff from other teams are the cheapest option over a year and the slowest to build, which is why that work has to start months ahead. Outsourced overflow is useful for a simple, repeatable call and poor for the anxious, complicated ones. Northern Rock’s Tuesday callers wanted to reinvest money, and that’s the sort of call that can be scripted. A call about why your savings feel unsafe isn’t.
Which brings up skill mix, the part that’s easiest to forget. On a normal day your Tier 1 to Tier 2 ratio might sit around six to two, as in the sample schedule in my earlier post. If the new calls are simple and repetitive, such as “reinvest my money” or “confirm my balance,” you want more Tier 1 and a tight script. If they’re complicated, such as “my payment didn’t go through and I’m panicking,” you want Tier 2 and a shorter route to them. Check the response expectations in your tiered SLA matrix before the day, so you know which customers get priority when you can’t answer everyone at once. Watch the call mix next to the call count, two charts, not one. A gap row that reads zero in total can still be four short in the skill you need.
What would I do on Monday?
Everything above scales down, and a small team needs it more, not less. Say you run a team of 12 and normally have 4 people on the phones from 9 to 11. A story breaks and arrivals double by 9:30. You don’t have a flexible pool of dozens. You have the 8 people doing other work. Which two can move without breaking something? Who’s on lunch? Can the person on the escalation desk take first-line calls for an hour? Those are three questions you can answer today in five minutes, or on the day in thirty.
So pick the next external event that could change why your customers call: a price change, a product recall, a regulator’s announcement, an annual statement run. For each one, write down three things on one page. The trigger that tells you the day has gone off-script, the named person who can change the roster, and the first places you’d find extra capacity, in order, with names. A list of six names and a rule, such as “ring in this order, ask for an hour, and thank them,” is more useful than a policy document. Under pressure people choose whichever option is nearest, and nearest is rarely best.
Then run a ten-minute drill. Pretend arrivals are double the forecast at 10:30 and ask who does what. If the answer takes longer than a minute, the plan isn’t ready, and it’s much cheaper to find that out on a quiet morning than on the morning it matters. A schedule tells you how many people you’d like to have. Intraday staffing is what you do when you’ve got the wrong number.
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.
