On the 25th of the month, a voice process sits at 78% service level against a contractual 80%. Five days remain. The first instinct on many floors is to call for “all hands” and hope. A better first step takes two minutes with a calculator: work out what service level the remaining days must deliver. Sometimes the answer is achievable. Sometimes it is mathematically impossible, and that changes the conversation completely.
Service level is a ratio of calls, not of days
Service level is usually defined as the percentage of calls answered within a threshold, for example 80% within 20 seconds. Contracts differ on details, especially how abandoned calls are treated, so always use the contract’s exact formula. Whatever the formula, the monthly figure is calculated from all calls in the month, not as an average of daily percentages.
This matters because days have different volumes. Suppose Monday has 10,000 calls at 70% and Sunday has 2,000 calls at 95%. The simple average of the two days is 82.5%. The correct combined figure is (7,000 + 1,900) ÷ 12,000, about 74.2%. Teams that track a running average of daily percentages routinely believe they are closer to target than they are, because the heaviest days, which are usually the worst, are under-weighted.
The recovery formula
Let the month so far have O calls at service level S, and let the forecast for the remaining days be R calls. To finish at target T, the remaining days must achieve:
Required SL = (T × (O + R) − S × O) ÷ R
In words: the total number of calls that need to be within threshold for the whole month, minus those already achieved, divided by the calls still to come.
Take the process above: 100,000 calls so far at 78%, with 20,000 forecast for the last five days. The month needs 0.80 × 120,000 = 96,000 calls within threshold. 78,000 are already in. The remaining 20,000 calls must deliver 18,000 within threshold, which is 90%. Hard but possibly achievable with the right levers.
Now change one input: suppose the month so far is at 76%. The same formula gives (96,000 − 76,000) ÷ 20,000 = 100%. Every single call for five days would need to be answered within threshold. That is not a recovery plan; it is a sign that the month is lost. The useful response shifts to limiting the shortfall (it still matters whether you finish at 77% or 79% if penalties are graded), explaining the cause to the client, and protecting next month.
Two refinements make the calculation more honest. Use the forecast volume for the remaining days, not an average day, since month-end can be heavier in billing or banking processes. And recalculate daily, because each day’s result changes what the rest of the month needs.
Intraday levers, in order of cost
Intraday management is the work, usually led by real-time analysts in the workforce management team, of comparing what is happening interval by interval with the plan and adjusting. When service level drops sharply in an hour, say to 55% against 80%, the first question is why: was volume above forecast, was handle time higher, or were fewer agents available than scheduled? Each points to different levers.
- Get scheduled agents into “available”. Check agents in long after-call work, unscheduled aux codes or extended breaks. This is the cheapest capacity because those people are already paid for.
- Move off-phone activities. Postpone coaching sessions, team meetings and training scheduled in the affected intervals, and reschedule them into quieter ones.
- Use multi-skilled agents. Pull agents from a queue that is over-performing or from back-office work into the struggling queue, if they are trained for it.
- Shift breaks within allowed limits. Stagger breaks away from peak intervals, respecting labour rules and agent welfare.
- Offer voluntary overtime or early log-ins. Effective but costly, and only fair if offered transparently.
- Reduce demand where permitted. Accurate IVR messages about a known outage, callback options instead of holding, or temporarily pausing outbound campaigns that generate inbound calls.
Pressure to cut handle time is notably absent from this list. Rushing calls during a backlog usually produces repeat calls the next day, which makes the remaining days harder.
Where the hours actually help
Service level responds very non-linearly to staffing, which is a consequence of queueing behaviour described by models such as Erlang C. An interval that is two agents short can fall from 80% to well below 60%, while an interval that is already well staffed gains almost nothing from extra agents. So recovery hours should go to the intervals that are forecast to be short, usually the morning peaks and post-lunch peaks, not spread evenly across the day. Real-time analysts typically re-run interval requirements against the latest staffing to see where each extra hour buys the most service level.
Avoiding the same position next month
A month that needs a heroic finish usually had warning signs early. Useful habits include tracking the required-SL figure from the first week, not the last, so that a small daily gap is visible when it is still cheap to close; reviewing forecast accuracy each week, since a persistent under-forecast is the commonest cause; and checking shrinkage against plan, because higher-than-planned leave or absenteeism quietly removes the agents the schedule assumed.
It is also worth being careful with tactics that improve the number without improving the experience, such as changing how short abandons are counted or deflecting calls in ways customers dislike. Clients eventually notice when service level is met while complaints rise, and the trust lost costs more than one missed month.
Written by the RP Strativa Team. Volumes and percentages are illustrative.
Leave a Reply