Reading the intraday — service is the symptom, not the problem

Real-time management · ~7 minute read

Service level is the symptom; volume, AHT and availability are the drivers that produce it. The amateur reacts to the service tile. The professional asks which driver moved — because the same 78% from three different causes needs three completely different fixes.

Service is the outcome, not the diagnosis

When service slips, the instinct on most real-time desks is to treat the service number itself — pull people back, sound the alarm, throw bodies. But service is an outcome, driven by the interaction of three things: volume (how much work is arriving), AHT (how long each piece takes), and availability (how many agents are actually handling it). A dip from a volume spike, a dip from AHT creep, and a dip from collapsed availability look identical on the service tile.

They are not identical problems. Volume above forecast needs more capacity or deferral. AHT above forecast means each contact is taking longer — a system slowdown, a complex issue type spiking, new starters — and crucially, more bodies will not fix rising AHT as efficiently as fixing its cause. Availability below plan means the people are not there: shrinkage running hot, adherence slipping, sickness. Misdiagnose, and you pull the wrong lever — people thrown at an AHT problem that a CRM fix would have solved.

Read all four together, against plan

No dimension means anything in the absolute. Volume is read against forecast, AHT against forecast, availability against requirement, service against target — and all four together, because the diagnosis lives in the comparison. A desk that watches the service tile alone is a tile-watcher; a desk that decomposes every service movement into its driver is doing the core analytical work of the role: symptom to driver to lever.

The drivers also interact, sometimes viciously. High volume drives up occupancy, which under pressure drives up AHT and erodes adherence — people with no clean moment to come back from breaks. A small initial divergence cascades across all three drivers within the hour. Reading the interaction, and acting on the trajectory rather than the snapshot, is what stops a one-driver problem becoming a three-driver crisis.

Quantify the variance — three views

Noticing the gap is not enough; the skill is to quantify it, because the number sizes the response. Track three views live: the interval variance (this half-hour, actual versus forecast, per driver), the cumulative variance (day-so-far against day-so-far — the running story), and the trend (widening, stable or closing). A desk that watches only the interval reacts to noise; the cumulative and the trend reveal the real divergence.

The discipline shows in the language. “Volume’s up a bit” might be +3% (noise — hold) or +15% (a real surge needing capacity); “busy today” supports no decision at all. “Voice volume running +12% on the day, +18% this interval and widening” supports a precise one: how much extra capacity, for how long, starting when. Vague awareness is the tell of a desk that will either over-react or under-react — at random.

Blip versus divergence

The hardest live judgement: a single interval above forecast might be random bounce, a transient already passing, or the first interval of a surge that runs all day. Reacting to a blip wastes effort and credibility; missing a divergence breaches service. The tells of a real divergence: it shows in the cumulative, it persists across several intervals, it usually has an identifiable cause (a campaign send, an outage, weather, a bill run), and the trend holds or widens rather than snapping back.

So the rule is: hold through the blip, act on the divergence — and always seek the cause. A divergence whose cause you have found can be projected forward and acted on proportionately; a divergence without a cause can only be chased. The desk that asks “what changed?” before reaching for a lever pulls fewer levers, and pulls the right ones.

Close the loop to planning

Forecast-versus-actual is not just for today. Every divergence the desk observes, with its cause attached, is intelligence the planning team needs: a forecast that ran 15% light because of an un-modelled campaign should teach the next forecast. The desk that logs variance and cause, and feeds them back, makes the whole operation’s forecasting better over time.

The desk that does not is the common failure: the same bill-run surprises the operation every month, the real-time team fights the same fire on the same date, and planning never finds out why. Today’s variance is two things at once — a gap to manage now, and a lesson for the plan later. Most operations capture the first and waste the second.

Symptom to driver to lever When service moves, check ▸ Volume vs forecast ▸ AHT vs forecast ▸ Availability vs plan ▸ Interval variance (this half-hour) ▸ Cumulative variance (day so far) ▸ Trend — widening or closing? ▸ An identifiable cause Then act ▸ Volume off → capacity or defer ▸ AHT off → fix the cause, not bodies ▸ Availability off → recover it ▸ Blip → hold ▸ Divergence → act, proportionately ▸ Watch the cascade between drivers ▸ Log variance + cause for planning Same service dip · different cause · different fix

The closing principle

Service is the symptom; volume, AHT and availability are the drivers. Read all four together, against plan and in trajectory; quantify the variance; hold the blip and act on the divergence — and feed every caused variance back to planning. Symptom to driver to lever is the habit everything else depends on.

See also