Field mapping

What to put in each field well, and the one mistake that produces confident, wrong staffing numbers.

The five fields

Field well	Required	Put in it
Interval	Yes	The start of each staffing interval. A date/time column is best. Any sortable category works, provided it separates one day from the next — see below.
Forecast contacts	Yes	Contacts offered in the interval. A plain `SUM`.
Average handle time (seconds)	Yes	A weighted average at the interval grain. See below.
Scheduled agents	No	Agents rostered for the interval. Staffing level, not a sum of row values.

The interval field has to separate the days

Intervals are keyed on the value you map here. A bare time-of-day with the date discarded keys every Monday 09:00 to the same interval as every Tuesday 09:00, so a week folds into a single day and every folded interval carries several days’ worth of contacts. Nothing errors; the chart just looks like one day.

A date/time column avoids this by construction. If your model splits date and time into separate columns, make sure the field you map here is still the one that distinguishes them.

To check: hover any interval and read Queues in the tooltip. It should equal the number of queues you run. Three queues over five days reporting 15 is the signature of folded days. | Queue or group | No | Splits the calculation per queue. Used for filtering, selection and tooltips. |

Without the first three the visual tells you which one is missing rather than drawing anything.

Average handle time is the field to get right

Three of the four numbers are simple sums. This one is not, and it is the most likely source of a wrong answer that still looks plausible.

Dropping a column called AHT into the field well makes Power BI average it — and an average of averages is not the average. If one interval handled 500 contacts at 200 seconds and the next handled 5 contacts at 900 seconds, the true weighted AHT is 207 seconds. Averaging the two averages gives 550. Offered load comes out nearly three times too high, and the visual will confidently recommend staffing for it.

The visual cannot detect this. Nothing in the data view distinguishes a correctly weighted measure from a badly weighted one.

Store handle seconds, not handle time, and divide:

Weighted AHT =
DIVIDE(
    SUM( Interactions[HandleSeconds] ),
    SUM( Interactions[Contacts] )
)

The sample dataset in sample-data/ stores HandleSeconds for exactly this reason — it makes the correct measure the natural one to write.

If your source only gives you an average per interval, reconstruct the total first:

Weighted AHT =
DIVIDE(
    SUMX( Interactions, Interactions[AvgHandleTime] * Interactions[Contacts] ),
    SUM( Interactions[Contacts] )
)

Scheduled agents is a level, not a total

Required and scheduled staffing are both headcounts at a moment, not quantities that accumulate over an interval. If your roster table has one row per agent per interval, COUNTROWS is right. If it already stores a headcount per interval, SUM over that one row is right. What is wrong is any measure that grows when the interval gets longer.

A quick check: switch the interval length setting from 30 to 60 minutes. Your scheduled staffing should stay roughly the same. If it doubles, the measure is summing rather than levelling.

Interval length is a setting, not an inference

The Interval length setting in the Format pane has to match the grain of your Interval field. It is not detected from the data.

That is deliberate. Detection would have to guess from the spacing between values, and it would guess wrong in exactly the cases that matter — overnight gaps, missing intervals after filtering, a queue that only opens at 08:00. A wrong guess silently changes every staffing number on the chart, so the setting is explicit and the default is 30 minutes.

Queues

Mapping a queue field calculates each queue separately and adds the resulting agent counts together on the chart.

This is the no-pooling assumption: each queue is staffed by its own people. It is the conservative reading, and the one that matches how most contact centres actually schedule. If your agents genuinely handle all queues from one pool, do not map the queue field — pooling more traffic into one queue needs fewer total agents, and calculating them separately will overstate what you need.

Service level and occupancy are not aggregated across queues, because there is no honest single number for “the service level across three queues”. A multi-queue interval shows the per-queue breakdown in its tooltip instead.

Reading the chart

Mark	Means
Solid line	Required scheduled agents — the answer, after shrinkage
Dashed line	Scheduled agents — what you have rostered
Shaded band	The gap between them. Red and hatched when short, green when surplus
Faint dotted line	Required productive agents, before shrinkage. Off by default

The hatch on an understaffed band is deliberate, not decoration. Colour-vision deficiency, greyscale printing and Windows high-contrast mode each remove hue as a channel, so shortfall is marked by texture as well as colour and survives all three.

Letting report readers adjust the parameters

The Format pane is author-only. Someone reading a published report cannot open it, so out of the box they cannot ask “what if shrinkage were 35%?” — the exact question this visual exists to answer.

Four optional field wells fix that. Bind a Power BI what-if parameter to one and the reader gets a slicer.

Field well	Overrides
Target service level %	The share half of the SLA
Answer within (seconds)	The seconds half of the SLA. Seconds, not a percentage.
Shrinkage %	The Format pane’s shrinkage
Maximum occupancy %	The occupancy cap

Whatever is not mapped keeps its Format-pane value, so you can expose just shrinkage and leave the rest fixed.

Setting one up

Modeling → New parameter → Numeric range.
Name it Shrinkage, minimum 0, maximum 95, increment 5, default 30.
Leave Add slicer to this page ticked.
Drag the generated Shrinkage Value measure into the Shrinkage % well.

The reader moves the slicer; required staffing recalculates.

Units: whole-number percentages

These wells take 80, not 0.8 — the same convention as the Format pane.

A what-if parameter built over 0 to 1 is the easy mistake, and it would otherwise compute staffing for a 0.8% service-level target quite happily. So a percentage field carrying a value between 0 and 1 is rejected with a message naming the fix, rather than calculated with. Zero is allowed: 0% shrinkage is a real setting.

The chart says which parameters it used

Under the summary metrics is a line reading something like:

80% in 20s · 30% shrinkage · 85% max occupancy · 30-min intervals

It reflects whatever actually drove the calculation, so a reader moving a what-if slicer sees the assumptions move with it. A staffing number is not a fact on its own — it is a fact given those parameters — and a screenshot of the chart would otherwise be unattributable.

It trims to the two parameters that move staffing most on a narrow visual, and carries the full line as a tooltip. Turn it off under Chart → Show assumptions line. Screen readers get it too, as part of the chart’s label.

Interval length is deliberately not overridable. It has to match the grain of your data, which is a property of the model rather than a question a reader should be able to change.

What you can change

Every card in the Format pane:

Card	Controls
Staffing parameters	Service level (SLA): the share of contacts to answer and the seconds to answer within. Staffing and intervals: shrinkage, maximum occupancy, interval length
Chart	Scheduled line, gap shading, required-productive line
X axis	Label format, rotation, and text formatting
Y axis	Show/hide, text formatting, number format
Axis titles	Show/hide, text formatting
Legend	Show/hide, text formatting
Lines	Style (solid / dashed / dotted) and width for required and scheduled
Data labels	Show/hide, which series, text formatting, number format
Summary metrics	Show/hide, text formatting, number format
Assumptions line	Show/hide, text formatting
Colours	Required, scheduled, understaffed, surplus

Text formatting

Every card that draws words carries the same set: font family, size, bold, italic, underline, alignment and colour.

Colour has a Match report theme switch, on by default. Leave it on and the text follows the report’s own colour, so the visual stays readable on a dark theme. Turn it off and the colour picker takes over — an explicit choice rather than something that silently breaks when someone applies a dark theme.

Two colours ignore the setting on purpose, because they carry information rather than decoration: an understaffed metric stays red, and so does the evaluation-mode truncation notice.

Alignment means what it can for each element. In the summary tiles and the assumptions line it is alignment within the box. For axis ticks, data labels and the legend — text with no box — it shifts the label relative to the point it is pinned to. Y-axis ticks stay right-anchored regardless, because any other anchor walks them into the plot.

Number format

Summary metrics, data labels and the Y axis each take display units and decimal places.

Display units default to Auto, which only abbreviates once the digits stop being individually meaningful — six figures. Turning “1,200 agents” into “1.2K” loses precision a staffing plan needs, so that is not something the visual does uninvited. Set Thousands or Millions explicitly if you want it sooner.

Decimal places default to -1, meaning “let the value decide”: whole agent counts print whole, percentages get one decimal, and an abbreviated value keeps a decimal so “1.2K” does not collapse to “1K”.

The summary metrics

Metric	Means
Forecast SLA	The service level the scheduled staffing is expected to deliver, across everything visible. Red when it falls below your target.
Understaffed	Intervals where scheduled staffing is below requirement
Largest deficit	The worst single-interval shortfall
Peak required	The highest scheduled requirement in the range
Intervals on target	The share of intervals that individually clear the target

Forecast SLA and Intervals on target are not the same number, and confusing them is easy. One says what service level the day will deliver; the other says how many intervals clear the bar. A day can have two thirds of its intervals on target and still deliver well under target overall, if the misses land where the volume is.

Forecast SLA is weighted by contacts, because service level is a ratio of contacts answered in time to contacts offered. Averaging the per-interval percentages instead would let a quiet overnight interval at 100% cancel out a peak interval at 40% — on the sample data that inflates the figure from 62.4% to 75.3%, which would be a comfortable and completely wrong number to plan against.

Zero-volume intervals carry no weight, so an overnight stretch cannot drag the forecast either way. An overloaded interval counts as 0%, not as missing — dropping it would flatter the result.

Data labels

Off by default, because an interval chart is dense and the shape usually matters more than the individual numbers. Turn them on under Data labels and choose required, scheduled or both.

Two things happen automatically:

They thin out. 96 numbers across 600 pixels is a smear, not information, so labels drop to every second or third interval the same way axis labels do. A larger text size means fewer labels.
They avoid each other. With both series labelled, each number sits on the far side of its own line from the other one — so they do not stack up where the lines run close together, which is exactly when someone turns labels on.

Each label carries a halo in the background colour so digits stay legible over a gridline or a shaded gap band.

Axis label format

X axis → Label format offers presets — time only, date only, weekday and time, and so on — or Custom for your own pattern.

Token	Gives	Token	Gives
`yyyy` `yy`	2026, 26	`HH` `H`	14, 14
`MMMM` `MMM`	March, Mar	`hh` `h`	02, 2
`MM` `M`	03, 3	`mm` `m`	30, 30
`dddd` `ddd`	Monday, Mon	`ss` `s`	00, 0
`dd` `d`	02, 2	`tt`	PM

Put literal text in single quotes. HH'h'mm gives 14h30; without the quotes the h is read as a token and you get 14230. '' is an apostrophe, so HH:mm 'o''clock' gives 14:30 o'clock. Month and weekday names follow the report’s locale.

Non-date intervals are left alone — a category axis is already whatever you made it.

Axis label rotation

X axis → Label rotation: Automatic, Horizontal, Angled (45°) or Vertical (90°). Rotated labels take far less horizontal room, so many more of them survive before the axis starts dropping every second or third one. The chart reserves the extra height automatically.

Automatic stays horizontal until the labels stop fitting, then angles them. It never goes vertical on its own — 90° costs a lot of chart height, and that is a trade worth opting into rather than having sprung on you by a resize.

The y-axis is titled Agents; the x-axis carries the display name of whichever column you mapped to Interval, so a mismapped field shows up at a glance. Both can be switched off under Chart → Show axis titles, and both drop out automatically below 300 x 140.

The legend appears above the chart when the visual is at least 380px wide and 150px tall. Below that the chart is a sparkline and there is little left to label.

Getting a sanity check

The quickest test that a mapping is right: pick your busiest interval and check it by hand.

Read the forecast contacts and weighted AHT from the tooltip.
Offered load is contacts × AHT ÷ interval seconds. For 300 contacts at 280 seconds in a 30-minute interval: 300 × 280 ÷ 1800 = 46.7 Erlangs.
Required productive agents will be somewhere just above that — typically 10 to 20 percent higher at an 80/20 target.
Required scheduled will be that divided by 1 − shrinkage.

If required staffing comes out at three times offered load, or below it, the AHT measure is the first thing to check.