Proposing a Business Definition

Proposing a business definition is the step before an entry in The Definitions Register. When a term is disputed, used two ways, or about to land on a scorecard, someone writes a short definition proposal. It states the problem, shows the competing formulas side by side, sets out the options, recommends one, works an example, says what will change in reported numbers, names what the data owners must build, and lists the open questions. The proposal is the "proposal form" that Definitions and Source-of-Truth Change Control (generic process) (GP-606) names as the supporting artifact of its first step. The register records what was agreed; the proposal records why, and which alternatives were rejected.

The pattern is borrowed from fields that settle shared standards in writing before adopting them:

• Internet standards move through numbered, versioned documents that are reviewed in the open before adoption.
• Open-source language communities ask enhancement proposals to state their motivation and specification, and to record the alternatives considered.
• Software architecture uses a lighter form, the decision record: one short document per decision, giving title, context, decision, status and consequences.
• Metadata-registry standards pair each agreed concept with its representation under one identifier.

A workforce definition proposal combines all four: open review, a record of rejected alternatives, the brevity of a decision record, and the registry's one-term-one-identifier rule.

When a proposal is needed

A term needs a proposal rather than a quick register edit when any of these hold:

• One name, several formulas. The same metric name is computed differently by different platforms, teams or vendors.
• Two terms used interchangeably. Occupancy and utilization are the standard case. The register's rule for them is to define both together or not at all.
• A scorecard or contract is about to depend on it. Once a number is in a scorecard or a vendor schedule, changing its definition is a series break with consequences.
• The instrument cannot produce it. If a platform cannot emit the components, the proposal has to say so. A segment that cannot compute a metric is not assessable on it, which is a different finding from scoring low.

The sections of a proposal

Section	Holds	Test it must pass
Problem	The divergence today, with each competing formula quoted exactly and its source system named	A reader can see the disagreement is real arithmetic, not wording
Scope	Layer (planning, operational, business or contractual), channels, units, exclusions	The term answers one question for one audience
Ledger	The buckets every term reads from, usually tiers of the Agent Time Accounting Model	Every term is a named bucket or an adopted definition
Proposed definition	A plain sentence, then the formula by channel; unit, grain, aggregation rule, pairing	Two analysts given the same data get the same number
Options considered	Each candidate formula, what it measures, and why it is recommended or rejected	The rejected options stay on record
Worked example	Graded synthetic numbers for every option, plus the number as reported today	The arithmetic is recomputed independently
Reporting impact	Series breaks, segments that become not assessable, pages and reports that change, who is briefed first, the baseline period	Nobody learns about the change from the scorecard
Instrument asks	What the data owners must build: state mapping, source of truth, grain, lineage	Each ask has an owner and a reason
Open questions	Numbered; each one closed, or parked with an owner and a date	None left unowned at adoption
Register entry and sign-off	The ten register fields completed, signatories by seat, plus a target policy note	The entry can be copied into the register

Target policy is a note carried alongside the entry, not one of the ten fields. It records whether the metric is an input, an output or a key performance indicator, and whether a target may be set. The target value belongs to the capacity cycle; as Capacity Standards Workshop for Time Accounting puts it, no number changes by agreeing a definition. The policy is still worth recording. Occupancy shows why: under queueing theory, the occupancy a pool can sustain at a given service level depends on the pool's size, so a single fixed occupancy target builds a service failure into the plan for some pools and slack into others. A definition that arrives with a target attached also tends to be managed to the number before anyone knows what normal looks like. That is Goodhart's observation, in its best-known restatement: a measure that becomes a target ceases to be a good measure.

Ownership is split the way the register splits it. The operational function owns the definition and its pairing; the analytics function owns the instrument, the source of truth and the lineage. A proposal names both owners from the start, because a definition the instrument cannot compute is not yet a definition.

Worked example: occupancy and utilization

The example is synthetic. Its inputs are graded [E] under Human Gates and Number Grades, and every ratio computed from them is [C] on [E] inputs.

Problem

Discovery in a multi-platform operation typically finds three formulas under the one name occupancy:

• handling time ÷ (staffed time − auxiliary time)
• handling time ÷ (handling time + ready time × mean concurrency)
• handling time ÷ (handling time + available time)

It also finds a utilization formula that counts idle time in its numerator and keeps auxiliary time in its denominator. The two measures can move in opposite directions: when auxiliary time is converted into available time, occupancy falls while that utilization rises. A leader who hears one while thinking of the other then misreads the staffing position.

Published reference pages carry the same spread:

• Agent Utilization vs Occupancy Definitions and Tradeoffs counts handling, idle, scheduled breaks and meetings as utilized time.
• Occupancy and the ledger table on Agent Time Accounting Model use handle time over paid time.

Each form is legitimate. The organization has to choose one and record why.

Scope and proposed definition

	Occupancy	Utilization
Layer	Workforce planning and real time	Business and scorecard
Question	Of the time an agent is open for customer work, how much is spent handling it?	Of the paid time, how much is deployed to customer work, in any channel?
Formula	Live channels: (talk + hold + after-call work) ÷ (talk + hold + after-call work + available) Email: work ÷ (work + idle in queue) Chat: focus time ÷ chat open time, with concurrency reported separately, as Chat Concurrency and Handle Time explains	Open time (every state in which the agent is available for, or doing, customer work) ÷ paid time
Aggregation	Ratio of sums across intervals and agents	Ratio of sums across agents and weeks
Pairing	Service level and pool size; actual against planned	Paid-basis shrinkage (the two sum to 100%) and a companion team time percentage
Target policy	An output, never a fixed target; see The Occupancy Trap	No target until a clean baseline exists

Email occupancy sits near or above 95% whenever items are waiting, because working an item counts as occupied. It is therefore treated as a planning parameter, an assumption checked against measured idle time, rather than as a performance measure.

Options considered (utilization)

Option	Formula	What it measures	Decision
A	Open ÷ paid	Paid time deployed to customer work	Recommended. Closes to 100% with paid-basis Shrinkage, and is unaffected by mis-coding between channels
B	(Open + paid breaks + training + coaching + meetings) ÷ paid	Planned non-customer time counted as utilized, the form on the tradeoffs page	Rejected. Breaks and development disappear into a busy number. Readiness is left out, as on that page
C	In-work ÷ paid	Handle density	Rejected as the headline, because it equals A × blended occupancy and inherits channel-mix effects. Kept as a diagnostic

Training, coaching, meetings and readiness time are published as the companion team time percentage, so time spent developing people stays visible. The denominator is paid time, because it ties to cost and full-time equivalents. A scheduled-time view is published alongside it to show seasonality, because leave is seasonal and a paid-time denominator dips with it.

Worked example

One agent, one week: 40.0 paid hours [E].

• 3.0 hours are off schedule.
• 7.5 hours are in schedule but not open: paid breaks 2.5, training 1.5, coaching 1.0, meetings 1.0, readiness 0.75, unplanned auxiliary time 0.75.
• That leaves 29.5 open hours:
  • voice: talk and hold 13.5, after-call work 2.5, available 4.0
  • email: work 9.0, idle in queue 0.5

Measure	Calculation	Result [C]
Voice occupancy	16.0 ÷ 20.0	80.0%
Email occupancy	9.0 ÷ 9.5	94.7%
Blended occupancy (diagnostic)	25.0 ÷ 29.5	84.7%
Utilization, option A	29.5 ÷ 40.0	73.75%
Paid-basis shrinkage	10.5 ÷ 40.0	26.25%
Team time	4.25 ÷ 40.0	10.6%
Option B	35.5 ÷ 40.0	88.75%
Option C, handle density	25.0 ÷ 40.0	62.5% (= 73.75% × 84.7%)

As reported today. Suppose 3 of the 9 email hours were worked while the voice state showed available, the contamination Production Accounting for Blended Agents describes:

• voice occupancy reports 16.0 ÷ 23.0 = 69.6% against a true 80.0%
• email shows 6.0 work and 0.5 idle, so 92.3%
• utilization stays at 73.75%, because the mis-coding moves minutes between channels but never out of open time

The ledger page works a larger version of the same effect. The difference in robustness is the main argument for option A as the business headline.

Reporting impact, asks and open questions

Reporting impact

• Voice occupancy in blended teams will rise once email has its own state code, with no change in workload. Leaders are briefed before the first report, and the series break is marked.
• Adopting option A changes the utilization formula used in the ledger table on Agent Time Accounting Model and the comparison table on Occupancy. Both pages are listed for amendment through GP-606.
• No utilization target is set until eight weeks of clean baseline exist, outside the holiday season.

Instrument asks

#	Ask	Owner	Reason
1	A separate email state, so email work is never logged as voice available	Analytics lead	Removes the contamination above
2	Blended-cohort rows added to the precedence table on Agent State Taxonomy and Activity Codes, for minutes an agent spends in two channel states at once	Analytics lead with the head of WFM	Each minute counted once, by the table's rule rather than a private one

Open questions

#	Question	Status	Owner	Date
1	Paid or scheduled time as the headline denominator?	Closed: paid, with scheduled published alongside	Head of WFM	11 November
2	Can every segment capture email idle-in-queue time?	Parked: proxy accepted by exception until capture is built	Analytics lead	4 January

Register entry and sign-off

Field	Utilization (new entry, UTL-01)
ID	UTL-01
Definition	Share of paid time deployed to customer work in any channel: open time ÷ paid time
Instrument	State-to-tier map v1 per platform and vendor (system, field and version recorded per row); payroll hours, or billed hours for vendors, as the denominator
Resolution	State minutes at interval grain; a team's weekly figure resolves to about half a point [C]; reported weekly as a ratio of sums
Pairing	Paid-basis shrinkage; team time; scheduled-basis view
Owner	Head of WFM
Signatories	Head of WFM, analytics lead; executive sponsor as well, because the denominator moves a scorecard
Version and effective date	v1, adopted 11 November, effective 4 January
Carry rule	Graded [A] in a new platform, vendor or state set until its state-to-tier map is re-measured; history before v1 is not restated
Change history	v0 proposed 20 October; v1 adopted 11 November

The occupancy entry, OCC-01, is amended in the same proposal, with the per-channel formulas in its Definition field. The target policy notes travel with both entries. Adoption then follows the four steps of the register and GP-606; the proposal is filed with the entries it produced. When a definition is revised later, the next proposal starts from the last one's rejected options rather than reopening them. A Capacity Standards Workshop for Time Accounting is a natural setting for working several proposals at once, with the ledger agreed first.

Maturity Model Position

Four scales on this wiki use the word level; the launch page states which is which. On the WFM Labs Maturity Model™, the need for a proposal usually surfaces at Level 2, when the first scorecard reconciliation fails. Writing proposals against a ledger and a register, with rejected options on record, is Level 3 practice. At Level 4, every scorecard metric traces to an adopted proposal, and an agent asking what a column means can read the reasoning as well as the formula.

See Also

• The Definitions Register — the ten-field entry a proposal completes, and its change control
• Definitions and Source-of-Truth Change Control (generic process) — GP-606, whose first step uses this proposal form
• The Data Definition Card — the full field set an adopted definition carries
• Agent Time Accounting Model — the ledger tiers every numerator and denominator reads from
• Workforce Data Definitions Framework — the hub for ledger, registers, channels and practices
• Agent Utilization vs Occupancy Definitions and Tradeoffs — the two measures and their trade-offs
• Occupancy and The Occupancy Trap — occupancy as a pool-dependent output
• Production Accounting for Blended Agents — the email-in-available contamination
• Human Gates and Number Grades — the grades on the example's numbers