Operate · item 25
Capacity and surge plan
A shared service is judged on quarter close, not on an average Tuesday. Every committed level on this estate is priced against a throughput assumption and a number of humans who must be awake to approve things. This page states those assumptions, forecasts the queue into each peak, and says plainly where the plan is short of people before the peak arrives rather than after.
Health
Whether there are enough hands and agents
Plans covered, tight and short, against the queues forecast to breach their service level.
Peak plans
8
One per named peak on the calendar
Short of humans at peak
5
2 tight · 1 covered
Peak against base week
3.5x
64,265 units in a base week · 222,250 at peak
Playbooks never run
3
5 have been rehearsed at least once
Surge levers
21
6 of them have never been pulled
Forecast periods at risk
5
2 high · 3 elevated
Human coverage at peak
115.0 full-time equivalents needed · 108.6 available
Record-to-Report quarter close-24.2% headroom
needed
14.9
available
11.3
Quarter close · Final two business days of the quarter plus the following four · 7.5 review minutes per unit · short at peak
Source-to-Pay year-end payment run+3.8% headroom
needed
18.4
available
19.1
Year end · Final ten business days of the fiscal year · 3.2 review minutes per unit · tight at peak
Hire-to-Retire open enrollment-16.0% headroom
needed
24.3
available
20.4
Open enrollment · Three weeks in the second month of the fourth quarter · 5.5 review minutes per unit · short at peak
Order-to-Cash quarter-end order surge-3.9% headroom
needed
23.3
available
22.4
Quarter-end order surge · Final five business days of the quarter · 4.1 review minutes per unit · short at peak
Customer Service major incident surge-12.3% headroom
needed
7.3
available
6.4
Product incident · Unpredictable, typically 6 to 48 hours · 2.4 review minutes per unit · short at peak
Finance external audit window+7.0% headroom
needed
12.8
available
13.7
External audit · Six weeks following the year-end close · 24.0 review minutes per unit · tight at peak
Human Resources hiring surge+14.2% headroom
needed
12.0
available
13.7
Hiring surge · First eight weeks of the fiscal year · 9.5 review minutes per unit · covered at peak
Marketing campaign launch-20.0% headroom
needed
2.0
available
1.6
Campaign launch · Two weeks around each major launch · 6.0 review minutes per unit · short at peak
Human need at peak is derived, not asserted: peak volume times the share of units that still route to a person, times the review minutes each of those takes, spread over the working hours in the peak window. 5 of the 8 plans come out short. Nothing on this page rounds that away.
Record-to-Report quarter close
Quarter close · Final two business days of the quarter plus the following four · owner Ana Duarte
Short at peak
A shared service is judged on quarter close, not on an average Tuesday. This plan is the one that matters and it is the one with the thinnest human margin.
Base week
7,294
units
Peak week
24,800
3.4x the base week
Agent throughput
412/hr
units the agents clear per hour
Humans, base
4.4
full-time equivalents
Humans, peak
14.9
11.3 available
Headroom
-24.2%
Short at peak
Queue depth forecast into the peak
W-5
0
178h need
424h have
actual 7,905
W-4
0
159h need
424h have
actual 7,070
W-3
0
218h need
424h have
actual 9,674
W-2
0
220h need
424h have
actual 9,771
W-1
0
323h need
424h have
actual 14,358
Peak week
798
524h need
424h have
actual 23,265
W+1
0
335h need
424h have
W+2
0
185h need
424h have
bars are items left waiting at the end of the week · scaled within this plan · peak week highlighted · forecast by Prior four closes, weighted to the most recent, with a manual overlay for known entity changes, historical error 11.4%
Surge levers
Extend agent autonomy on low-materiality journals from A2 to A3 for the windowpulled before
Removes roughly 2,100 approvals from the human queue
lead time Same day, requires a named approval
Draw four reviewers from the Source-to-Pay towerpulled before
Adds about 130 review hours
lead time 48 hours
Defer non-statutory reporting by five business daysnever pulled
Frees roughly 90 hours
lead time Requires controller sign-off
Peak playbook
Freeze changes 72 hours before period end. Move the intake gate to continuous release with a rostered releaser. Stand up a twice-daily control call. Raise the escalation tier by one for the duration so that medium items reach a human faster rather than slower.
rehearsed 2 months ago on Jun 8, 2026
Assumptions behind the committed level
Assumes the same entity count as the prior close, no late journal categories, and that the two controllers who sign are both available in the window.
The entity count has grown in three of the last four closes. The forecast has not been re-based, so it is structurally low.
Source-to-Pay year-end payment run
Year end · Final ten business days of the fiscal year · owner Tobias Renn
Tight at peak
Suppliers submit against the year-end cutoff and the arrival curve is steeper than any other period on the estate.
Base week
22,778
units
Peak week
61,500
2.7x the base week
Agent throughput
1,840/hr
units the agents clear per hour
Humans, base
6.8
full-time equivalents
Humans, peak
18.4
19.1 available
Headroom
+3.8%
Tight at peak
Queue depth forecast into the peak
W-5
0
275h need
716h have
actual 24,561
W-4
0
259h need
716h have
actual 23,110
W-3
0
313h need
716h have
actual 27,931
W-2
0
333h need
716h have
actual 29,714
W-1
0
518h need
716h have
actual 46,241
Peak week
0
680h need
716h have
actual 60,695
W+1
0
440h need
716h have
W+2
0
289h need
716h have
bars are items left waiting at the end of the week · scaled within this plan · peak week highlighted · forecast by Seasonal index on three prior years, adjusted for supplier count, historical error 8.1%
Surge levers
Raise the touchless threshold on matched three-way invoicespulled before
Removes roughly 6,400 reviews
lead time Same day
Add a second review shift in the Manila centerpulled before
Adds about 260 review hours
lead time Two weeks, needs rostering
Hold non-urgent supplier categories to the first week of the new yearnever pulled
Defers roughly 18% of volume
lead time Requires procurement sign-off
Peak playbook
Pre-clear the exception backlog to under 200 items ten days out. Roster continuous quarantine release. Pre-approve the touchless threshold change so it does not need a decision under pressure. Hold a daily payment run rather than twice weekly.
rehearsed 8 months ago on Dec 28, 2025
Assumptions behind the committed level
Assumes supplier behavior matches prior years and that the payment gateway is available throughout. The gateway is one of the two systems with no live feed.
The payment gateway connector has never been exercised at year-end volume because it has never been connected.
Hire-to-Retire open enrollment
Open enrollment · Three weeks in the second month of the fourth quarter · owner Priya Raman
Short at peak
Every employee interacts with the tower inside a three-week window. Volume is predictable and the human review need is not, because the questions are novel each year.
Base week
9,317
units
Peak week
38,200
4.1x the base week
Agent throughput
2,600/hr
units the agents clear per hour
Humans, base
5.9
full-time equivalents
Humans, peak
24.3
20.4 available
Headroom
-16.0%
Short at peak
Queue depth forecast into the peak
W-5
0
251h need
765h have
actual 10,530
W-4
0
207h need
765h have
actual 8,671
W-3
0
318h need
765h have
actual 13,342
W-2
0
270h need
765h have
actual 11,343
W-1
0
473h need
765h have
actual 19,834
Peak week
805
839h need
765h have
actual 35,194
W+1
0
577h need
765h have
W+2
0
241h need
765h have
bars are items left waiting at the end of the week · scaled within this plan · peak week highlighted · forecast by Headcount times prior-year participation rate, with a flat uplift for plan changes, historical error 16.8%
Surge levers
Publish the twelve highest-volume answers as self-service before the window openspulled before
Historically removes about 30% of contacts
lead time Four weeks
Bring in six seasonal reviewersnever pulled
Adds roughly 190 review hours
lead time Six weeks, requires a budget decision
Extend the window by five dayspulled before
Flattens the curve by roughly 20%
lead time Requires a benefits decision
Peak playbook
Stand up the seasonal roster four weeks out. Move the escalation tier for benefits questions to a named human within the hour. Publish a daily unanswered-question digest so that novel questions become self-service inside 24 hours rather than at the end of the window.
never rehearsed — this playbook has only ever been read
Assumptions behind the committed level
Assumes participation matches last year and that plan changes generate the same question volume per change as last year. Neither assumption has been tested.
The uplift for plan changes is a single flat number applied regardless of how substantial the change is. In a year with a carrier change this plan is materially wrong.
Order-to-Cash quarter-end order surge
Quarter-end order surge · Final five business days of the quarter · owner Morgan Idris
Short at peak
Roughly a third of quarterly order value arrives in the last five days, and it arrives with the tightest committed cycle time on the estate attached to it.
Base week
12,154
units
Peak week
47,400
3.9x the base week
Agent throughput
1,420/hr
units the agents clear per hour
Humans, base
6.0
full-time equivalents
Humans, peak
23.3
22.4 available
Headroom
-3.9%
Short at peak
Queue depth forecast into the peak
W-5
0
249h need
840h have
actual 13,484
W-4
0
229h need
840h have
actual 12,393
W-3
0
299h need
840h have
actual 16,204
W-2
0
284h need
840h have
actual 15,378
W-1
0
452h need
840h have
actual 24,513
Peak week
0
833h need
840h have
actual 45,166
W+1
0
566h need
840h have
W+2
0
258h need
840h have
bars are items left waiting at the end of the week · scaled within this plan · peak week highlighted · forecast by Pipeline-weighted arrival model refreshed weekly from the customer platform, historical error 9.6%
Surge levers
Pre-approve credit for the top 200 accountspulled before
Removes roughly 40% of holds
lead time One week
Stand up a deal desk bridge with three additional approverspulled before
Adds about 95 approval hours
lead time 72 hours
Suspend the non-standard terms path and route it to Legal directlypulled before
Removes about 400 items
lead time Same day
Peak playbook
Freeze discount band changes for the final ten days. Pre-clear credit on the named account list. Move the desk to continuous cover across both hemispheres. Run the order queue snapshot hourly rather than daily so that the oldest item is visible while it can still be recovered.
rehearsed 34 days ago on Jul 15, 2026
Assumptions behind the committed level
Assumes credit checks clear at the historical rate and that the discount approval band does not change during the window.
The discount band was tightened mid-quarter twice in the last year, each time pushing volume into the human queue during the surge.
Customer Service major incident surge
Product incident · Unpredictable, typically 6 to 48 hours · owner Avery Chen
Short at peak
The only plan on the estate with no calendar. Volume arrives in minutes and the externally facing credit clause is attached to the same queue.
Base week
4,774
units
Peak week
29,600
6.2x the base week
Agent throughput
3,100/hr
units the agents clear per hour
Humans, base
1.2
full-time equivalents
Humans, peak
7.3
6.4 available
Headroom
-12.3%
Short at peak
Queue depth forecast into the peak
W-5
0
44h need
240h have
actual 4,747
W-4
0
45h need
240h have
actual 4,894
W-3
0
53h need
240h have
actual 5,707
W-2
0
61h need
240h have
actual 6,650
W-1
0
79h need
240h have
actual 8,627
Peak week
1,000
280h need
240h have
actual 30,433
W+1
0
170h need
240h have
W+2
0
48h need
240h have
bars are items left waiting at the end of the week · scaled within this plan · peak week highlighted · forecast by None. Capacity is sized against the largest incident in the last eight quarters., historical error 0.0%
Surge levers
Publish an incident banner and a holding responsepulled before
Historically deflects 35 to 50% of contacts
lead time 30 minutes
Move Tier 2 reviewers onto first responsepulled before
Adds about 40 response hours a day
lead time 2 hours
Raise agent autonomy on status-only responses to A3never pulled
Removes review from roughly 60% of contacts
lead time Same day, named approval
Peak playbook
Declare within 15 minutes. Publish the holding response before the queue builds. Move to a single merged queue rather than per-channel queues. Record every contact against the incident so that the credit exposure is known while the incident is running rather than at month end.
rehearsed 4 months ago on Apr 22, 2026
Assumptions behind the committed level
Assumes an incident no larger than the worst observed, and that the deflection content can be written and published inside two hours.
Sizing against the worst observed incident is not a forecast. A larger incident breaches the only externally facing commitment on the estate, and the credit is contractual.
Finance external audit window
External audit · Six weeks following the year-end close · owner Grace Abbott
Tight at peak
Evidence requests arrive against work already completed, so the surge is in retrieval and explanation rather than in processing.
Base week
2,306
units
Peak week
4,150
1.8x the base week
Agent throughput
260/hr
units the agents clear per hour
Humans, base
7.1
full-time equivalents
Humans, peak
12.8
13.7 available
Headroom
+7.0%
Tight at peak
Queue depth forecast into the peak
W-5
0
302h need
514h have
actual 2,604
W-4
0
253h need
514h have
actual 2,183
W-3
0
363h need
514h have
actual 3,129
W-2
0
344h need
514h have
actual 2,966
W-1
88
549h need
514h have
actual 4,731
Peak week
0
455h need
514h have
actual 3,925
W+1
0
305h need
514h have
W+2
0
290h need
514h have
bars are items left waiting at the end of the week · scaled within this plan · peak week highlighted · forecast by Prior-year request count by category, adjusted for scope changes agreed with the auditor, historical error 13.2%
Surge levers
Pre-generate evidence packs for the highest-volume request categoriespulled before
Removes roughly 300 retrieval hours
lead time Two weeks
Second reviewer from Internal Auditnever pulled
Adds about 60 hours
lead time One week
Peak playbook
Agree the request taxonomy with the auditor before the window opens. Pre-generate packs for the top categories. Route every request through a single intake so that duplicate requests are answered once.
never rehearsed — this playbook has only ever been read
Assumptions behind the committed level
Assumes the auditor accepts evidence packs generated from the spine and does not require a parallel manual reconstruction.
The auditor has not yet accepted a generated evidence pack as sufficient. If a parallel reconstruction is required, this plan understates effort by a wide margin.
Human Resources hiring surge
Hiring surge · First eight weeks of the fiscal year · owner Priya Raman
Covered at peak
Requisitions open against the new plan and the screening agent carries an open fairness alarm through the same window.
Base week
4,448
units
Peak week
12,900
2.9x the base week
Agent throughput
940/hr
units the agents clear per hour
Humans, base
4.1
full-time equivalents
Humans, peak
12.0
13.7 available
Headroom
+14.2%
Covered at peak
Queue depth forecast into the peak
W-5
0
182h need
514h have
actual 5,214
W-4
0
138h need
514h have
actual 3,958
W-3
0
222h need
514h have
actual 6,380
W-2
0
191h need
514h have
actual 5,477
W-1
0
330h need
514h have
actual 9,473
Peak week
0
400h need
514h have
actual 11,469
W+1
0
282h need
514h have
W+2
0
173h need
514h have
bars are items left waiting at the end of the week · scaled within this plan · peak week highlighted · forecast by Approved headcount plan times historical requisition-to-hire ratio, historical error 19.7%
Surge levers
Add two contract recruiterspulled before
Adds about 70 review hours a week
lead time Three weeks
Batch screening review into twice-daily blockspulled before
Improves reviewer throughput by roughly 15%
lead time Same day
Peak playbook
Stage requisition opening across the eight weeks rather than releasing the full plan at once. Hold a weekly fairness reading review during the surge instead of monthly.
rehearsed 7 months ago on Feb 3, 2026
Assumptions behind the committed level
Assumes the approved plan is executed evenly and that screening runs at current autonomy throughout.
Candidate screening carries an open adverse-impact alarm at 0.76. If autonomy is reduced in response, human review need roughly doubles and this plan does not allow for that.
Marketing campaign launch
Campaign launch · Two weeks around each major launch · owner Avery Chen
Short at peak
Asset approval volume triples and the review gate is a single named person for brand-sensitive material.
Base week
1,194
units
Peak week
3,700
3.1x the base week
Agent throughput
520/hr
units the agents clear per hour
Humans, base
0.6
full-time equivalents
Humans, peak
2.0
1.6 available
Headroom
-20.0%
Short at peak
Queue depth forecast into the peak
W-5
0
30h need
60h have
actual 1,514
W-4
0
21h need
60h have
actual 1,047
W-3
0
35h need
60h have
actual 1,725
W-2
0
30h need
60h have
actual 1,473
W-1
0
52h need
60h have
actual 2,606
Peak week
20
62h need
60h have
actual 3,099
W+1
0
46h need
60h have
W+2
0
27h need
60h have
bars are items left waiting at the end of the week · scaled within this plan · peak week highlighted · forecast by Launch calendar times average assets per launch, historical error 22.4%
Surge levers
Pre-approve template-conforming assetspulled before
Removes roughly 45% of reviews
lead time One week
Nominate a standing brand delegatenever pulled
Removes the single-approver constraint
lead time Requires a decision that has been open for two quarters
Peak playbook
Lock the asset list ten days out. Route template-conforming assets straight through. Publish the review queue so the bottleneck is visible to the launch owner rather than only to the reviewer.
never rehearsed — this playbook has only ever been read
Assumptions behind the committed level
Assumes assets per launch matches the historical average and that brand review remains a single approver.
Assets per launch has a wide spread and the average hides it. A single approver is a structural single point of failure in this plan and it has never been tested against a launch during a leave period.
Actions
What is waiting on a person
Capacity is the one place where the answer is almost always a person, because the shortfall is people.
Operations
What this desk is allowed to start
A surface that only reports is not operable. This is the work this page can set in motion, and the bound it runs into.
Trigger and bound
This desk can start a forecast run — project queue arrivals against declared capacity and publish a breach risk. It cannot hire, reassign a person, change a shift pattern, or move an agent ceiling. Every one of those is an operating decision owned by a function leader.
Live observability
What the record shows right now
Plan coverage. Tight means covered today with no margin.
Current distribution
8 plans
Covered113%
Tight225%
Short563%
Is policy and strategy coming to fruition
Whether the written intent is holding here
1 of 8 capacity plans are genuinely covered. 5 queues are forecast to breach.
Not holding on the record
The strategy position behind automation on this estate is that agent capacity absorbs volume growth so headcount does not have to. Against that, 5 of 8 plans are short today and 2 have no headroom, and 3 surge playbooks have never been exercised. Forecasts on this page are modeled from historical arrival rates — they are labeled as forecasts because none of them has yet been compared against what actually arrived.
What this page is, and what it is not
Drill results are measured against a modeled estate on a rehearsal calendar of our own making. Recovery times here are real measurements of a simulation, not evidence that a customer's production estate recovers in the same time.
Volumes, throughput and the review-minute assumptions are modeled. The method is the deliverable: state the peak, state the throughput and staffing the committed level was priced against, forecast the queue into the peak, and name the levers with their lead times. Worst forecast error on file is 22.4%, which is carried on the plan rather than hidden behind a point estimate.
3 of the 8 peak playbooks have never been run, and 6 of the 21 surge levers have never been pulled. An untested playbook is shown as untested here. It is not shown green because it exists.
This page covers surge capacity for peaks. It is not a workforce transition plan: what happens to the people whose work the agents now do is a separate question this estate does not yet answer.