River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Recruiting Capacity Plan Template

Four documents and three sheets, where the annual arithmetic says the plan fits but the monthly arithmetic is the one worth reading.

Free download  ·  No account needed

The same plan, tested two ways

89 percent a year. 267 percent in April

Orenda Retail Group, a fictional specialty retailer. 26 hires, one recruiter, one coordinator at 0.4 FTE.

The annual test
Recruiter attention the plan demands234.0 req-weeks
Available, after 8 weeks of PTO, holidays and non-search work264.0 req-weeks
Utilisation89%
Headroom30 req-weeks
The monthly test
Peak week, 4 May: requisitions open at once15
Concurrent ceiling, observed in their own data6
April utilisation, on three available weeks267%
Hires landing on or before the date asked for3 of 26

And Three Were Late On Day One

Working backwards from the dates the business gave, R-01 Store Manager had to be open on 24 December, R-21 Ecommerce Developer on 29 December, R-24 Finance Analyst on 30 December. The plan was approved on 12 January. No amount of recruiting capacity fixes a date in the past, and a plan built from a forecast rather than a backward pass finds this out in March.

A headcount plan is a list of hires with dates on them. Whether it can be delivered is a different question, and the annual arithmetic almost always answers it wrongly. Orenda Retail Group's 26 hires need 234 requisition-weeks of recruiter attention against 264 available: 89 percent, 30 weeks of headroom, nothing to discuss. Twenty-three of those hires land late. The reason is invisible until the same demand is laid out month by month, which is what this pack does.

The days come from your own closed requisitions by role type, not a benchmark. Orenda has 97 of them, and the spread is the point: 22 days for a distribution associate, 112 for an ecommerce developer. A generic 45-day assumption would have priced that one role type at 6.4 weeks a search instead of 16, understating it by 29 requisition-weeks, which is almost the entire headroom the annual test claimed to find. Weak groups get flagged rather than hidden.

Then two tests instead of one. The backward pass from each need-by date says when the requisition had to be open, and three of Orenda's had to be open before the plan was approved, which is a finding no forward forecast can reach. The concurrency count says the peak week wants 15 open requisitions against a ceiling of 6 that came out of their own data rather than a rule of thumb. Where a single requisition's trade-off belongs at intake, this is the whole plan at once.

Three of twenty-six land on their date, and the annual figure says 89 percent

The backward pass, the month it breaks, and the ceiling at which raising capacity stops changing the answer.

Role Pipeline

Orenda Retail Group, FY2026, 26 hires. Medians from 97 closed requisitions over 24 months, requisition approved to offer signed. Twelve of the 26 rows shown.

ReqRole typeTNeed-byMedian
days
nMust openModel
fills
Slip
R-01Store Manager12 Mar681124 Dec 2521 Mar+19
R-02Store Manager16 Apr681128 Jan28 May+52
R-11Distribution Associate111 May223719 Apr26 May+15
R-13Distribution Associate125 May22373 May26 May+1
R-16Merchandise Planner113 Apr84619 Jan13 Apr0
R-17Merchandise Planner211 May84616 Feb6 Jul+56
R-18Merchandise Planner38 Jun84616 Mar15 Nov+160
R-19Buyer14 May97527 Jan4 May0
R-20Buyer229 Jun97524 Mar22 Sep+85
R-21Ecommerce Developer120 Apr112429 Dec 254 May+14
R-22Ecommerce Developer14 May112412 Jan4 May0
R-24Finance Analyst216 Mar76730 Dec 2529 Mar+13
26 hires, 8 role types, 235.86 requisition-weeks of demand. 3 land on time, 23 late by 1 to 160 days, 1,042 days of slip in total. Four of the eight role types rest on fewer than seven closed requisitions and account for nine of the hires.

Read the must-open column before anything else. It is need-by minus the role type's median, and it is the only arithmetic on this page with no model behind it. Three rows land in December 2025 against a plan approved on 12 January 2026, so they are late before a recruiter has done anything. That finding is unavailable to any plan that forecasts forward from today, which is every plan.

Recruiter Load

The same 26 requisitions, month by month. Available weeks are 52 less eight: three PTO, two holiday, three of named non-search work. The eight are placed in the months they actually fall in.

MonthPeak openCeilingOverWeek lost toWeeks
avail
Req-weeks
demanded
Req-weeks
available
Util
Dec 202536before approval01.860n/a
Jan 202676+1careers site rebuild319.2918107%
Feb 202696+3EEO-1 filing330.7118171%
Mar 2026116+5441.8624174%
Apr 2026146+8PTO348.0018267%
May 2026156+9545.8630153%
Jun 202696+3417.572473%
Jul 202656PTO416.572469%
Aug 202646PTO311.431863%
Sep 202626employer brand32.711815%
Oct-Dec 2026062 holiday weeks120.00720%
Total 2026156+98 of 52 weeks44234.0026489%

The bottom row and the April row are the same plan. April is the sharpest cell because it carries 267 percent of capacity in the month the recruiter has a week of leave booked, and the annual 89 percent is arithmetically correct while telling you nothing. Twenty of the 39 weeks the plan spans run over the ceiling. Six is not a rule of thumb either: above seven open requisitions Orenda's median days to fill moved from 36 to 58, and the same threshold was four before the coordinator joined at 0.4 FTE.

Gap to Plan

Hires the plan wanted against hires the forward pass delivers at a ceiling of six, cumulative. Requisitions queue on their must-open date and take a slot when one frees, highest priority tier first.

MonthPlannedDeliveredCum
planned
Cum
delivered
BehindLanding that month
Mar 202621211R-01
Apr 202632532R-16 R-24
May 20268813112R-02 R-05 R-11 R-12 R-13 R-19 R-21 R-22
Jun 20268321147R-06 R-14 R-15
Jul 20261222166R-07 R-17
Aug 20262324195R-03 R-09 R-26
Sep 20262326224R-04 R-10 R-23
Oct 20260226242R-08 R-20
Nov 20260126251R-18
Dec 20260126260R-25

Both readings of this table are arithmetically true and they say opposite things. Judged on need-by dates, the plan is seven hires behind at the end of June and 23 of 26 hires are late. Judged on total hires by 31 December, it delivers 26 of 26 and the scorecard is green. Which reading the review uses is a choice somebody has to make in January, and the assumptions document makes it an explicit line rather than something discovered at the December board meeting.

Ceiling Sweep

The same simulation re-run at higher concurrent ceilings. Recruiter-equivalents are the ceiling divided by the current one, so the answer is in people rather than slots.

CeilingRecruiter-equivalentsLand on timeTier 1 lateTotal slip
61.003 of 2671,042 d
81.3386378 d
91.50105256 d
111.8316492 d
122.0020365 d
152.5023246 d
183.0023246 d
264.3323246 d
Fifteen is where it stops. 2.50 recruiter-equivalents, which is the same 15 over 6 the peak week already reported. The residual 46 days is R-01, R-21 and R-24, whose open dates had already passed.

The flat tail is the useful part of this table. Past 15 the curve does not move, because the only requisitions still late are the three the backward pass already condemned, and that is the sentence to take into the review: not that more recruiters are needed, but that 2.5 recruiter-equivalents delivers everything deliverable and three of these dates were never available. What Orenda actually did was cheaper. Re-dating the 13 requisitions with no document behind their date, buying two searches outside at 55,200, withdrawing one and pooling six assistant manager searches into a single evergreen pipeline cut total slip from 1,042 days to 103 and the peak from 15 to 8. Still over the ceiling, and the pack says so.

What is in the pack

01

Role Pipeline

One row per hire carrying the role type's median days to fill, the number of closed requisitions behind that median, the requisition-weeks it costs, and the date it must already be open. Then the modelled open, the modelled fill, the slip in days, and a column saying whether the row is late at any capacity at all.

02

The backward pass

Need-by date minus the role type's own median. Two numbers, no model, and it is the step that finds the requisitions late before the plan was signed. Three of twenty-six in the worked example, against a plan approved on 12 January.

03

Recruiter Load

Requisition-weeks demanded and available in every month, with peak concurrent open requisitions against the ceiling. Weeks lost to PTO, holidays and named non-search projects are placed in the months they fall in rather than deducted off the top, because April losing a week matters and October losing one does not.

04

The concurrency ceiling

Not a rule of thumb. The load at which your own median days to fill starts rising, computed from the same export as the medians. Orenda's is six, and it was four before a coordinator joined at 0.4 FTE. Every peak figure in the pack is printed next to it.

05

Gap to Plan

Hires planned against hires delivered by month, cumulative, with the requisitions named in both columns so any row can be traced. Plus the point the two readings of it diverge: seven hires behind at the end of June on need-by dates, twenty-six of twenty-six on year-end totals.

06

Plan Assumptions

Nine observations separated from two choices, with a section listing where the model is optimistic. The forward pass does not stop for PTO, no requisition is assumed to fall through, and the ceiling is treated as flat. If the plan gets argued about, the argument is really about a line in here.

07

Prioritization Note

The output that goes to the executive team. Annual utilisation and worst monthly utilisation in the same sentence, what lands, the requisitions nobody can save, then the three levers priced: raise capacity, buy the peak, or move the dates. Total cost and cost per requisition recovered, because the fees usually look cheaper and buy less.

08

Sequencing Decision

Written after the review. Every requisition assigned to exactly one outcome so the partition adds up, a name against every accepted slip, the document behind every immovable date, and the model re-run against the decision. Including the one requisition the decision made worse, which is the part a before-and-after table usually leaves out.

How it works

  1. 1

    Send the plan and two years of closed requisitions

    With the requisition approval date and the offer-signed date. Medians come back by role type with the group size printed beside each one, and any group under five flagged in the same row rather than in a footnote.

  2. 2

    The backward pass runs first

    Need-by minus median gives the date each requisition must already be open. Anything landing before the plan approval date comes back named, with the days already gone and the three real options, none of which is trying harder.

  3. 3

    Then the month you cannot deliver

    Requisition-weeks and peak concurrency per month against the ceiling, with the annual figure printed beside the worst monthly one so the gap between them cannot be skimmed.

  4. 4

    Three levers, priced, and the decision recorded

    Capacity with its ramp, agency fees per requisition recovered, and re-dating the requisitions that have no document behind their date. Then every requisition assigned to one outcome and the model re-run against it.

Frequently asked questions

Why not just divide the hires by the recruiters?

Because that is the annual test, and it passes. Twenty-six hires against 264 available requisition-weeks is 89 percent utilisation with 30 weeks spare. The same plan runs at 267 percent in April and delivers three hires on their date. Demand is not spread evenly across a year and the annual figure cannot see that.

Where does the concurrent ceiling come from?

Your own export. Group closed requisitions by how many were open at the time and find the load where median days to fill starts rising. Orenda's median went from 36 days at six or fewer to 58 at seven or more, so the ceiling is six. A guessed ceiling invalidates everything downstream, so the pack asks for it before anything else.

What if we have no history for a role type?

Use the nearest adjacent type in your export, say which one you substituted, and record it in the assumptions. Never a benchmark. In the worked example a generic 45-day figure would have understated one role type by 29 requisition-weeks, which was almost the whole headroom the annual test found.

Do the medians need two years of records?

They need whatever you have retained. Title VII's rule sets a floor: records having to do with hiring are preserved for one year from the making of the record or the personnel action. Larger federal contractors sit at not less than two years. Floors, not targets, and a 24-month median needs a policy above them.

Three requisitions were late before the plan was approved. Really?

It is the most common finding, and it is arithmetic rather than blame. A 112-day median and a 20 April date means opening on 29 December. Adding recruiters does not help; the pack tests that at unlimited capacity and the same three stay late by the same margins. Move the date, buy the search, or sign for the slip.

So the answer is always hire more recruiters?

Usually not. The cheapest lever is re-dating the requisitions that have no document behind their date, and it is the one nobody proposes. Orenda re-dated thirteen, bought two searches outside, withdrew one and pooled six searches into one pipeline. Slip fell from 1,042 days to 103 without a single extra recruiter arriving in time to matter.

Where does this sit against the rest of hiring?

Above it. Intake settles one requisition's trade-off, the requirement register turns it into a posting, and funnel analysis says why an open role is stuck. This asks whether the plan was ever deliverable. Four documents and three CSV sheets, free.

Find out in January, not in June

Send the headcount plan and two years of closed requisitions. The first thing back is the list of requisitions that needed to be open before the plan was approved, with the days already gone.

Edit with AI