Peak Season Capacity Planning
What to build in advance, from Chuseok to year-end

This year Chuseok — Korea's autumn holiday — falls on September 25, with the break running September 24 through 27. As this piece goes out, that's one month away.

Nothing in warehousing is stranger than peak season. The dates are fixed years in advance, it came last year and the year before, even the magnitude is roughly predictable — and every year the same warehouses collapse in the same ways. Not because it's a disaster. Peak season is not a disaster; it is an exam with a printed date, and warehouses fail it not because the problems are hard but because they start studying the night before.

This piece is about doing that preparation in numbers: how to compute the demand, how to back-compute labor, space, and process capacity, and what price to put on that capacity. If you've read the costing series, the tools will be familiar — standard time, volume, the ledger. Peak season is their live-fire exercise.

Peak season isn't one wave — the calendar

Start with the calendar. A Korean warehouse's fourth quarter is not one wave but a chain of them.

There are real measurements to calibrate against. Korea's transport ministry runs a special parcel-management period of about four weeks around Chuseok every year; for Chuseok 2025, parcel volume was projected 13.5% above normal (the August average) at 20.3 million boxes a day, and the major carriers added about 5,500 workers across sortation, linehaul, and delivery. Lunar New Year 2026 ran at +5%, around 18.7 million boxes a day.

Handle those numbers with care, though. They are averages across the national parcel network. An individual warehouse's wave is far sharper. When the nation rises 13.5%, a warehouse carrying gift-set shippers doubles, and a single shipper riding a Singles' Day promotion can run 5× on its own volume. The costing series' lesson — averages don't represent individuals — applies here exactly: your warehouse's peak multiple lives in last year's ledger, not in national statistics.

And the last row — January's returns wave — rarely makes it onto anyone's calendar. In an era when return rates have doubled in five years ("The Structure of Reverse Logistics"), a fourth-quarter outbound wave always comes back as a January inbound-and-inspection wave. Preparing for the outbound peak but not the returns peak is answering only the front page of the exam.

Capacity is four things — and the bottleneck is one

"We're short on capacity" is not a diagnosis. A warehouse's capacity is at least four different resources, each with a different lead time to expand.

An old conclusion from the Theory of Constraints applies to peak season precisely — system throughput is set by the bottleneck. Double your pickers and, with six pack stations, outbound still moves at the speed of six pack stations. The first question of peak prep is not "what do we add" but "where did we jam last year."

The answer belongs to the ledger, not to memory. Open last year's peak-week throughput and queue records by activity and the bottleneck usually converges on one or two spots — and that's the address for this year's investment. Memory points to where it hurt last; the ledger points to where it jammed first. They are often different places.

Compute demand first — the shipper questionnaire

Before building capacity you need the denominator. Peak demand forecasting has three ingredients.

① Last year's ledger. Peak multiples by shipper and by week. Not the company average — peaks are made by a handful of shippers, not by the total.

② Shipper promotion calendars. These only surface if you ask. Run the quote questionnaire from "From Cost to Price" in its peak-season edition — expected volume bands per wave, promotion dates and target SKUs, gift sets or not (kitting, consolidation), pre-stocking plans and timing. Now — late August to early September — is the last good moment to send it.

③ SKU profile change. Peak orders are not an enlarged copy of normal orders. Gift sets are new SKUs, kitting is a new process, and the consolidation share climbs. In the language of the time equation: it's not just order count that jumps — standard time per order itself changes. If last year "volume was 1.8× but hours ran 2.3×," that difference is the profile shift, and this year's math has to include it.

Labor capacity — back-compute it from standard time

Once demand stands, labor is arithmetic. Using the costing series' example center: 20,000 orders a month in normal times — about 900 a day over 22 working days — at a single-line standard time of 5.5 minutes.

• Normal requirement: 900 × 5.5 min = 82.5 hours/day → about 10 people at 8 hours each
• Peak (2×) requirement: 1,800 × 5.5 min = 165 hours/day → about 21 people
• The gap: 10 people's worth — 82.5 hours a day

Two cards cover that gap — overtime and hiring — and each has a ceiling and a trap.

The overtime card's ceiling is set by law. Under Korea's 52-hour workweek, overtime is capped at 12 hours per person per week. Ten incumbents all maxing out give you 120 hours a week — barely 30% of the peak gap (412 hours a week). A plan to cross peak season on overtime alone doesn't just fail the arithmetic; in a four-week wave it starts eating efficiency from week two. Overtime is a card for spikes (days), not for waves (weeks).

The hiring card's trap is ramp-up. Because of the learning curve from Part 1 — 70% in weeks 1–3, 85% in weeks 4–8 — ten new hires dropped in on D-day are seven people's worth, not ten. Which turns hiring into a question not of headcount but of start date.

Same gap; the start date moves the hiring plan from 15 people to 10. If the Chuseok peak starts mid-September, D-2w is early September — meaning the job posting goes out this week.

Compare the costs, too. An overtime hour runs about 1.5× the standard rate with the 50% premium — around KRW 22,100. A new hire's effective hour in the 70% band costs 14,760 ÷ 0.7 ≈ KRW 21,100. Nearly identical. So overtime versus hiring is not a cost question — it's a question of wave length (a spike or four weeks), the legal ceiling, and the second-order effect of fatigue on efficiency. The favorable-rate-variance trap from "Variance Analysis and Cost Control" — cheap labor bleeding into efficiency losses — gets realized most often in peak season. In the first report after the hires land, read the rate and efficiency variances together.

Space capacity — the 85% wall

Peak-season space trouble starts not with outbound but with pre-stocking. Shippers start pushing gift sets in from early September, and the moment occupancy climbs from a normal 72% to 88%, the warehouse slows down before it fills up.

There is theory under this. Queueing theory — formalized by Erlang for telephone networks in 1904 and generalized in Kingman's formula — concludes that as utilization approaches 100%, waiting grows not linearly but explosively. The industry's hard-won "85% rule" — past 85% occupancy, putaway starts hunting for slots, double-handling multiplies, picking paths jam, and throughput actually bends downward — is that mathematics showing up on the floor. It's also why "Costing Storage" set target occupancy at 85%.

So space prep runs in three rungs.

1. Empty it. Clear slow-turning inventory before the season. If your contracts carry a long-term storage clause, now is when it fires — send shippers the 90-day no-movement list and have them choose: withdraw, dispose, or extend for a fee. A clause left sleeping is precisely the leak the price waterfall in "From Cost to Price" described.
2. Manage pre-stocking by slot. "Send it whenever" is what creates the second-week-of-September dock crush. Put per-shipper pre-stock volumes on a weekly reservation system — flatten the inbound and the same docks take twice as much.
3. Overflow by ladder. Densify internally first (temporarily narrowed aisles, upper levels — mind the honeycombing warning from the storage piece) → external overflow (on-demand warehousing — market rates run a 15–30% premium over standard shared storage; in peak season, judge by per-shipper contribution whether the volume is worth that premium) → and the last rung is refusal. Accepting every pallet of pre-stock in front of the 85% wall isn't service — it's selling every other shipper's outbound speed.

Process capacity — change the flow before buying more of anything

Of the four capacities, process has the shortest lead time. It costs almost nothing, changes in a week or two, and is therefore the card that stays playable the longest.

• Pre-kitting. Don't assemble gift sets in the peak week. Build kits in the first two weeks of September — the trough between waves — and put them away as finished SKUs; the peak week's per-order standard time drops by the whole assembly component. The heart of peak prep is moving peak work into the trough.
• Switch to batch picking. Promotion spikes concentrate orders on a few SKUs, so batch picking pays far more than in normal times. Set the trigger in advance (n+ orders of the same SKU) and flip the switch in peak week.
• Seasonal re-slotting. The ABC layout in "A Slotting Strategy That Cuts Picking Travel by 30%" is built on normal-season turnover — and the ranking flips in peak season. Run one seasonal slotting pass two weeks before the peak, promoting gift sets and promotion SKUs into the golden zone.
• Cutoffs and waves. Pull the order cutoff earlier for the peak weeks (the defensive line for same-day promises), and shorten wave cycles so work doesn't pool at the bottleneck station.
• Split the returns queue. During the peak, returns inspection competes for the same people as outbound. Put non-urgent returns inspection in a separate queue so an outbound peak doesn't cascade into a returns pile-up — with a hard deadline to digest it before the January wave.

Peak-season pricing — don't sell the capacity for free

Everything above was about building the capacity. The last part is its price.

Peak capacity is the most expensively bought resource of a warehouse's year — short-term labor at a premium, external space at a premium, an overtime budget spent forward. A contract that hands that resource out unlimited at normal-season rates makes peak season the time of year when margin falls as revenue rises. The market is already settling the answer — the 15–30% premium on on-demand warehousing means a market price now exists for "flexibility without commitment," and that logic applies equally to your own rate card.

Three practical devices, all extensions of "From Cost to Price":

• Volume bands. Inside the announced, committed band: contract rates. Beyond it: an overage rate. The commitment lets the warehouse build capacity; the shipper's notice buys rate stability.
• Peak surcharges in writing. Rush shipments, unannounced spikes, holiday-period work — put the surcharges in the contract. A surcharge that isn't written down doesn't get billed in the middle of the peak. The "unbilled" step of the price waterfall is at its widest exactly now.
• The price of holiday work. Commerce doesn't stop for the four-day Chuseok break, and no-holiday delivery has long since become the retail standard. Whether to take holiday shipments, and at what rate, is something to agree with shippers before the season — not on a phone call the first morning of the holiday.

What's different about peak season in 2026

① The peak became a time problem, not a volume problem. In the annual US 3PL survey, 65% run below 90% of capacity and 22% report stagnant volumes. Over the year, capacity is abundant — it's a few weeks that are short. From an era of owning capacity year-round to one of borrowing it for the weeks you need and handing it back — the axis has moved from "how much do we have" to "how much can we mobilize for the weeks that matter."

② The short-term labor market is tightening. US holiday e-commerce set a record $257.8 billion (+6.8%) in 2025, while seasonal retail hiring in the same season fell below 500,000 — a 16-year low — in a market where Amazon alone takes 250,000 at an average $23 an hour. Volume up, seasonal labor pool down; Korea's direction is the same. Good short-term workers are not on the market right before the peak — they're on the roster of the warehouse that moved early.

③ Flexible capacity became a real option. Robots-as-a-Service — rent robots for the season, return them after — is entering the warehouse, with subscription deployments rated 3–5× faster than traditional installs. Combined with on-demand space, a large share of peak capacity can now be assembled without owning it. One condition, though: borrowed capacity has a return date, so only warehouses that know from when to when, and how much — in numbers — can use the option. Which brings it back, as always, to demand math and the ledger.

What to start this week

A peak-prep piece can't end with "the first month." It's one month to Chuseok; the D-2w hiring deadline is half of that.

1. Send the shipper promotion questionnaire today. Volume bands per wave, promotion dates, gift sets or not, pre-stocking plans. Reply deadline: one week.
2. Build the one-page report on last year's peak. Peak multiple by shipper, the weekly volume curve, and where things jammed first. Out of the ledger.
3. Build the labor back-computation table and open hiring. Forecast volume × standard time → required heads → hires by start date with ramp-up applied. Put the D-2w deadline on the calendar.
4. Send the occupancy forecast and the inventory-cleanup notice. Chart weekly occupancy with pre-stocking included; if any week crosses 85%, start the long-term-inventory notices and the pre-stock slot system together.
5. Review peak surcharge and volume-band clauses. Separate the contracts that need them added at next renewal from the ones that have them but have never fired them.
6. Fix the date for the D-2w rehearsal. Pick a day and load-test at forecast peak volume — find out where the bottleneck appears on the floor rather than in the plan, and finish the emergency-contact and alternate-flow runbook that same day.

The exam is only as hard as the preparation

When the season ends, two things remain: January's returns wave — and the data.

The peak-season ledger is the densest observation of the year. Measured peak multiples, bottleneck locations, the actual ramp-up curve of your new hires, what the profile shift did to hours. Use it to sharpen next year's questionnaire and to re-examine standard times — with one caution: don't mix peak-season actuals into normal-season standards. Numbers made of overtime, new hires, and improvised flows are observations of the peak, not the baseline of normal times. Per the revision discipline in "Variance Analysis and Cost Control," tag the peak period and read it separately.

The dates are fixed. The volume comes from the questionnaire and the ledger. The headcount comes out of standard time. The limit of the space stands at a number: 85%. If peak season is an annual exam, it has one peculiarity — the exam paper is published in advance.