How to Build a Storage System That Scales

A shoebox works at five hundred cards and seizes up at five thousand. The fix is not more boxes. It is an address, an index, and one rule about what never gets sorted twice.

A shoebox is a perfectly good storage system for five hundred cards. At five thousand it is a problem, and at twenty-five thousand it is not a system at all — it is a landfill you happen to own. The box is not the failure. The failure is that most collections are organised by memory, and memory does not scale. Structure does.

THE SYSTEM, IN ORDER

1. 01 Count in rows before you buy a box
2. 02 Split the pile into four tiers, once
3. 03 Give every box a permanent address
4. 04 Index boxes, never cards
5. 05 Run one intake box and one rule
6. 06 Leave a fifth of every box empty
7. 07 Audit twice a year, on a clock

PHASE A

Size the pile

COUNT → TIER → ADDRESS

1. Count in rows before you buy a box

Nobody counts cards one at a time, so count rows. A standard 800-count box is one row of eight hundred. The hobby calls the big one a monster box; BCW's catalogue calls the same object a 3,200-count storage box, and it is four of those rows side by side. Lay the pile out in row-lengths and you have a real number in four minutes.

Then size for four-fifths full. Say the pile comes to 9,400 cards: divided by 640 — eight hundred at four-fifths — that is 14.7, so you buy fifteen boxes. Buy twelve and you spend next year shuffling.

2. Split the pile into four tiers, once

Value first, because value decides the holder. Under a quarter is bulk — unsleeved, set order only, into the monster box. Tier C, a quarter to five dollars: penny sleeve, sorted box. Tier B, five to fifty: sleeve and a 35pt toploader, upright. Tier A, fifty to two hundred: a magnetic one-touch and a line of its own. Tier S, over two hundred, is a grading conversation and a semi-rigid until it ships. Same breakpoints as the protection stack, deliberately.

The ratio matters more than the tiers. In every bulk lot I have sorted, the large majority of cards never left bulk. Build for that majority.

3. Give every box a permanent address

A row letter and a box number — A-01 through D-06 — written on both ends and on the lid, in marker, before anything goes in. The address names a place on a shelf. It never names the contents.

The moment a box is called "Pokémon 3" its name is a promise about what is inside, and the first set you add breaks it and renames everything downstream. Places hold still. Contents do not. Label the thing that holds still.

+ SPEC — THE FILL LINE

640 of 800

A box packed to the lid is not full. It is frozen. Every card you file after that shoves the row into the next box.

FOUR-FIFTHS TARGET FILL · HOLD THE AIR

UNDER 400

half empty — you are paying an index line for nothing; merge it

600 – 660

the working range — a card files in one motion

OVER 720

nearly full and climbing — every insert becomes a shelf-wide shuffle

A FIFTH OF EVERY BOX IS NOT WASTED SPACE. IT IS THE ONLY REASON THE SYSTEM STILL WORKS IN TWO YEARS.

FIG. 01

What "scales" actually means

SCROLL THE CHART SIDEWAYS — OR READ THE NUMBERS BELOW IT

MINUTES AT 500 · 5,000 · 25,000 · 60,000 CARDS
one pile, no order — 3 · 26 · 115 · 240
sorted by set, no index — 2 · 6 · 11 · 17
addressed + indexed — 2 · 2 · 3 · 3

PHASE B

The index

ONE LINE PER BOX

4. Index boxes, never cards

One line per box: address, contents range, count, date counted. Twenty-four boxes is twenty-four lines. It fits in a phone note and it survives a Sunday afternoon, which is the entire specification.

Index card by card and those same twenty-four boxes become 19,200 rows — a spreadsheet you abandon in five weeks and distrust for years. The exception is the top of the pile: Tier A and Tier S get a line each, because there are thirty-five of those, not nineteen thousand. Small enough to maintain is the only test that matters.

5. Run one intake box and one rule

Everything new lands in a single box marked INTAKE, and nothing else gets to be a staging area — no desk piles, no "I'll deal with that later" stack on the shelf. A sort session empties intake through the tiers and every card leaves for a final address.

The rule: a card that has left intake never re-enters it. That one sentence is what stops a collection being sorted twice, and re-sorting is where the hours actually go. My gate-one pass runs about fifteen cards a minute, so a thirty-minute session clears roughly 450 cards. Sessions, not marathons — marathons are how cards end up in the wrong box at midnight.

FIG. 02

Twenty-four boxes, twenty-four lines

SCROLL THE WALL SIDEWAYS — ROWS A TO D, SIX BOXES EACH

THE INDEX LINE FOR THE GOLD BOX
C-02 — 800 slots, 640 used — SWSH Base → Evolving Skies — counted 2026-06-14

PHASE C

Keeping it alive

AIR + AUDIT

6. Leave a fifth of every box empty

Fill to 640 and stop. A box crammed to the lid is not efficient, it is rigid: the next card you file pushes the row along, the overflow lands in the neighbouring box, and now two index lines are wrong instead of none. The empty fifth is the difference between filing one card and rebuilding a shelf — and you cannot pull a card cleanly from a packed row, which is how corners get bent.

7. Audit twice a year, on a clock

Two hours, twice a year. Count the rows, reconcile the index, and read the hygrometer while you are in there — keep relative humidity steady and inside the 35–55% band from the Library of Congress’s preservation FAQ. Twenty-four boxes at two minutes each is forty-eight of those minutes; the rest is fixing.

Then re-price Tier A and Tier S against what eBay calls sold listings — closed sales, not asking prices — because the tiers are defined by value and value moves. A card that has crossed a breakpoint needs a different holder and, sometimes, a different opinion of what the binder is worth. Boxes drift; two audits a year keeps the drift to a rounding error.