The cost of keeping everything

Keeping everything looks like the risk-free option. Storage is cheap, nothing is ever missing, and no decision has to be made. It is genuinely the right default for the first few years of a system.

It stops being free at a point that’s easy to miss, because the bill arrives as friction rather than as a charge. The cost of an over-full archive is paid entirely at retrieval, and it is paid by whoever is looking — usually you, usually under time pressure.

What volume actually costs

Search precision falls as volume rises. A search that returns four results is an answer. The same search across ten years returns sixty, of which fifty-two are dispatch notices, duplicates, drafts and things that merely mention the same vendor. The search worked. You still have to read sixty things.

Duplicates make you verify rather than read. Three copies of one invoice, in slightly different states, means every retrieval now includes a question: which of these is the one? That question is short and it recurs every single time.

Undecided items accumulate at the front. Anything you couldn’t classify sits where you’ll trip over it. Over years, the unclassifiable fraction becomes the layer you have to dig through to reach everything else.

The archive stops being reviewable. A batch you can assess as a unit takes seconds to make a retention decision about. A folder of four thousand unsorted items cannot be assessed at all, which means running a disposal review becomes impossible, which means the volume grows unchecked. The loop reinforces itself.

Migration cost scales with volume. Every move to new storage, every backup verification, every restore takes time proportional to the size of the thing. Most of that time is spent on material you would not have kept if anyone had ever asked.

None of these is dramatic. Together they are the difference between finding a receipt in ninety seconds and finding it in forty minutes, which is the difference between a system that gets used and one that gets avoided.

The part that isn’t about convenience

There is a second cost and it is categorically different.

Records contain personal information — yours, and other people’s. A contractor’s expense claim carries their card fragment and their travel. A customer’s order carries their address. Your own receipts carry a purchase history that describes your movements, health, household and habits in more detail than you’d volunteer. What a receipt reveals goes through the specifics.

Holding that data has consequences that grow with volume, and they run in the opposite direction from the “keep everything” instinct: the more you hold, the more there is to lose in a breach, the more there is to hand over in a dispute, and — depending on where you are and what you do — the more there is that you may have an obligation to not still be holding.

Whether such an obligation applies to you is a legal question that varies by jurisdiction and activity, and this site doesn’t answer it. What can be said generally is that data you never collected cannot leak, and data you disposed of on schedule cannot leak either.

Where the line sits

The useful question is not “what can I throw away” but “what is this copy for?” Applied honestly, it disposes of a surprising amount without touching anything that supports a record.

Three things are almost always worth keeping: the record itself, the evidence of payment where it isn’t on the record, and anything establishing the reason for the spend. That triad is the substance — it runs through what makes a receipt useful later.

Almost everything else in a typical archive is one of: a second copy of one of those three, a notification about the transaction rather than a record of it, or an intermediate state of something that was later finalised. Notifications and superseded drafts have no retrieval value and considerable noise value.

The exception, and it’s a real one: something whose purpose you can’t determine is not the same as something with no purpose. Records relating to a matter you’ve forgotten about, or supporting an asset you still own, look like noise and aren’t. This is why disposal works by class and trigger rather than by instinct, and why “anything live overrides the schedule” is the rule that stops a review destroying something.

Keep or bin

KEEP OR BIN — earning archive space

  · The record, the payment evidence, the
    reason
                    → KEEP. The substance. Never the
                      problem.

  · Anything relating to a matter that is
    still open
                    → KEEP. Looks like noise, isn't.

  · Second and third copies of one
    document
                    → BIN all but one. Duplicates
                      convert reading into verifying.

  · Dispatch, delivery and marketing mail
    about a purchase
                    → BIN. They match every search
                      for the real record and answer
                      none of them.

  · Superseded drafts and quotes where the
    final version exists
                    → BIN the drafts, keep the final —
                      unless the quote itself is the
                      record of what was agreed.

  · An unnamed folder of images from four
    years ago
                    → TRIAGE, don't keep and don't
                      bin. Newest first, an hour at a
                      time.

  · Data about other people, past the point
    of needing it
                    → ASK LOCALLY. Retention and
                      minimisation obligations vary by
                      jurisdiction and activity.

Do less, not none

The failure mode of this argument is over-correction: someone reads it, resolves to be ruthless, and deletes something that mattered. That is a much worse outcome than a cluttered archive, and it’s worth being explicit that the recommendation is not asceticism.

Practically: be generous with the substance and ruthless with the noise. Duplicates, notifications and superseded drafts can go on sight, at any time, with no scheme and no thought — they are the majority of the volume and none of the value. Anything that might be a record stays until a scheduled review considers it as part of a class with a trigger, and anything live stays regardless.

That split gives you most of the benefit with none of the risk, because the ruthless half of it never touches a document that could be a record.

What this doesn’t settle

Whether you are permitted to dispose of anything, whether you are required to, what the periods are, and whether obligations differ for records about other people. Whether a copy may be kept once an original goes, or the reverse.

Those come from your tax authority, an adviser, or the data-protection regime where you are, and they vary too much for a number here to be useful — how long to keep things covers building a scheme around whatever they turn out to be. What this page settles is the part no rule addresses: an archive is a retrieval tool, and every item in it that answers no question makes the ones that do harder to reach.