Wedding Photo Backup: Why 3-2-1 Protects Your Archive and Not Your Wedding Day

The 3-2-1 backup rule, being three copies of your files on two kinds of media with one copy off-site, is worth following and it does not protect the thing wedding photographers are actually afraid of. Drive failure is rare and, against three copies, close to solved: Backblaze recorded 1,030 failures across 341,263 drives in the first quarter of 2026, an annualised failure rate of 1.24%. Weddings are not lost from a three-copy archive. They are lost in the hours between a card leaving the camera and a second copy existing, a window in which the number of copies is one and no backup rule has started applying yet.

Almost every backup article written for photographers is a description of 3-2-1 followed by a product recommendation. That leaves the interesting questions untouched. How likely is the failure you are defending against, really? Where does the rule stop applying? And if the archive is the safe part, why is everyone's anxiety pointed at it?

Where did the 3-2-1 rule actually come from?

From photography, which is a better piece of trivia than it first appears.

Backblaze, a storage company whose failure data carries the rest of this article, states in its own explainer on the rule that it is thought to have originated in the world of photography in Peter Krogh's 2009 book, The DAM Book: Digital Asset Management for Photographers. It defines the rule as three copies of your data on two different media with one copy stored off-site. From there it left the field entirely and became the baseline that enterprise IT and government cybersecurity guidance still point to.

So the situation is slightly absurd. A rule written for photographers, by a photographer, about photographs, is now taught back to photographers by storage vendors, several translations later, stripped of the working context it was written in. That context is the part worth recovering, because Krogh was describing what to do with an archive of images you already have. He was not describing a Saturday.

How often do drives actually fail?

This is the number the whole anxiety rests on, and almost nobody in this industry quotes it, despite it being published quarterly, for free, on the largest public dataset of its kind.

In its Q1 2026 Drive Stats report, Backblaze recorded 1,030 drive failures across 341,263 hard drives over 30,203,180 drive days, giving a quarterly annualised failure rate of 1.24%. Its lifetime figure across roughly 530 million cumulative drive days is 1.39%. Newer high-capacity drives ran better still, with the 20TB and larger pool at 0.85%.

Sit with that for a second, because it reframes the problem. Roughly one drive in eighty fails in a year. Now apply the rule: if a wedding exists on three independent copies, losing it requires three failures inside the same window, and independent events at that rate essentially do not coincide. The archive, correctly built, is not your exposure. It is the best-defended part of your entire operation.

The stretch the rule was never written to cover

Count the copies through an actual wedding day and the gap shows up immediately.

• Frame taken, single card, no second slot in use: one copy. Redundancy zero.
• Frame taken, dual card slots writing simultaneously: two copies, both in the same camera body, in the same room, exposed to the same theft, the same drink, the same bag left on the same chair. Two copies, one location.
• Cards in a pocket or a bag on the drive home: still two copies at most, moving together, in the same vehicle.
• Offload finished to one drive, cards not yet formatted: three copies, two locations. This is the first moment the day resembles 3-2-1 at all.
• Cards formatted for tomorrow before a second copy exists: back to one copy, and this is a decision people make on purpose, for convenience.

Everything above the fourth line happens before any backup rule engages. Krogh's rule assumes files that exist in an archive. On a wedding day the files exist in exactly one place for hours, and often in one building overnight, and the entire discipline that photographers have absorbed is aimed at the period after that. It is the equivalent of a very good lock on a door you have already carried the valuables past.

Note what this changes about where effort should go. Buying a fourth archive drive improves a number that is already extremely good. Changing what happens between 11pm and the following morning improves the number that is actually one.

It is one in the morning and the drive back took longer than it should have. The cards are in the front pocket of the bag, which is by the door, because carrying it upstairs felt like a decision you could make tomorrow. You know the offload takes about forty minutes and you know you are going to be useless for it, and both of those things are true. What you are actually doing, without framing it this way, is leaving the single existing copy of somebody's wedding in a bag by a door for eight hours, in a house, near a street. Nothing happens, almost every time. That is exactly why the habit survives.

What to do in the window

None of this requires new equipment, and most of it is a decision made once rather than a task performed weekly.

• Use both card slots when the body has them, and treat the second slot as a backup rather than as overflow capacity. Overflow gives you more room and no redundancy, which is the opposite trade to the one you want.
• Stop formatting cards on location. A full card that comes home unformatted is a real copy of the wedding. Buy more cards instead; the cost of another card is trivially less than the cost of the conversation you would otherwise be having.
• Separate the physical copies as early as you can. Two cards in one bag are two copies in one place, and most of the ways a wedding disappears are location events rather than media events: a stolen bag, a spilled drink, a car window.
• If you shoot with a second shooter, their cards are a single copy of a substantial part of the day, sitting in someone else's bag, subject to someone else's habits. Agree the handover and the offload rule before the day, which belongs in the same conversation as everything else in our piece on working with a second shooter.
• Offload before sleeping, or accept explicitly that you are not going to and decide where the cards sleep instead. The failure here is not laziness, it is leaving the decision unmade until you are too tired to make it.
• Verify the copy before the cards are reused. A finished progress bar is not the same claim as a readable file.

Are SSDs actually safer?

Less than the price difference implies, and the reason to buy one is not the reason usually given.

Backblaze has run the same comparison on its own fleet and found the gap smaller than the folklore: the annualised failure rates of SSDs and hard drives sat close enough that each fell within the other's 95% confidence interval, and SSD rates ran at or below roughly 1% through the first four years. On population statistics alone, paying several times more per terabyte for archive storage buys less reliability than most photographers assume it does.

The genuine argument for solid state in wedding work is a different one and it maps exactly onto the gap identified above. An SSD has no moving parts to damage while it is being carried, knocked, or run warm in a bag on a summer Saturday. That is an argument about the drive you take to the wedding and offload onto in a hotel room, not the drive that sits on a shelf. Buy it for the window, not for the archive.

Which is the shape of this whole topic, really. The archive question has a published answer and it is reassuring. The interesting risk sits earlier, in a few unglamorous hours nobody writes articles about, and it is addressed with habits rather than purchases. The other half of the problem, being how long any of this has to be kept once it is safe, turns out to be a legal question rather than a storage one, and it is covered in how long to keep wedding files. What you promised the couple about availability sits in your contract and your delivery terms. Two follow-ups read the fine print on the safety nets themselves: what cloud durability figures actually guarantee, and what the memory card warranty covers when a card fails. The one moment the archive itself is exposed comes every few years, when it moves to new drives and a normal copy carries silent corruption with it.