How do I move from a shoebox/paper system to digital without losing history?

What counts as "history" for your business?

History has three parts, and the plan is judged against all three:

• Source documents. IRS Publication 583 says purchases, sales, payroll and other transactions generate supporting documents, and that these support the entries in your books and on your tax return.
• The record made from them. This is your ledger, if you keep one, and every return and report already filed.
• The link between the two. The IRS's summary of Rev. Proc. 97-22, its procedure for electronically stored books and records (on its record-retention FAQ page for tax-exempt bonds), requires stored records to be cross-referenced with the books in a manner that provides an audit trail to the source documents.

If a ledger sits behind the paper, tie each batch to it as well as to the statements. If the paper is the only record, the bank and card statements are the only outside check and cash items have none beyond the count-in, so also tie to any other outside record you hold for the period, such as card-processor or payroll-service reports, and read every document that matches no statement line against its original, on top of the random sample.

In what order should the move happen?

Nearly every way of losing history here is a step taken too early. Run each batch through these phases in order:

1. Inventory. List every box and folder by period and document type, marking closed or reported periods and documents that still support something open.
2. Storage first. Set up the archive and its copies under the 3-2-1 rule described below, and the owner's own login, before the first scan. At least one copy must be a backup that keeps earlier versions or is kept offline, not a live sync of the archive.
3. Cut-over. Declare and write down the cut-over date; from it, new documents go to go-forward batches and the backlog stays in its own boxes and batch numbers.
4. Batch preparation. Take one period and one document type, remove staples, and write the number of documents and pages, counting every back that carries anything, on a batch sheet before scanning.
5. Capture and index. Scan the batch and enter each file's index fields in the same sitting.
6. Verification gate. Run the batch check described below, and fix and recheck a batch that fails; no original leaves your control before its batch passes. The one planned exception is outsourced scanning: boxes leave only under the signed handover list described under "What if someone else does the scanning?", and every original must be back and reconciled before the batch can pass.
7. Originals back in place. Return the paper to its labeled box, in order, where the index says it is.
8. Routine change last. Stop working from paper only after the first go-forward batches pass the same gate.

End a work session only at a batch boundary, and label any batch left part-way as unverified. Nothing in this sequence discards anything.

What does a scan have to be to stand in for the paper?

IRS Publication 583 says all requirements that apply to hard copy books and records also apply to electronic storage systems that maintain tax books and records. It requires the system to index, store, preserve, retrieve and reproduce the records in legible format. The IRS's FAQ summary of Rev. Proc. 97-22 adds an accurate and complete transfer of the hardcopy records, and hardcopy reproductions with a high degree of legibility and readability.

The National Archives' rule for federal agencies digitizing temporary records is a sound model for completeness: capture all information contained in the source records, and include all the pages or parts. A capture that can stand in for the paper passes three tests, however fine its thumbnail looks:

• It shows the whole document, with totals, signatures and stamps inside the frame.
• It includes every page and attachment, and the back of any page that carries anything, whether print, handwriting, a stamp or an endorsement.
• It holds exactly one document per file, and every letter and figure, including names, dates and descriptions as well as amounts, reads at normal zoom and on a printed copy without guessing.

Which labels does each file need?

Rev. Proc. 97-22 gives as an example of an index assigning each stored document a unique identification number with a separate database describing the records, and says the indexing requirement is met when the index is functionally comparable to a reasonable hardcopy filing system. Enter these fields at capture:

• ID. Give each file a unique number, such as batch plus sequence.
• Date, type, party and amount. Record the document's own date, what it is, who it is from or to, and its total.
• Period. Record the tax year and month; Publication 583 suggests organizing supporting documents by year and type of income or expense.
• Link. Note the ledger entry, statement line or return the document supports.
• Original's location. Note the box and batch where the paper sits.
• Status. Mark the period as open or closed and add any gap note.

How to name files and folders is covered in a separate guide.

When do new documents switch to digital, and what happens to the backlog?

Declare a cut-over date, such as the first of a month, write it down, and route every document by one rule:

Keep the backlog in its own boxes and batch numbers so a document sits in only one stream. While both streams run, capture nothing that is not on a batch sheet, and before accepting a batch, search the index for the same date, party and amount to catch doubles; treat a hit as a double only when both files show the same physical document (same number, marks and pages), and otherwise keep both and note the match in the index. Put the parallel run's end date in writing: an open-ended parallel run is how documents land in whichever system was nearer, or in neither.

How far back should you scan?

Each scope needs its own check and its own way of finding a document later:

Decide on these criteria:

• Period. Capture open periods first, and treat closed or reported periods as a separate series.
• Materiality. Capture documents behind large or unusual amounts before routine small receipts.
• Open obligations. Capture early anything still supporting something live, such as an asset you own, a loan, a dispute or a return that could still be questioned.
• Pending requests. Handle periods under an examination or a lender or insurer request as described further on.

Paper you do not capture stays the record for its period. How long records are kept is a separate question.

How do you prove nothing was lost?

A batch passes only when all five checks pass:

• Count in, count out. The documents and pages on the batch sheet, counted before scanning, equal the files and page images in the archive.
• Tie-out. Every line of the batch's own kind on the period's bank and card statements (money out for an expense batch, money in for an income batch), downloaded by the owner from the owner's own bank access, matches an indexed document, an item that never has paper (such as a bank fee) or a gap-log entry. Where a ledger exists, its entries for the period are tied the same way. A period's statements are fully cleared only when all its batches have passed.
• Every file opens. Each file opens and displays in full.
• Sample read. A random sample is read end to end against the paper, with every page and back present, every word and figure legible and the index fields correct, with one sampled document printed and read the same way.
• Restore. One file is restored from the off-site backup copy as it stood at an earlier date, and opened.

Start the tie-out from the statements, never from the scans: a list built from what was scanned cannot show what was never scanned. Where staffing allows, someone other than the scanner reads the sample; in a one-person business, read it on a different day, against the paper.

For sample size, the National Archives' rule for federal agencies digitizing permanent records is a useful yardstick. It allows a visual inspection of a random sample of a minimum of ten records or 10% of each batch, whichever is larger. Any failure in the sample means finding the cause, correcting the affected files and drawing a fresh sample.

What does a passing check look like on one batch?

Batch 2024-03-EXP holds March 2024 expense documents for a business whose paper is its only record:

• Counted going in. The batch sheet shows 64 documents and 81 pages, 6 with anything on the back, so 87 images are expected.
• What came back. There are 64 files but 85 images, because two backs were missed; rescanning those two documents gives 87.
• Tied against. The March bank and card statements carry 71 outgoing lines: 59 match indexed documents, 9 are bank fees and transfers between the business's own accounts, which never have paper, and 3 have no document. A search of neighboring boxes finds none of the 3, so all go into the gap log (59 + 9 + 3 = 71). The other 5 documents are cash purchases with no statement line (59 + 5 = 64); the count-in is their only check, so all 5 are also read against their originals.
• Readability. All 64 files open. Ten percent of 64 is 6.4, so ten documents are drawn at random and a second person reads each against its original, with one also printed and read the same way. One long receipt has its total cut off, so the cause is found (the scanner clips receipts longer than its page setting), all seven long receipts are rescanned, and a fresh sample of ten passes.
• Restore. One file restored from the off-site backup copy as it stood at an earlier date opens.

The batch sheet is then signed and dated, and the originals go back into box 2024-03 in order.

When can the originals and the paper routine go?

No original is shredded, returned, sent away or moved out of reach, and the paper routine does not stop, until its batch has passed. The one planned exception is outsourced scanning: boxes leave only under the signed handover list described under "What if someone else does the scanning?", and every original must be back and reconciled before the batch can pass. Rev. Proc. 97-22 shows the stakes: it says the IRS may issue a Notice of Inadequate Records where books and records exist only electronically and the storage system fails its requirements, while the penalties it describes may not apply if the taxpayer maintains its original books and records.

Passing the batch check shows that a batch came across; it is not the testing the IRS describes for discarding originals. Publication 583 allows the original hard copy books and records to be destroyed only where the electronic storage system has been tested to establish that they are being reproduced in compliance with IRS requirements for an electronic storage system, and procedures are established to ensure continued compliance with all applicable rules and regulations. It adds that you still have the responsibility of retaining any other books and records that are required to be retained. Whether to let paper go once those conditions are met, and whether to keep paper, digital or both, are separate questions.

How do you keep the archive from becoming a single point of failure?

Build these protections in before the first batch, so all history never sits in one account or on one device:

• Copies. CISA's undated Back Up Business Data page recommends the 3-2-1 rule: 3 copies of important files, on 2 different types of storage media, with 1 copy stored off-site, away from your business location. At least one copy must be a backup that keeps earlier versions or is kept offline, not a live sync of the archive; CISA lists offline copies among the protections for backups.
• A proven restore. The same page says, "Test backup procedure to make sure your team can rapidly restore data both fully and partially, and to ensure you can roll back data at least seven days if needed." Those seven days are CISA's roll-back interval for backups, not a records-retention period. It adds that a backup plan is only helpful if everyone knows how to use it.
• Lasting formats. The Library of Congress describes PDF/A's purpose as preserving a document's appearance over time, independent of the tools and systems used to create, store or render it. Rev. Proc. 97-22 treats stored records as destroyed if you stop maintaining the hardware and software needed to meet its conditions, unless they remain available to the IRS in conformity with it.
• A regular read. The IRS's FAQ summary of Rev. Proc. 97-22 requires regular inspections and evaluations of the system; the Library of Congress's personal-archiving page suggests checking at least once a year that files can still be read, a workable interval.
• Access beyond one person. Hold the archive in an account the business owns, with the owner's own administrator login; a bookkeeper is an added user, never the account holder. If there is no bookkeeper, name a second person, such as a partner, family member or adviser, add them as a user or second administrator on the account, keep the system description where they can find it, and check at each regular read that they can still get in. The IRS's FAQ summary of Rev. Proc. 97-22 requires you to maintain, and provide on request, a complete description of the system, including its procedures and indexing system, so write that description down where a second person can reach it.

How do you keep the move from changing figures already reported?

The move changes the medium, not the record. The IRS's FAQ summary of Rev. Proc. 97-22 requires controls that prevent and detect unauthorized alteration of stored records; leave each stored copy exactly as captured. Separately, this project changes no figures: re-enter, re-categorize or correct nothing, and leave the ledger and filed returns as they are.

Where part of the backlog covers closed or reported periods, capture them in their own batch series, separate from open-period work, and preserve their documents exactly as found, notes and stamps included. The figures they support are fixed, so a correction made while scanning would be a restatement.

If a document seems to show an error, log it with the document's ID and what looks wrong, finish the batch unchanged, and take the log to whoever keeps your books or prepares your returns as a separate matter. Catching up bookkeeping that was never recorded is a separate project too.

What do you do with damaged, illegible or missing documents?

Record every gap, so the digital set never implies a completeness it lacks. Keep a gap log inside the archive, backed up with it, that records:

• What is missing or damaged, and its batch
• What it relates to, from the statement line or ledger entry
• Where you searched
• What, if anything, stands in for it

Scan a damaged or faded original anyway at the best quality you can, flag it in the index, and keep the paper. Rev. Proc. 97-22 says records-management practices may include retaining hardcopies of books or records that are illegible or that cannot be accurately or completely transferred. Records lost in bulk have their own guide; what to do about a single receipt you cannot find is a separate question.

What if an examination, lender or insurer request is pending?

The moment one arrives or is expected, freeze the periods it covers: no boxing, shipping or scanning of them until it closes, and any of their batches already under way is either finished and checked or put back in its box in order. On the index-only path, produce the paper itself and capture the item after the request closes. The IRS's FAQ summary of Rev. Proc. 97-22 says that during an examination you must retrieve and reproduce hardcopies of the stored records requested and provide the resources necessary to locate, retrieve, read and reproduce them. Every document must stay producible at every point.

What if someone else does the scanning?

Rev. Proc. 97-22 says using a third party to provide the storage system does not relieve you of its responsibilities. Before any box leaves, count documents and pages per box on a handover list that you and the vendor both sign. Send no box for a period under, or expected to come under, an examination, lender or insurer request, and agree in writing how quickly the vendor will return any box you ask for. When boxes and images return, reconcile them box by box against that list, then run your own batch check on the output before accepting it. All originals come back to you, with no destruction by the vendor. The IRS's FAQ summary also says the system must not be subject, in whole or in part, to any agreement that would limit the IRS's access to it, so check the contract for that too.