Knowing how to retire legacy accounting software after a migration depends on one question: will you still need to answer from the old books? If audit, tax or lender questions can only be answered inside the old system, keep it read-only for a defined period. If a verified export covers those questions, export, archive and cancel. Telling the two cases apart takes a few checks, and closing the old system without losing history takes a few steps.
Why the old books still matter after the cutover
A migration moves the daily work to a new system, but it does not move every question. The old accounting software still holds the history: invoices sent before the cutover, fractional CFO companies in Miami bills paid through the previous bank feed, journal entries from the last weeks of parallel work, and the attachments someone scanned years ago. Once the new system is live, those records stop changing, which is exactly why they tend to be forgotten.
Many owners treat the old subscription as a line item to cancel as soon as the new system works. That instinct is understandable, but it skips the real question. Whether the old system can go depends on what the business may still need to prove from it. Two companies can finish a migration on the same weekend and reach different answers, because one has a lender expecting prior-period statements and the other only needs its last tax return. The same action, cancelling, is safe for one and risky for the other.
Miami-based fractional CFO and interim CFO work, controller support and outsourced bookkeeping all rest on one base: a ledger that can be trusted. When books move from one system to another, the finance function inherits whatever the cutover left behind. Clean opening balances, reconciled bank accounts and a clear map of where each piece of history lives make later reporting, budgeting and cash forecasting far easier. The narrower question here is what happens to the old software once the new one carries the work.
The risk is rarely dramatic at the start. It usually appears months later, as a missing invoice during an audit request, a bank reconciliation that cannot be rebuilt, or a buyer's accountant asking for a ledger nobody can open. Read-only access is one way to prevent those problems, but it is not free. It means another login, another renewal and another person who needs to know where things are kept.
The decision therefore comes down to cost against exposure. Keeping a legacy system read-only costs a subscription or hosting fee and some staff attention. Shutting it down without a plan can cost much more if a question arrives after access is gone and the data can only be recovered through a support request. Those two costs are rarely equal, and the difference usually depends on questions you can list before anything is cancelled.
Which questions the old system still has to answer
Before choosing between read-only access and a full export, list the questions the old books might need to answer. Each one needs a different kind of record, and each one makes the export test harder or easier. The table below is a starting point, not a checklist that fits every company.
QuestionWhat it needsUsually enough from a clean export?
Audit or review requestTransaction detail, support documents and approval trailOften, if attachments and approvals come along Tax or sales tax inquiryFiled returns, tax codes and rates in force at the timeOften, if tax codes are kept on each transaction Lender covenant or compliance certificateMonthly statements as previously reported, including later adjustmentsOnly if prior versions can be reproduced Buyer due diligenceMulti-year ledger, reconciliations and revenue schedulesDepends on the reporting the old system can still run Revenue recognition review under ASC 606Contract-level schedules and deferred revenue historyOften needs the original reports, not just raw data Payroll correctionPay runs, withholdings and employer filingsUsually answered from payroll provider records
The middle column shows what each question needs, and the last column is where most mistakes happen. An export gives you data. A report gives you an answer. The two differ when someone asks for the balance sheet as it was reported in March, after several adjustments were posted later. A general ledger file can be rebuilt into that balance sheet, but the rebuild is slow and error-prone unless the export keeps posting dates, entry history and reversal records.
Consider a hypothetical company with two entities, one bank account each and a lender that requires monthly statements. Its new system holds the current balances correctly. Its old system holds the statements that were actually sent to the lender, including a reclassification made in the third month. If the old system goes away, the company can still show its current position, but it cannot prove what the lender saw. That gap is the one to test for.
Tax questions tend to be more predictable, since filed returns are fixed documents and tax codes can often be preserved on each transaction. Still, the tax preparer should confirm which records they need and in what format. Revenue questions are harder. A review of revenue recognition under ASC 606 often depends on contract-level schedules built inside the old system, and a raw export may not carry the logic that produced them.
The practical lesson is to match each question to a specific record before you cancel. Ask the person responsible for each question what they would need to see, and whether a file or a rebuilt report would satisfy them. Write the answers down. A question nobody can name is the one most likely to show up later, when the old access is no longer available.
When keeping read-only access is worth the cost
Read-only access makes sense when the old system can still answer questions that an export would answer poorly. The most common case is a need to reproduce a report exactly as it was produced at the time. A lender, auditor or buyer may ask for a figure that changed after later postings, and only the original system can show the version that was reported. In that situation, a copy of the data is not the same thing as the old report.
It is also worth keeping access when the export loses information that matters. Some systems keep attachments in a way that does not travel with the data, some store approval trails separately, and some tie custom fields or user-defined tax codes to records in a format the new system cannot read. If the export is missing any of these, read-only access may be the only way to show where a number came from.
Another case is a partial migration. Some companies move the core ledger but leave behind a legacy module, such as a project costing tool, an inventory add-on or a separate entity that was never consolidated. In that situation the old system is still part of the reporting chain, and shutting it down would break a link nobody has documented. Read-only access keeps that chain intact while the team decides what to rebuild.
These signs point toward keeping read-only access:
A lender, auditor or buyer may ask for prior reports that only the old system can reproduce. The export test loses attachments, approvals or tax codes. The company has more than one entity, and some of them are still linked to the old system. A sale, financing round or audit is expected within the period the old data must cover. The team has not yet agreed who owns the archive and who answers questions about it.
The cost is real, so the decision should be explicit. Write down the subscription or hosting fee, the logins that must stay active and the person responsible for them. If those costs are small compared with the exposure described above, read-only access is usually the sensible choice.
When a verified export is enough without read-only access
A verified export is often enough when the business is simple and the history is short. A single-entity company with one bank account, a clean chart of accounts and a new system that received correct opening balances may not need the old software at all. In that case the export is the archive, and keeping a login alive adds cost without adding much protection.
The word verified matters. An export that nobody has opened is a hope, not an archive. Verification means reconciling the exported trial balance to the last closed period in the old system, opening a sample of attachments, and confirming that the file can be read without the original software. If those checks pass and the same person can explain them to an outside reviewer, the export has done its job.
Consider a hypothetical service firm with one entity, one operating account and no inventory. Its last audit was clean, its lender requires only annual statements, and any sale is not yet planned. The owner exports the full general ledger, the customer and vendor lists, the reconciliations and the attachments, then tests the files on a second computer. Read-only access would add little in that case.
Even then, the decision should not rest on feeling. Ask whether any question in the earlier table would be answered poorly by the export. Ask whether the new system's opening balances match the old closing balances to the cent. Ask whether anyone at the company could explain the export to an outside accountant in a single meeting. If the answers are yes, the export is probably enough. If one is no, the case for read-only access returns.
A quieter reason to avoid unnecessary access is security. Every legacy login is a place where someone could change or delete records, and a retired system that still accepts passwords can become an access point nobody is watching. Closing a system that no longer serves a function reduces that exposure, provided the archive is kept somewhere controlled.
Setting the read-only period without guessing
The read-only period should end when the questions that need the old data are closed, not when the team gets tired of paying for it. That sounds obvious, but many companies set the date by feel, usually at the end of a subscription cycle or when someone remembers to ask. A better approach is to name the events that would still need the old data, then keep access until each one is resolved.
Those events vary by company. A tax preparer may need access until the relevant filing period is closed and any follow-up notice has been resolved. A lender may need records until the reporting obligations under the loan agreement are satisfied. A buyer's counsel may need history for a period set in the purchase process. An auditor may need records through the end of the engagement and any follow-up. Each professional can tell you which of these applies to you. None of them should be guessed.
Here is how it can look in a hypothetical case. A company has a line of credit with reporting requirements, and a tax filing that depends on prior-year records. The team writes two triggers: the last monthly statement due under the agreement, and the close of the tax filing cycle plus any notice that follows. Until both are met, the old system stays read-only. Once both are met, the team reviews whether the archive still needs live access.
A period set this way has an owner. Someone must watch the triggers, remind the others when a trigger is met, and sign off on the shutdown. Without an owner, the license tends to renew automatically and nobody remembers why it exists. Writing the owner's name next to each trigger is a small step that prevents a common problem.
It also helps to decide what happens if a trigger moves. A sale or financing round may extend the period, and an unexpected inquiry may reopen it. The plan should allow for both, so the company does not have to reconstruct the reasoning in the middle of a deal or a review.
Freezing the legacy system in the right order
Once the decision is made, the order of steps matters more than speed. Closing the old system before the export is verified, or cancelling a subscription before access to the archive is confirmed, recreates the problems described above. The sequence below works for many companies, though the details depend on the software and on the records involved.
Confirm that the new system's opening balances match the last reconciled close in the old system, account by account. Export the full data set and the standard reports, and save each file with its date and the period it covers. Lock the closed periods in the old system so that no new postings can change them, and record who performed the lock. Remove or limit user logins, keeping only named read-only accounts for the people who need them. Document where each archive lives, who holds the credentials and how a request for information would be handled. Cancel or downgrade the subscription only after a test restore has passed and the documentation is complete.
Step one is the one most often skipped. Owners compare the new system's current balances to the old system's last report, which is useful but not enough. The comparison has to cover the closed periods, the accounts that move less often, such as fixed assets and accrued liabilities, and any reconciling items carried forward. A small difference is worth tracing before it becomes an audit question.
Step three deserves a separate check. Many systems allow a period to be closed while still letting an administrator reopen it, and that permission can outlast the migration. Confirm that the lock is in place, that the permission is removed from users who do not need it, and that the record of who changed the lock is preserved. A lock that anyone can open is not much of a lock.
Steps four and five protect against two different failures. Limiting logins stops accidental edits, but it does not stop a former employee or a shared password from reaching the system. Documenting the archive protects against the situation where the only person who knew where things were kept has left, changed roles or simply forgotten. Both are easy to skip and both are hard to fix once the old system is gone.
The sixth step is the point of no return. Before cancelling, confirm the vendor's rules for data after the subscription ends, since each vendor sets its own terms for how long data stays available and how it can be retrieved. Those terms are in the contract and in the vendor's support documentation. Read them, do not rely on memory of a sales call, and keep a copy of the answer with the archive records.
Checking the archive before you cancel the subscription
An archive is only useful if it can be read, understood and tied back to the books. A test restore is the simplest way to find out whether it works. Choose a closed period, rebuild the trial balance and the main financial statements from the export, and compare them to the figures that were reported at the time. If they match, the archive is probably sound. If they do not, the gap should be explained before the old system goes away.
The test should also cover the records that exports often leave out. Open a sample of attachments and check that they display correctly. Look for voided, deleted or reversed entries, since those are often needed to explain why a balance changed. Confirm that approval records and user activity logs are included, or at least that they exist somewhere else. Check the dates too. An export that shifts entries by a day because of a time zone or date format setting can cause small but confusing differences.
File format deserves attention. Systems such as QuickBooks, NetSuite and Sage Intacct each produce exports and native backups in different ways, so check what your version actually delivers. A file that only the original software can open is a weak archive. Common formats such as spreadsheets and plain text are easy to read later but may flatten relationships between records. A reasonable approach is to keep both: a readable export for review and the native backup, if the vendor provides one, for full fidelity. Ask whether the native backup can be restored without the subscription, and under what conditions.
Here is a short example. A hypothetical company exports its ledger as spreadsheets and rebuilds March's balance sheet. The totals match, but one accrued liability is off by a reversing entry posted in April and tagged to March. The team traces the entry, adds a note to the archive and keeps the native backup, because the tagging cannot be reproduced in the spreadsheet. The extra step costs an afternoon, and it removes a question that would otherwise come up during the next review.
Document the test itself. Record the date, the person who ran it, the periods tested, the differences found and how each was resolved. That record is part of the archive. If someone asks later whether the old books were checked before the subscription ended, the answer should be a document, not a memory.
What to confirm with your accountant or lawyer before you decide
Some of the choices in this process depend on rules, contracts and plans that only your advisors can see in full. The questions below are the ones to bring to them before you cancel anything. They are not legal or tax advice, and the answers will change with the company's structure, industry and location.
Which records must be kept, for how long and in what form, given the tax filings and entity types involved? Does the archive need to stay in its native system, or is a verified export acceptable to the auditor or lender? Which lender, investor or buyer obligations could require access to prior-period reports? Who in the company will hold the archive, the credentials and the authority to answer requests? What do the vendor's contract and support documentation say about data retention and retrieval after the subscription ends? If a sale, financing or audit is expected soon, what changes to the read-only period should be planned now?
Bring the answers to a single meeting if you can. Ask the accountant to confirm which questions in the earlier table apply to the company. Ask the lawyer to review any contract clause that mentions records, data access or retention. Ask the person who will run the archive to explain the test restore in plain language. If someone cannot explain it, the archive is not ready.
A reasonable outcome is a short written decision: read-only access kept until a named trigger, or a verified export archived with the test results and a named owner. Either can be correct. What matters is that the decision was made on purpose, with the questions answered, and that the person who made it can be found when a question arrives.
Expect the decision to be revisited. Companies change, and a period that made sense before a financing round may not make sense after it. Put a date on the review, even a loose one, and tie it to events rather than to the calendar alone. A sale process, a new lender or a change in tax filing obligations should each trigger a second look.