Journal entries in accounting: the definitive guide

A journal entry can balance perfectly and still be wrong.

An auditor pulls a sample of 25 journal entries from the prior quarter. Entry number 14 is a $340,000 reclassification between cost of goods sold and operating expenses. The description reads "reclass per controller." There is no support attached. No memo explaining the business rationale. No evidence of who reviewed it or when. The preparer left the company two months ago. Nobody on the current team can explain why the entry was made, whether it should have been reversed, or whether the underlying issue it was meant to correct has since been resolved. That single entry, which took seven seconds to post and had no support behind it, is now an audit finding.

This is what makes journal entries one of the highest-leverage control points in accounting. Every reconciliation, every variance explanation, every line on the financial statements traces back to a journal entry. When the entry is correct, supported, and approved by the right person, the rest of the close holds together. When it is not, the problems compound into every downstream process.

What a journal entry actually is, and why most definitions are wrong

A journal entry is the formal record of a financial transaction in an organization’s accounting system. It captures the accounts affected, the amounts debited and credited, the date, a description of the business purpose behind the transaction, and the evidence that supports it. Under double-entry accounting, every journal entry must balance: total debits must equal total credits. ASC 105 matters for a different reason: it frames the FASB Accounting Standards Codification as the source of authoritative GAAP, which is why the accounting conclusion behind the entry must be supportable.

Most definitions stop there. They describe the syntax of recording transactions but miss the control dimension entirely.

A journal entry is not a bookkeeping note. It is part of the formal process by which financial statements are prepared in conformity with GAAP. It is a statement about economic reality: this expense belongs in this period, this liability existed at month-end, this balance sheet account was misstated and required correction, this cost belongs in COGS rather than SG&A. Every one of those claims needs to be supported, reviewable, and approved by someone with the authority to challenge it.

The debits and credits are the syntax. The rationale, support, policy basis, approval, and evidence trail are what make the entry defensible.

That is why "data entry" is the wrong mental model. A journal entry is a controlled accounting assertion. Under SOX Section 404, PCAOB AS 2201, and the COSO Internal Control Framework, the approval and documentation around a journal entry are not administrative steps. They are control activities that support reliable financial reporting. The entry is not complete when it balances. It is complete when the accounting conclusion, support, and approval trail can stand on their own

The anatomy of a complete journal entry

A complete journal entry has two layers: the record itself and the packet around it. The record contains the visible mechanics. The packet is what makes the entry reviewable.

Controllers who have been through a SOX audit or a close that went sideways know that every component below creates a specific problem when it is missing.

Beyond the entry record itself, the surrounding documentation (evidence, approval trail, and policy basis) is what makes it reviewable. The JE packet section below defines that standard.

The description field does more work than most teams give it credit for. "Reclass per controller" is not a description. "Reclassify $42,000 of March freight-in costs from SG&A (account 630030) to COGS (account 500100) per direct fulfillment cost policy. Original AP vendor mapping used incorrect default account. See attached policy memo and original invoice" is a description. The difference is whether the next person who opens the entry can understand it without calling anyone.

The types of journal entries you will actually book

Not every journal entry carries the same risk. A recurring payroll summary is a different kind of problem from a top-side correction posted late in close. Teams that treat them the same usually end up over-reviewing low-risk entries and under-reviewing high-risk ones.

A useful controller question is not "what kind of JE is this?" but "what risk does this JE create, and what level of support and approval should follow from that?" The approval matrix in a later section formalizes this distinction.

Three journal entries you would actually see at close

Example 1: unbilled revenue accrual with reversal

A SaaS company completed a $90,000 implementation project for a customer in March. The customer confirmed completion on March 28, but the invoice will not be sent until April 5. Revenue belongs in March under the accrual basis.

March 31 accrual entry:

Memo: "Accrue $90,000 implementation revenue for [Customer]. SOW #2024-0147, all deliverables completed 3/28. Invoice to be issued 4/5. Reverses 4/1."

April 1 reversing entry:

When the invoice posts through AR on April 5, it debits Accounts Receivable and credits Service Revenue. The reversal prevents double-counting. Without it, March and April both show $90,000 of revenue for the same project.

The mechanics are simple. The lifecycle control is where teams fail. If the reversal does not happen, April double-counts revenue. If it reverses twice, April understates revenue. If it reverses in the wrong period, the financials are misstated in both months.

Evidence: Statement of work with completion date, customer completion confirmation, revenue recognition policy memo confirming point-in-time recognition for implementation services, billing schedule showing April 5 invoice date.

Example 2: expense reclassification from SG&A to COGS

Reviewing operating expenses during the March close, the controller spots $42,000 of packaging materials coded to Marketing Expense (630020) by the AP team. These costs are direct materials and belong in Cost of Goods Sold (500100). The expense was recorded correctly in AP, but the account mapping on the vendor record defaulted to marketing.

March 31 reclassification entry:

Memo: "Reclassify $42,000 of packaging materials from Marketing Expense to COGS. Vendor: PackageCo, PO #3847. Original AP coding used default vendor account mapping. Corrected per updated cost allocation policy. No reversal needed. Vendor record updated to prevent recurrence."

This entry is net-zero to the income statement total but moves $42,000 between line items. Gross margin decreases and operating expenses decrease by the same amount. A reviewer who only checks that total expenses are unchanged will miss this entirely. Reclassifications need the same level of support as any other entry.

Evidence: Original AP invoice, purchase order, updated vendor record showing corrected account mapping, controller approval of reclassification.

Example 3: prior-period error correction (ASC 250)

During Q2 close, the controller discovers $175,000 of SaaS implementation labor capitalized as contract fulfillment assets over the prior two quarters, Q4 of the prior year and Q1 of the current year. After reviewing the company’s capitalization policy and ASC 340-40 guidance, the controller concludes the costs should have been expensed as incurred rather than capitalized.

The first question is materiality. Is $175,000 material to the financial statements? For a company with $80 million in annual revenue and $11.7 million in operating income, $175,000 represents approximately 0.2% of revenue and roughly 1.5% of operating income. Under a purely quantitative assessment, this may fall below typical materiality thresholds. But SEC Staff Accounting Bulletin No. 99 (SAB 99) makes clear that quantitative thresholds alone are not sufficient. The controller considers qualitative factors: the error affects gross margin (a metric the board tracks closely), it spans two reporting periods, and it involves a judgment area auditors are likely to examine closely. The controller concludes the error is not material to the prior periods and that correcting it in the current quarter would not materially misstate current-period results. The company records an out-of-period correction rather than restating previously issued financial statements.

June 30 correcting entry:

Memo: "Record out-of-period correction for prior-period capitalization error. $175,000 of SaaS implementation labor incorrectly capitalized as contract fulfillment assets in Q4 FY25 ($95,000) and Q1 FY26 ($80,000). Management concluded the costs should have been expensed under company policy and ASC 340-40 guidance. Immaterial to prior periods individually (SAB 99 analysis attached). Corrected as out-of-period adjustment per ASC 250. No restatement required. See attached materiality analysis and policy memo."

Evidence: Materiality analysis (quantitative and qualitative per SAB 99), ASC 340 policy memo with analysis of capitalization criteria, detail of costs by quarter, controller sign-off, discussion with external auditors documented.

Journal entries are not a mechanical exercise. The correcting entry itself is two lines. The judgment behind it, whether to restate or correct in the current period, requires understanding ASC 250, SAB 99, and the company’s specific facts and circumstances. That judgment, and the evidence trail supporting it, is what separates a defensible correction from an audit finding.

The JE packet minimum standard

The anatomy section covered what belongs on the journal entry record. The packet standard covers what makes that record ready for review.

Most JE review failures are not caused by wrong numbers. They are caused by incomplete packets that force the reviewer to chase evidence after the fact.

Every non-trivial journal entry packet should include the following elements before it is submitted for review:

• Business rationale. Why does this entry exist? Is it driven by an invoice, a contract, a schedule, a policy, an estimate, or a management decision? The rationale connects the entry to the underlying economic event. "Record accrual" is not a rationale. "Accrue $90,000 implementation revenue for [Customer], SOW #2024-0147 completed 3/28, reverses 4/1" is.

• Source evidence attached. The supporting document itself, not just a reference to it. Invoice, contract, calculation workbook, email approval, vendor statement, payroll register, or source-system report. Attach it to the entry in the ERP or close management system. If the evidence lives only on someone’s desktop, it does not exist for audit purposes.

• Tie-out to schedule or calculation. The entry should reference the specific document, schedule, or calculation that produces the amount. "Per amortization schedule v3, tab Q1, row 14" is a tie-out. "Per schedule" is not.

• Policy basis or accounting conclusion. For entries involving judgment, such as accruals, reclassifications, or corrections, the packet should state the accounting policy or standard that supports the treatment. This is especially important for entries that touch ASC 250, ASC 340, or revenue recognition.

• Preparer and reviewer/approver trail. Who created the entry and who reviewed and approved it for posting. Under PCAOB AS 2201, auditors evaluate whether the control operated at the right point in the process.

• Review timestamp before posting. Timestamps matter because they demonstrate that the review happened before the entry posted, not after. A pre-posting review is a preventive control. A post-hoc review is a detective control. In SOX 404 and COSO-style control design, the question is not only whether the journal entry posted correctly. It is whether the evidence shows the control operated: who reviewed it, what they reviewed, when they reviewed it, and whether exceptions were addressed.

• Reversal ownership if applicable. If the entry reverses, when does it reverse? Is the reversal automatic or manual? If manual, who is responsible for posting it? Undocumented reversal rules are how accruals get double-counted and how "temporary" reclassifications become permanent fixtures in the trial balance.

• Exception or root-cause note if applicable. If the entry corrects an error, fixes an upstream coding issue, or overrides a standard process, the packet should document what went wrong and what preventive steps will avoid recurrence.

The JE approval matrix

Journal entry approval is a sequence: the preparer submits the entry with evidence attached, the workflow routes it to an approver matched to its risk level, the approver evaluates the accounting conclusion and the support, and only approved entries post to the general ledger.

Journal entries carry different levels of risk. A $500 office supply entry and a $2 million top-side consolidation adjustment do not warrant the same scrutiny. The approval matrix below provides a concrete framework that a mid-market controller can adapt. The specific triggers will vary by company size and risk appetite, but the principle holds: match the level of review to the level of risk.

Two principles matter more than the specific triggers:

• First, the approver should never be the preparer. That is the foundational segregation of duties control for journal entries, and it applies regardless of amount.

• Second, approval should happen before the entry posts, not after. A post-hoc review is a detective control. A pre-posting approval is a preventive control.

A useful calibration check: if nearly every entry routes to the controller, the triggers are probably too tight. If almost nothing routes to the controller, they may be too loose for the company’s risk profile. Review the matrix periodically and calibrate with the risk profile of the close

How this works in NetSuite

For teams running on NetSuite, journal entry workflow is built into the ERP, but the native behaviors and limitations are worth understanding before designing a control process around them.

NetSuite provides two approaches to journal entry approval. The basic approach uses the "Require Approvals on Journal Entries" accounting preference, which adds an Approved checkbox to each journal entry record. Users with Journal Approval permission must check this box before the entry posts to the general ledger. The more structured approach uses SuiteFlow or custom approval routing to direct entries to specific approvers by amount, subsidiary, or other criteria.

Both approaches enforce the core control: unapproved entries do not post. But the default configuration does not prevent self-approval, so teams should verify that the workflow explicitly routes entries to a different approver than the preparer.

NetSuite supports memorized transactions for recurring journal entries. These allow teams to create templates for entries that repeat on a predictable schedule, such as monthly depreciation, rent accruals, or standard allocations. However, teams should verify that memorized transactions created before enabling the Approval Routing feature are governed by the new workflow. After workflow changes, teams should confirm that existing memorized templates still flow through the intended approval process; older templates may need review or re-creation.

For teams approving high JE volumes during close, approval queues can still become a bottleneck. Routine months are manageable. Quarter-end and year-end add real friction.

Once a journal entry is approved and posted, many teams avoid editing the original record and instead use a controlled reversal and replacement process. If an error is discovered after approval, correcting it typically requires posting a separate reversing entry and then resubmitting a new corrected entry through the approval workflow. This is the correct accounting treatment, but it creates operational overhead. Teams that catch errors before approval save significant rework.

For intercompany journal entries, NetSuite OneWorld creates mirror postings across subsidiaries. Depending on configuration, approval may apply to each subsidiary independently, which means the same entry may be approved on one side and pending on the other. Controllers managing multi-entity closes should track approval status across the group, as the reconciliation cannot be finalized until both sides are approved.

In practice, many NetSuite teams still supplement the native workflow with spreadsheets or external tools because the preparation layer (support assembly, rationale documentation, transformation logic) sits outside the journal record. The posting step is system-native. Everything before it often is not.

Where teams get stuck

The concept of a journal entry is straightforward. The operational problems that consume time during close are not. They arise when entries are reviewed under pressure, when documentation standards slip, and when patterns that started as temporary fixes become permanent fixtures.

• Net-zero entries that hide real mistakes. A $150,000 reclassification entry nets to zero on the trial balance. Total expenses are unchanged. Total assets are unchanged. In a quick review, the entry looks clean. But the reclassification moved costs from one segment to another, changing the gross margin on the segment the board reviews most closely. Net-zero entries receive less scrutiny because they do not change totals, which is precisely why they are the most common vehicle for misclassification errors.

• Copy-paste accruals that never get revisited. A $65,000 monthly accrual for legal fees was set up 14 months ago based on a litigation estimate from outside counsel. The litigation settled eight months ago for $30,000. The accrual still posts every month at $65,000 because it is a recurring entry and nobody has reviewed the underlying assumption. The settlement month carried one final legitimate posting, but the seven months since then have added $455,000 of unnecessary accrual to the balance sheet. Without a defined expiration or reassessment date, recurring entries become the most dangerous items in the ledger because they operate on autopilot.

• "Temporary" reclassifications that become permanent. A controller posts a reclassification in January to move $200,000 of hosting costs from Engineering to Cost of Revenue while the cost allocation methodology is being finalized. The reclassification carries forward into February, then March. By Q2, the "temporary" classification has become the de facto treatment. When the final methodology is approved in Q3, it differs from the reclassification by $45,000 per month. Unwinding five months of incorrect classifications takes longer than the original entry took to post.

• Missing support that surfaces during audit. The most expensive version of incomplete documentation is not the entry posted without support. It is the entry posted with support that no one can find. The preparer saved the calculation locally. The approver reviewed it on screen but did not attach it. The invoice was in the AP inbox but never linked to the manual JE. When the auditor requests support six months later, the team spends two hours reconstructing evidence for an entry that took five minutes to post.

• Approvers who review without re-performing. An approver who clicks "Approve" without inspecting the support is not performing a control. They are performing a ceremony. For the approval to function as a control, the reviewer needs to verify the amount against the support, confirm the account mapping, check the period, and assess whether the entry is consistent with prior periods. For high-risk entries, the reviewer should be able to re-perform the calculation independently. The distinction between approval as ceremony and approval as control is what auditors evaluate under PCAOB AS 2201.

What changes at month-end

Journal entry work during the month is relatively steady. A few standard entries, an occasional reclassification, the normal flow of business. At month-end, the volume spikes and the nature of the work changes.

The volume problem is real but predictable. A company that posts 30 to 40 journal entries during the month may post 150 to 200 during the close window. Accruals, deferrals, depreciation runs, prepaid amortizations, intercompany eliminations, and correcting entries all converge into the same 72- to 96-hour period. The review queue fills faster than designated reviewers can clear it.

The quality problem is less obvious but more dangerous. Under time pressure, descriptions get shorter, support gets attached later, and reviewers approve batches instead of individual entries. The entries that receive the least scrutiny are often the ones that carry the most risk: late accruals booked on the last day, top-side adjustments made after the preliminary trial balance is pulled, and correcting entries created under pressure to make the numbers tie. They deserve more review, not less.

The interdependency problem compounds both. Reconciliation may surface an unrecorded liability that requires an accrual. That accrual changes the trial balance, which changes the variance analysis, which may trigger a new explanation. A correcting entry posted at 4 PM on day three of close can invalidate flux commentary written at 9 AM the same day. The close is a network of dependencies, and journal entries are the mechanism through which every adjustment flows.

The solution is not more reviewers. It is fewer entries that require manual assembly in the first place.

What agentic AI changes about journal entries

What AI changes is not the accounting principle. It is the preparation burden around the entry.

The most time-consuming part of journal entry work is not posting the entry. It is everything that happens before the entry is ready to post: gathering the source data from an upstream system, formatting it into the structure the ERP expects, determining the correct accounts and amounts, writing a description that explains the transaction, attaching the evidence, and routing it to the right approver. For a single entry, this takes minutes. For 150 entries during a close window, this is where the days go.

This is where platforms like Maxima change the workflow. Preparation shifts from the accountant to an agent, and the accountant’s role shifts from assembling entries to evaluating the proposed accounting.

• For recurring, rule-driven entries like bank cash activity, payroll, and benefits, a controller configures matching rules that map transaction types from source data to journal entry templates. New transactions arrive from bank feeds or payroll providers, and the platform applies the rules, populates the header and line detail, writes the description, and routes the drafted entry for approval. Where source data needs transformation, the platform handles that too. Employee-level payroll data can be aggregated to the department level before posting, keeping sensitive compensation detail out of the GL while producing the summary entry the ledger requires. Computed columns can derive amounts the source system does not provide directly, such as a PTO accrual calculated from balance hours and annual salary. The entries arrive pre-prepared with source data attached. The controller evaluates the output, not the input.

• For complex entries that require judgment, like stock-based compensation from Carta or multi-step accrual calculations, a controller can instruct the agent in plain language, the same way they would brief a junior accountant. The agent reads the source data, builds the intermediate calculations, produces the final figures, and feeds them into the journal entry workflow. Before executing, the agent presents its plan and waits for approval. Every step is logged.

• For structured close tasks with documented procedures, the agent reads the standard operating procedures attached to each checklist task and executes the workflow: run the journal entry automation, run the reconciliation, prepare manual entries for any unreconciled items. The controller opens prepared work with the source data used, rules applied, calculations performed, and exceptions surfaced.

In all three modes, the control pattern is clear: the agent prepares, and the human evaluates, challenges, and approves. In a controlled setup, proposed entries should not post without approval, and the human does not spend time on data gathering, formatting, or description writing to perform an informed review. Preparation and approval are separated not by policy alone but by architecture. The audit trail captures every action the agent took, every source it referenced, and every rule it applied. Evidence is assembled as the work is performed, not reconstructed after the fact.

The goal is not to remove accountants from the journal entry process. It is to remove the data assembly, the formatting, and the manual preparation so accountants spend their time on what actually requires judgment: evaluating whether the entry is correct, whether the support is sufficient, and whether the financial statements tell an accurate story. Accountants do not stop owning the accounting. They stop owning the spreadsheet assembly work that too often gets mistaken for accounting.

What remains human-owned is exactly what should remain human-owned: policy judgment, materiality assessment, approval decisions, non-routine exception review, error escalation, and final sign-off. That is what agent-prepared, human-reviewed means in accounting practice.

The goal is not simply faster posting. It is stronger evidence, cleaner packets, and entries that still hold up when someone other than the preparer opens the file. That is the difference between data entry and accounting.

See how Maxima helps accounting teams prepare entries that arrive ready for review.

Frequently asked questions

What is a journal entry in accounting?

A journal entry is the formal record of a financial transaction in an organization’s accounting system. It captures the accounts affected, the amounts debited and credited, the date, a description of the business purpose behind the transaction, and the evidence that supports it. Under double-entry accounting, every journal entry must balance, but balance alone does not make the entry review-ready. The evidence, rationale, and approval trail are what make it defensible.

What are the main types of journal entries?

The eight types most commonly encountered during close are standard, adjusting, reversing, reclassifying, accrual, deferral, consolidation and elimination, and top-side entries. They differ not just by form but by risk and control burden: adjusting entries align balances with accrual accounting, reversing entries prevent double-counting of prior-period accruals, reclassifying entries correct account or department coding, and consolidation entries eliminate intercompany balances for group reporting.

Who should approve journal entries?

Approval should depend on the type and risk level of the entry. Standard and recurring entries can be approved by a designated reviewer. Error corrections, material reclassifications, consolidation entries, and top-side entries should require controller review. Prior-period corrections under ASC 250 should escalate to controller and CFO/CAO if the error is material or affects prior reporting. The preparer should never be the approver.

What is the difference between a standard and an adjusting journal entry?

A standard journal entry records a routine transaction or manual movement that is not captured through a subledger, such as a bank fee, wire transfer, or recurring allocation. An adjusting journal entry is posted at period-end to bring the books into the correct accounting position. Accruals, deferrals, depreciation, allowance updates, and inventory valuation adjustments are common examples. In practice, adjusting entries usually require more support because they often involve cutoff, estimates, or management judgment.

How do you document a journal entry for audit?

A journal entry packet should include the business rationale, source evidence, tie-out to the calculation or schedule, policy basis where judgment is involved, preparer and reviewer/approver trail, and evidence that review occurred before posting. If the entry reverses or corrects an error, the packet should also document reversal ownership or root cause. The goal is that a reviewer or auditor can re-perform the logic without reconstructing the entry from emails, spreadsheets, or memory.

Can journal entries be automated?

Yes, but the best automation opportunity is usually the preparation layer, not just the posting step. Systems can pull source data, transform it, apply rule logic, prepare journal lines, assemble support, and route the proposed entry to the right reviewer. Routine and rule-driven entries are the strongest candidates. Entries involving estimates, materiality, error correction, or policy judgment still require human evaluation and approval.

What is a reversing journal entry and when do you use it?

A reversing entry is posted at the beginning of a new period to reverse a prior-period accrual. It is used when the accrual was recorded to recognize revenue or expense in the correct period, and the actual invoice or payment will post through the normal subledger workflow in the subsequent period. The reversal prevents the transaction from being counted twice. Reversing entries should be flagged at the time the original accrual is created, and the reversal date should be documented in the entry.

What is a top-side journal entry?

A top-side entry is a manual adjustment booked at a parent or consolidation level rather than through the operating subledger workflow. Because it often bypasses normal entity-level controls, it typically requires stronger support and approval. Controllers and auditors apply heightened scrutiny to top-side entries because they sit above the operational process where most preventive controls operate.

What makes a journal entry high risk?

Non-routine nature, high materiality, prior-period correction, manual top-side posting, KPI sensitivity, missing support, unusual account combinations, and weak approval lineage all increase JE risk. The most dangerous journal entry is often not the largest one. It is the one the team has normalized.