TL;DR
- Core decision: Reconciliation is the only point where payroll, finance and the general ledger have to agree in writing, so the question is how to keep that agreement honest when the close window shrinks.
- When doing nothing is right: When a single experienced owner reconciles six named items each month, can reproduce the variance report on request, and the audit committee has not asked a harder question.
- What has to be true: Someone has to own the reconciliation as a defined deliverable, not as a residual task after the journal is posted.
- How the options split: By who carries the variance explanation. In-house owner, embedded analyst, or external reviewer.
- Decision rule: Promote reconciliation to a signed deliverable before you add another control on top of a process that has none of them.
- Outcome to expect: A close that finishes on the same day every month, with variance lines that already have an explanation attached.
A Thursday at 4:47pm
A payroll manager in a mid-sized professional services firm opens the variance report for the period just closed. The gross-to-net difference against last month is three hundred thousand. She knows the headcount grew, the bonus cycle paid out, and two contractors moved to PAYE. She also knows that, on her current footing, she will sign the journal at five and investigate the variance next week, because next week is when she has time, and next week is when the explanation will be harder to build. So she signs it.
The journal posts. The close completes. The number that the firm reports to its auditors carries a variance the signatory could not fully explain, because the close deadline arrived first. This isn't a mistake made by a careless person. It's what happens when the control that catches everything the process missed is treated as the cheapest task on the calendar. But the real issue isn't the missed investigation. It's that the variance doesn't disappear when the month ends. It rolls forward, joins the next month's noise, and becomes harder to find in the period after that. That's what reconciliation is supposed to prevent, and that's what gets compressed when the calendar tightens.
When You Genuinely Do Not Need to Act Yet
Your current setup is genuinely fine. A payroll lead at a thirty-person agency runs payroll on the same day every month, has the same person review the same six lines every cycle, and can produce the variance explanation from memory before the journal posts. The general ledger owner receives the payroll journal and posts it the same day. The audit committee has not asked a harder question in two years. This isn't a firm that needs a new reconciliation programme. It needs the existing one written down so it survives staff turnover. The cost of action here's mostly the cost of distraction, and the cost of inaction is roughly zero. The reason this stage is comfortable is that the variance categories don't change, the inputs are predictable, and the close has been run the same way for long enough that the people involved have stopped noticing the discipline. That stability is the asset.
Best tools for Payroll & Compensation
Friction, not risk. A finance team at a two-hundred-person software company has outgrown a spreadsheet that tracks payroll against the ledger by account. The variances are usually small and explained, but the explanation lives in one person's inbox. When she is on leave, the close slips by a day. The number is still right. The audit trail is just thinner than it should be, and the dependency on a single individual is the only thing standing between an on-time close and a slipped one. This is the most common position in the market and the hardest one to argue for change inside, because nothing has visibly broken. The hidden exposure is the audit committee finding out about the dependency at the same moment as the team does, which is usually when the named owner hands in her notice and the next close takes four days instead of one. The fix here is a working paper, not a programme.
Real risk is now visible. A growing retailer with eight hundred employees and three payroll runs a month has just had its first qualification letter from the auditors, asking why inter-entity payroll recharges aren't reconciled to the receiving entities' ledgers. The payroll team can explain the gross and net figures. They can't explain the recharge side, because nobody owns it. The exposure is no longer theoretical. The question is no longer whether to build a programme. It's how quickly one can be stood up without disrupting a close that's already busy. The risk inside this stage is that the team responds to the auditor by buying a tool before it has agreed who owns the recharge process, which produces a tidier spreadsheet and the same gap.
The edge case. A group with a recent acquisition has two payroll systems, two calendars, and a finance team that didn't know that until the first post-acquisition close. The reconciliation programme can't be designed until someone decides which entity runs payroll for the acquired staff in the next cycle, and that decision has not been made. Here the reconciliation problem is downstream of an operating decision that has not landed, and any process design done before that decision is work that will be redone. Pause, make the operating decision, then build. The cost of skipping the pause is a reconciliation procedure that has to be re-papered the moment the operating model is confirmed, which is usually the second close after the cut-over.
The Questions You Ask Yourself at 11pm
Why did the gross move like that when headcount barely moved? The honest answer is usually a pay rise effective mid-cycle, a bonus that landed in this period, or overtime that compressed into a five-week month. The reasoning behind the question is that gross pay is the most stable line on a payroll report, because the inputs that feed it (headcount, base salary, hours) change slowly. A movement here means one of three things, and the right one needs to be named without checking three places. If you can't name which one without checking three places, your inputs aren't joined up. If you guess, the audit committee will catch it eventually, because the guess is what gets repeated in the working paper, and the working paper is what the auditor reads first. A wrong answer here becomes a control finding in the management letter.
Who actually owns the variance explanation when I am not here? The honest answer is the name of a person, not a team. The reasoning is that variance explanation is investigative work, and investigative work doesn't survive "everyone". If the answer is a name, and that person can do your job on a Tuesday morning without you, you've a process. If the answer is "everyone", you've a habit that survives until the first holiday season, when the habit runs out of slack and the close misses by three days. The cost of getting this wrong is a close that depends on a calendar you don't control.
Could I reproduce last month's variance report from the records alone? Without your inbox, your memory, or a chat thread. The honest answer is yes, and the working paper is the test. If you can't, the reconciliation isn't documented. It's performed, and performed work disappears the moment the performer leaves. If the auditor asks for the variance file next quarter and the only copy is in a personal folder, the audit trail fails on the first request.
What is the smallest variance I would still investigate? The honest answer is a number written down and agreed with finance, because a tolerance without an agreement is a preference. If the answer is a number written down and agreed with finance, you've a materiality policy. If the answer is "anything that looks weird", you've a judgement call that varies by who is on shift, which is how immaterial variances become material ones. A tolerance that lives in someone's head also tends to drift upward, because the easiest way to keep the close short is to raise the bar for what gets investigated.
If the auditors asked for the reconciliation tomorrow, what would I send? A folder, a narrative, a working paper, or an apology and a promise. The first three are right answers, and the difference between them is the difference between a defensible close and an anxious one. The reasoning behind the question is that audit requests rarely give more than a working day's notice, so the answer you give on the spot is the answer that describes the state of the file. If the answer is the fourth option, the auditor has already learned that the close depends on the signatory's availability rather than the records.
Three Honest Categories the Approaches Split Into
In-house owner, single point of accountability. One named person on the payroll or finance team carries the reconciliation as a defined deliverable. They produce a variance report each cycle, investigate every line above a documented threshold, and sign the reconciliation alongside the payroll sign-off. The category works when the payroll volume fits inside one person's working week and the variance categories are stable, because the owner needs to know the data well enough to spot a movement that doesn't fit the pattern. It fails when the same person also runs the payroll, because the reconciliation has to disagree with the operator sometimes, and operators don't enjoy disagreeing with themselves. The same-person problem also produces a single point of failure that the calendar exposes the moment the owner is away. The concrete example is a payroll lead at a professional services firm who owns the close calendar, produces the variance pack, and is the only person who can explain a movement to the auditors. It works until it doesn't, and the day it stops working is the day she goes on leave and the close slips because the deputy can run the payroll but can't defend the reconciliation.
Embedded analyst inside the finance team. The reconciliation lives with a finance analyst who sits close to the general ledger owner, with read access to the payroll system but not operating responsibility for it. The category works when the reconciliation problem is mostly a ledger problem: intercompany recharges, accruals, and postings that the payroll team doesn't see clearly, because the analyst has the ledger context that the payroll lead doesn't. It fails when the variance is inside payroll, because the analyst can describe the symptom but can't reach the cause. The analyst will end up escalating every payroll-side variance back to the payroll team, where they queue behind the next run, and the queue is the same problem the category was meant to solve. The concrete example is a finance analyst at a multi-entity retailer who reconciles the payroll journals to the receiving ledgers each month but escalates payroll-side variances back to the payroll team, where they queue behind the next run. The category produces a clean ledger tie and a noisy payroll explanation.
External reviewer, periodic. A firm of accountants or an independent payroll specialist reviews the reconciliation on a quarterly or annual basis, tests the lines, and reports findings back to the audit committee. The category works when the in-house team is competent but thin, and the board wants independent eyes on the close without building an internal control function, because the reviewer brings a set of expectations that the in-house team can use to benchmark their own work. It fails when the reviewer is asked to operate the reconciliation rather than test it, because the moment an external party owns the control, the in-house team stops trying to understand the variance and waits for the review. The concrete example is an annual review engagement that produces a management letter with two findings, both of which the in-house team could have produced themselves if the same discipline had been applied monthly. The review was useful. It was not a substitute, and the in-house team will be tempted to treat it as one because the review carries more weight than the monthly walk.
Five Diagnostic Questions You Can Self-Assess Against
Can you name, from memory, the largest variance in last cycle's reconciliation, what caused it, and the evidence you kept? To answer this for your own organisation, open the most recent working paper and check whether the largest variance has a written explanation, a document reference, and a date. If yes, the work is performed in your head, which is the most fragile place for it, because the memory is what survives a system change. If no, the work isn't being performed at all. The fix is a variance report that requires the owner to write the explanation down before the journal posts.
Is the reconciliation written down anywhere a person who didn't do the work could follow it? To answer this for your own organisation, hand the procedure document to a colleague who wasn't involved in last month's close and ask them to walk you through what they would do on day one. Pull the procedure document. If it doesn't exist, write one. If it exists and is shorter than two pages, it's a diagram, not a procedure. The fix is a working paper that walks through each reconciled item with the source, the comparison, the variance, and the explanation.
Does the reconciliation happen before the journal posts, or after? To answer this for your own organisation, find the calendar entry for the sign-off and count the working days between the variance walk and the journal post. Before means the reconciliation can still change what gets posted. After means it can only explain what already went through. The fix is a calendar entry that moves the reconciliation ahead of the sign-off by at least one working day.
Who can challenge a variance you've signed off, and do they actually do it? To answer this for your own organisation, look at the last three sign-off records and check whether any of them carry a second-line comment that disagreed with the first-line explanation. If the answer is the finance controller, and they do, you've a working second line. If the answer is "no one", the first line is the only line, and the audit committee will discover that the day they ask.
What is the longest a variance has sat in the file without an explanation? To answer this for your own organisation, run an aging report on the open-variance log and look at the maximum age, not the average. If the answer is a number of days, you've a queue and you've a control problem. If the answer is "I don't track that", the queue is invisible, which is worse, because invisible queues fill the working paper with stale items that no one can close.
What a Complete Reconciliation Actually Covers
Gross to Net Within the Run
This is the reconciliation inside payroll itself. Gross pay for the period against the sum of the deductions, the net, and any year-to-date adjustments that landed in this cycle. It earns a place because every other reconciliation depends on this number being right, and every error in this number compounds downstream, where the gross feeds the ledger and the net feeds the bank payment. The weakness is that a clean gross-to-net reconciliation can still hide a misclassified deduction or a payment to a leaver processed at the wrong rate, because the totals will agree even when the components are wrong. The check has to be done at the line level, not just the total. A line-level walk on a payroll run of any size takes longer than the total tie, and that is the trade the team has to accept.
Payroll to the Bank Payment
This is the agreement between the net pay calculated by payroll and the amount that left the bank account on the payment date. It earns a place because it's the only reconciliation that proves money actually moved to the right people in the right amounts. The weakness is that it confirms the payment without confirming it went to the right people. A payment to a closed account that bounced back and was reissued may net to zero against payroll and still indicate a control gap in the leaver process, because the reconciliation will treat the reissue as a new line rather than a failure. The reconciliation has to look at the failure cases, not only the totals.
Payroll to the General Ledger
This is the agreement between the payroll journals posted and the payroll accounts in the ledger, usually broken down by cost line and entity. It earns a place because the ledger is what the financial statements are built on, so any disagreement here becomes a disagreement in the reported numbers. The weakness is that a clean payroll-to-ledger reconciliation can still hide a misposted accrual or a payroll cost coded to the wrong department, because the total still ties. The check has to walk the payroll journals by account and by entity, not only in aggregate. The walk also has to be done in the same currency the journal was posted in, or a foreign-currency gain or loss will look like a variance.
Movement Against the Prior Period
This is the variance report: each payroll account compared to the prior month and the prior year, with an explanation for every movement outside an agreed band. It earns a place because it's the only reconciliation that forces someone to explain change, and unexplained change is the seed of every later problem. The weakness is that the band is set by the same person who does the reconciliation, which means a slowly drifting cost line can stay inside the band long enough to become the new normal. The band has to be reviewed periodically against a fresh set of eyes, and a cost line that has sat inside the band for several periods should be re-tested against a tighter band to check whether it is genuinely stable.
Statutory and Third-Party Liabilities
This is the agreement between the payroll system's calculated liabilities and the amounts actually paid to the relevant authorities and third parties, including pension contributions, court orders, and attachment of earnings. It earns a place because these are the liabilities that, if missed, create personal exposure for the signatory and reputational exposure for the firm. The weakness is that the reconciliation can confirm the payment was made without confirming it was made on time or at the right amount, because the payroll system may not capture the due date or the late-payment flag. The check has to be done against a calendar of expected payments, not only against the payroll system's view.
Balance Sheet Accounts Payroll Feeds
This is the agreement between the payroll-related balance sheet accounts and the underlying movements: the payroll control account, the bank clearing account, the tax and deduction control accounts. It earns a place because these accounts are where errors accumulate when no one is looking, and because the year-end audit will ask about them whether the close team thought to or not. The weakness is that the balance on the account is the result of many cycles of postings, so a single variance can be the sum of many small errors that are individually below the materiality threshold but jointly material. The check has to include an aging analysis of the balances, not only a reconciliation to the payroll system.
The Decision Table
| Situation | Scale | Setup | Primary Pain | Recommended Starting Point |
|---|---|---|---|---|
| Single entity, stable payroll, experienced owner | Small | One payroll system, one ledger, one person owns the close | The work lives in one person's head | Write the existing reconciliation down before anything changes |
| Multi-entity, single payroll system | Medium | One payroll system, several ledgers, finance team posts per entity | Intercompany recharges not agreed across entities | Reconcile the recharge side first, separately from the payroll side |
| Multi-system after an acquisition | Medium to large | Two payroll systems running in parallel, one ledger, one close calendar | Nobody knows which system owns which population this cycle | Pause reconciliation design, make the operating decision, then build |
| High headcount, complex pay (shifts, overtime, allowances) | Large | Single payroll system, complex inputs, multiple cost centres | Inputs to payroll do not tie to the time and attendance system | Reconcile the inputs before the outputs, because no payroll reconciliation will fix a bad input |
| Outsourced payroll with in-house finance | Any | Provider runs payroll, finance posts the journals, neither fully owns the variance | Variances fall into the gap between provider and finance | Name one party as the reconciliation owner in writing, and put it in the contract |
| Audit qualification or close miss in the last 12 months | Any | Any of the above | The control failure is now visible to someone outside the team | Build the programme around the specific line the auditors asked about, not a generic template |
Reading a Variance
| Variance Type | Likely Cause | Can It Wait |
|---|---|---|
| Gross pay up, headcount flat | Pay rise effective mid-cycle, bonus in this period, overtime compressed into a five-week month | No, but it is usually explained quickly |
| Net pay up more than gross pay up | A change in tax basis, a new deduction, a refund that reduced withholding | No, tax basis changes have downstream effects |
| Statutory liability up, payroll cost flat | A correction posted to the liability account, a backdated reclassification, a missed contribution | No, statutory exposures compound with penalties and interest |
| Bank payment less than net pay | A bounced payment, a manual hold, a payment run that did not complete | No, leavers and unresolved payments create employee relations risk quickly |
| Payroll cost in the ledger differs from payroll system | A posting error, a missed accrual, a timing difference on the cut-off | Depends, some timing differences are normal, but unexplained ones should not sit |
| Intercompany recharge does not tie to receiving entity | Wrong allocation key, entity set up late, missing recharge in one direction | No, this is a common audit finding |
| Balance on payroll control account growing | Unreconciled items accumulating, no one watching the aging | No, the longer it sits the harder it becomes |
| Year-to-date figure changing retroactively | Backdated pay rise, late leaver payment, correction for a prior period | Investigate before next cycle starts, because the next cycle will compound the error |
| New cost centre appearing with a small balance | Re-organisation, new hires, a transfer posted to the wrong line | Yes, but flag the cost centre for the next variance walk |
A Close Calendar That Holds
The reconciliation has to finish before the journal posts, which means it has to start before the last week of the period, not during it. Most close calendars put the reconciliation in the same week as the sign-off, which is the week with no slack, and that's why the reconciliation gets compressed.
A close calendar that holds treats the reconciliation as the critical-path activity, because it's the activity that determines whether the sign-off is honest. The payroll run completes mid-period. The variance report runs the next morning. The payroll lead walks the variances on day two and writes the explanations into the working paper. Day three is for the lines that need a finance colleague: the ledger tie, the intercompany recharge, the statutory liabilities. Day four is the second-line review, usually the finance controller, who challenges the explanations that are too thin. Day five is the sign-off and the journal posts.
The work that makes this calendar hold is the work done before the period ends. The cut-off has to be agreed and documented before the last working day, because a cut-off that moves mid-cycle will move every reconciliation that follows. The inputs to payroll, including the time and attendance data, the new joiners, and the leavers, have to be locked before the run, because a payroll run on inputs that were never locked will produce a reconciliation on figures that can still move. The statutory payment calendar has to be visible to the payroll lead before the period starts, not after, because a missed statutory payment is the kind of variance that doesn't wait for the investigation.
The discipline is to publish the close calendar before the period starts, and to hold the order. The most common failure is to start the variance walk on the day the payroll runs, which leaves two days for the work that needs three, and the third day is the day the variance explanations are usually written.
| Day | Activity | Owner |
|---|---|---|
| Mid-period | Cut-off agreed, inputs locked, payroll run executed | Payroll lead |
| Day 1 | Variance report generated, walkthrough started | Payroll lead |
| Day 2 | Explanations written into the working paper, escalations flagged | Payroll lead |
| Day 3 | Finance-side reconciliations: ledger tie, intercompany, statutory | Finance analyst |
| Day 4 | Second-line review, challenge of thin explanations | Finance controller |
| Day 5 | Sign-off and close | Payroll lead and finance controller |
What to Put in Writing
| Artefact | Who owns it | When it is written | What it prevents |
|---|---|---|---|
| Variance report for the period | Payroll lead | Before the journal posts | A signed journal with an unexplained variance |
| Reconciliation working paper with each line walked | Payroll lead or finance analyst | Before the journal posts | An audit finding that the reconciliation was not performed |
| Cut-off memo for the period | Payroll lead | At the start of the period | A cut-off that moves mid-cycle and breaks the calendar |
| Statutory payment calendar for the period | Finance controller | Before the period starts | A missed statutory payment that becomes a year-end issue |
| Second-line review notes | Finance controller | Before the sign-off | A thin explanation accepted because no one challenged it |
| Sign-off record with both signatures | Payroll lead and finance controller | At the close | A close that one person remembers differently from the other |
| Aged analysis of payroll control account balances | Finance analyst | Monthly | A balance that grew unnoticed because no one looked at the age |
| Reconciliation procedure document | Payroll lead or finance controller | Annually, refreshed | A process that survives only in one person's head |
| Contract allocation of reconciliation responsibility | Procurement or finance controller | At contract signature, reviewed annually | A variance that falls into the gap between provider and in-house team |
| Audit committee narrative on the close | Finance controller | Quarterly | A committee that learns about a control gap from the auditors rather than the team |
| Tolerance band and review log | Finance controller | At the start of the year, refreshed after any material change | A drift in the variance threshold that lets movements accumulate unnoticed |
Questions to Ask Before You Commit
Ownership. Who carries the variance explanation when it can't be produced in the time available? A bad answer sounds like "we'll figure it out as we go".
Calendar. What is the latest date the reconciliation can be completed without delaying the journal? A bad answer sounds like "we usually get it done in time".
Evidence. What artefact proves the reconciliation was performed, and is it stored somewhere a person who didn't do the work can find it? A bad answer sounds like "it's in my inbox".
Independence. Who challenges the reconciliation before it's signed, and do they have enough context to challenge it properly? A bad answer sounds like "the finance controller signs it off".
Scope. Which of the six reconciliations above are in scope for the first quarter, and which wait? A bad answer sounds like "everything, eventually".
Tolerance. What variance gets investigated, and who set that tolerance? A bad answer sounds like "anything that looks odd".
Hand-off. What happens to the reconciliation when the named owner is away for two weeks? A bad answer sounds like "we hold the close until they're back".
Provider. If payroll is outsourced, which party signs the reconciliation, and where in the contract is that written? A bad answer sounds like "we work it out between us".
Audit history. What did the auditors ask last year, and what did we change as a result? A bad answer sounds like "we addressed it in the management response".
Sustainability. What is the smallest change in headcount or pay complexity that would break this setup? A bad answer sounds like "we've plenty of room".
What Getting This Wrong Costs
The first cost is the one that shows up in the auditor's letter, and it's the cost most teams plan for. A reconciliation that doesn't hold under audit produces a finding, the finding produces a management response, and the management response produces a control the team has to build anyway. So the work gets done either way. The difference is whether it gets done before the audit, when it's cheap, or after, when the budget is smaller and the patience is thinner. The second cost is the one that shows up in next month's close. A variance that was not investigated doesn't vanish. It joins the next period's noise, and the explanation that would have been easy in the original period becomes a forensic exercise once it has been joined by a second unexplained variance and a third. The third cost is the one that shows up in the payroll lead's calendar. The investigation that was deferred gets done in the next close, and the next close gets compressed to make room, and the close after that compresses again. So the cost of getting it wrong is a slow, year-long tightening of the calendar that ends in a missed close or a qualified audit, and the team never quite identifies the moment it started. So the question isn't whether the team can afford the time to reconcile properly. The question is whether the team can afford a close that gets tighter every month and never quite breaks.
When You Are Ready to Go Further
If the questions above have surfaced a programme that needs to be built, or a programme that needs to be tested by someone outside the team, the next move is to see what others in your position have chosen and how those choices held up under audit. HROpsLab publishes independent comparison work across the payroll and HR operations category, written for the person who signs the journal and not for the person selling the software. The comparison library is built the same way this article is built: with the controls the team needs to own, the trade-offs the team will actually face, and the questions the team should be asking before anything is signed.
Frequently Asked Questions
What should I reconcile first if I am starting from nothing?
Start with gross to net within the payroll run, then payroll to the bank payment, then payroll to the general ledger. These three are the foundation. The movement against prior period, statutory liabilities, and balance sheet accounts come after, because they're most useful when the first three are stable.
What level of variance is acceptable without investigation?
The honest answer is the level your finance controller has agreed with you in writing and that the audit committee has seen. Any variance below that threshold can be carried forward with a note. Any variance above it needs an explanation before the journal posts, because a small unexplained variance carried forward becomes a large unexplained variance over a year.
Who should sign off the reconciliation?
The payroll lead and the finance controller, jointly. The payroll lead owns the payroll-side explanation. The finance controller owns the ledger-side explanation. Neither signature covers both, and a single signature is the most common audit finding in this area.
What do I do with a variance I cannot explain before the deadline?
Post the journal with the best explanation you've, flag the variance as open in the working paper, and commit to a date by which it will be investigated. A posted journal with a flagged open variance is defensible. A posted journal with an unsigned variance isn't, because the audit trail won't show that anyone tried.
Does the payroll provider reconcile for us?
Some do, some don't, and most contracts don't say. The reconciliation has to be owned by someone with the authority to investigate across systems, and that's usually the in-house finance team. Ask the provider to put in writing what they reconcile, what they don't, and what evidence they produce for each.
How do I reconcile after a mid-year system change?
Run both systems in parallel for at least one full cycle, reconcile the outputs to each other, and investigate every variance before the old system is decommissioned. The mid-year change is the most common source of year-end audit findings, because the parallel run is usually shorter than it should be and the variances are usually larger than expected.
What do I do when payroll and the ledger disagree and neither side will move?
Stop and walk the underlying transactions, not the totals. A disagreement at the total level is usually a disagreement at one specific transaction, and that transaction is usually identifiable. If it isn't, escalate to a named owner on each side with a deadline, because an unresolved disagreement at close becomes an audit finding at year end.
How often should the reconciliation procedure be reviewed?
Annually at minimum, and immediately after any material change: a system migration, an acquisition, a change in payroll provider, a significant change in pay complexity. A procedure that was right last year isn't necessarily right this year, and the review is the moment that surfaces that.
Independent comparisons for HR and payroll leaders. No vendor pays to appear here.
