How a firm keeps a record it can prove.

This playbook is written for three people who inherit the same problem from different angles.

• The managing partner who runs the firm from memory and spreadsheets, and discovers the gaps only when a matter goes wrong.
• The associate drowning in coordination: chasing signatures, reconciling numbers, and holding the state of a dozen matters in their head.
• The legal accountant or auditor who inherits the mess at year end and has to make it defensible after the fact.

If you recognize any of them, the chapters that follow are addressed to you.

Read it whichever way fits the hour you have.

• Cover to cover, about ninety minutes. Each chapter builds on the one before, from intake through conflicts, billing, receipts and the audit record.
• By problem. Jump to the chapter that names the thing keeping you up: a conflict you should have caught, an invoice you cannot correct cleanly, a receipt you cannot produce, a record you cannot defend.

Nothing here assumes you started at the beginning.

Every chapter states a principle in plain terms, then shows where the principle lives in the software. A block marked MERSAL IN ACTION is documentation, not marketing. It describes the implemented behavior: what the system records, what it refuses, and which role may do what. If a chapter claims something is enforced, the block tells you what enforces it.

Read the principle first. The block is there to show that the principle is not a slogan.

Legal work is performed in the present and questioned in the future. A fee is agreed in a meeting. A deadline is noted on a call. A document is handed across a desk. Months pass. Then someone asks what happened, and the honest answer is that no one is quite certain.

Memory decays, and it decays in a particular direction. People remember the version that favors them. This is not dishonesty; it is how recollection works. Two competent professionals can leave the same meeting with two accounts of what was agreed, and both can be entirely sincere.

A practice cannot run on sincerity. It runs on what can be produced when asked. The discipline is not to remember better. It is to record at the moment, so that memory is never the thing on trial.

Every firm answers to three tribunals, and each one arrives without warning.

• The client who questions a fee. They want to know what was billed, when, at what rate, and whether they ever agreed to it. A confident answer needs the invoice, its serial, its issue date, and the record that the client saw it.
• The auditor who samples serials. They do not read every invoice. They pull a few and check that the numbering is sequential, that nothing is missing, that a receipt exists wherever money was taken. A gap in a serial book is a question you will answer under pressure.
• The court or bar that asks for the file. They want the whole matter: correspondence, filings, instructions, dates. They want it complete and they want it soon. What you cannot assemble, you cannot rely on.

None of these tribunals cares what you meant to do. Each cares only what the record shows.

There is a line beneath which your word is enough, and above which it is not. A small query from a long-standing client sits below it. A formal dispute, an audit sample, a regulator's request sits above it.

Below the line, goodwill carries you. Above it, goodwill is irrelevant and only the record speaks. The difficulty is that you cannot predict which questions will cross the line, and by the time one does, the moment to have recorded the fact has already passed.

So the sound practice records everything as though it will cross the line, because some of it certainly will.

The distance between recollection and proof is easy to see once it is laid out plainly.

The left column is a feeling. The right column is a document. Only one of them survives a challenge.

Recording as you work costs minutes. Attaching an email to a matter, issuing an invoice that freezes at issue, letting a receipt write itself when a payment settles: none of these takes real time, and each happens once, at the moment the facts are fresh.

Reconstruction costs days. It means searching inboxes, calling colleagues, guessing at dates, and assembling an account after the fact that a skeptical reader will treat as exactly that, an account. Worse, it costs credibility, because a file built in hindsight always looks like a file built in hindsight.

The two efforts are not comparable. One is cheap and contemporaneous. The other is expensive and doubted. A firm that understands the asymmetry records first and argues later.

A matter does not end when the work does. It is questioned, audited and reopened long after everyone has moved on. This is why the record must survive the people who made it, intact and retrievable, without depending on who is still at the firm or on what they happen to recall.

An archived matter should keep its full record and remain exportable, so that a file requested years later is produced whole, not pieced together from what remains. The question is never whether the work was good. It is whether the work can still be shown.

Ask where a matter lives and the honest answer is usually a list, not a place. Most practices run on seven surfaces, each holding a fragment of the truth:

• Email, for correspondence, engagement terms and the occasional attachment that becomes the only copy.
• Chat applications, for the fast decisions that never reach the file.
• Spreadsheets, for fee estimates, time and whatever the accounting package will not do.
• Wall and shared calendars, for hearings and deadlines, kept in one person's head as much as on the wall.
• PDFs in shared drives, for invoices, filings and signed documents.
• Accounting software, for the numbers, disconnected from the matter that produced them.
• Memory, for everything nobody wrote down.

Each surface works in isolation. The failure is between them. A figure corrected in the spreadsheet is not corrected in the PDF already sent, and the calendar does not know the matter settled.

This is not a problem that ages out. Four forces are pulling the fragments further apart.

• Remote and hybrid work. The shared drive is no longer down the hall. The people who once reconciled the pieces by walking between desks now reconcile nothing.
• Multi-currency clients. A fee agreed in one currency, billed in another, paid in a third, with the rate drifting between estimate and settlement.
• Cross-border matters. Parties, counsel and counterparties in several jurisdictions, each with its own record-keeping expectation.
• Language mixing. Correspondence in one language, the invoice in another, the court document in a third, and no single file that holds all three cleanly.

Every one of these adds a surface, a translation step or a version. None of them removes one.

The direction of tax regulation is not in dispute, even where the timetable is. Many jurisdictions are moving the same way: toward electronic invoicing, structured and machine-readable records, and the active enforcement of sequential invoice numbering. VAT regimes generally require that invoices carry unbroken serial sequences, that exempt supplies be distinguished from taxable ones, and that a corrected invoice leave the original on the record rather than overwrite it.

A scattered practice cannot satisfy this on demand. Sequential numbering breaks the moment a second person issues an invoice from a second template. Exempt and taxable items blur when the rate lives in someone's memory rather than on the line. The reissue overwrites the original because the original was only ever a PDF.

Client expectation is moving in the same direction, from the other side. Clients now bank, file their taxes and track shipments through portals that show status, history and documents in their own language, at any hour. That habit does not stay with the bank. A client who can see every transaction on a banking application will not indefinitely accept a firm that answers "let me check and get back to you" when asked for the status of a matter or a copy of an invoice.

Counsel is being measured, quietly, against an interface the client already uses every day.

These two pressures meet at the audit. For now, a scattered practice can still assemble a defensible record after the fact: a late evening, a shared drive, and someone who remembers where things are. That grace period is narrowing. As numbering enforcement tightens and records are expected in structured form, the reconstruction stops being merely painful and begins to fail outright. An auditor who asks for the unbroken invoice sequence, the proof of delivery and the original of a corrected bill is asking for artifacts a folder of PDFs cannot produce.

The question is no longer whether the scatter is manageable today. It is whether it will survive the next audit, and the one after that.

Before a firm chooses software, before it designs a billing workflow, before it drafts a single retainer, it makes one decision that governs all the others. To what do records attach? The answer is usually never spoken aloud. It is inherited from whoever first set up the shared drive, and it quietly shapes every search, every handover and every reconciliation for years.

Get this decision right and the rest of the operation has somewhere to stand. Get it wrong and no amount of discipline downstream will save you. A firm can hire careful people, buy capable tools and still lose an afternoon to a question a filing clerk should have answered in seconds. The fault is not effort. It is foundation.

Most firms organize by one of three axes. Each is intuitive. Each breaks under load.

The common fault is that none of these axes has an end. A year rolls over. A client relationship continues. A lawyer's caseload never closes. Records need a container that opens, fills, closes and can be set down intact.

A matter is the smallest complete unit of legal work. It has parties, so conflicts can be checked against it. It has a docket of hearings and deadlines. It accumulates documents and correspondence. It generates money, in fees, disbursements and receipts. And, unlike a year or a client, it ends.

Because it ends, it can be reasoned about as a whole. Everything the firm does can anchor to it: a task with an owner and a due date, a hearing in each member's local time, an invoice, an email sent on the firm's letterhead. When the unit of record is the matter, the question of where a thing belongs has exactly one answer, and everyone gives the same one.

The symptoms are quiet and easy to normalize. Read the list and count how many describe an ordinary week at your firm.

• Documents are found by asking a person, not by opening a file. Knowledge lives in someone's memory rather than on the record.
• Invoices are reconciled from recollection. Nobody can reconstruct what was billed, and why, without a conversation.
• Two lawyers hold different versions of the same file, and each believes theirs is the current one.
• A handover takes weeks, because the departing lawyer is the index.
• The same client name is spelled three ways across three folders, and a conflict check cannot see all three at once.
• Closing a matter means dragging files into a folder nobody expects to open again.

One or two of these is friction. Four or more is a structural fault, and no individual is to blame for it. The container is wrong.

Once the matter is the unit, it needs a name that is stable, unique and legible. Two conventions dominate, and the choice is a question of preference rather than correctness.

What matters is not which you pick but that the convention is fixed, applied without exception, and owned by the firm rather than by a tool you might one day leave. A number a partner can read aloud on a telephone call is worth more than a clever scheme nobody remembers.

Most systems treat a closed matter as a body to be buried. The file is moved somewhere cold, and its record decays into an obligation nobody wants to disturb. This is a failure of imagination about what closing means.

A matter that ends is not a matter that disappears. It becomes a settled, searchable part of the firm's memory: the register a future conflict check must see, the precedent the next associate will reopen, the evidence an auditor may request years later. Archiving should be a lifecycle stage in which the full record is retained and remains exportable, not a graveyard from which nothing returns.

A conflict check that runs after the engagement letter is signed is not a check. It is a post-mortem. By then the firm has already taken confidential information, committed a lawyer, and perhaps filed. Unwinding a live matter for conflict is expensive, embarrassing, and in many jurisdictions reportable to a disciplinary body.

The clearance must gate the opening, not trail behind it. In practice that means one rule the whole firm can recite: no matter number issues, and no workspace opens, until the search has run and its result is on the record. The order is not a formality. It is the difference between a firm that declines a bad engagement and a firm that discovers one.

A name is the smallest part of a conflict. Adversity hides in structure. The client sitting across the table may be clean, while the parent company that owns him is the counterparty in a matter you closed last year. A complete pre-flight cross-references far more than the person who walked in.

A search that stops at the individual's name will clear a matter that a search of the corporate tree would have flagged.

Memory-based conflict checking works in a firm of four lawyers who have lunch together every day. It fails quietly as the firm grows. Every new hire arrives with a client history no colleague remembers. Every retirement carries a partner's mental index of decades of adversaries out the door. Every merged practice doubles the register and halves the shared memory of it.

The firm that relies on someone recognizing a name is protected exactly until the one person who would have recognized it is on leave, retired, or too new to know. A register does not age, forget, or take a holiday. It is the institution remembering, rather than an individual.

A match is not a verdict. Many conflicts are waivable: unrelated matters, consenting clients, information barriers that genuinely hold. The discipline is not to avoid every conflict. It is to resolve each one deliberately, and to leave behind a record that a regulator or a successor partner can read years later.

A waiver that lives in a partner's inbox is not a waiver. The record must show what the match was, who reviewed it, on what reasoning it was cleared, and when. Authority matters as much as reasoning: the decision to waive belongs to a partner, not to the associate who wants the matter opened this afternoon.

The most dangerous outcome is not a conflict found. It is a search that fails and is read as clear. A query that times out, a register that was never reached, a name misspelled into nothing: each returns an empty result, and an empty result looks exactly like a clean one.

This is the failure mode that ends careers, because everyone acted in good faith on a screen that told them nothing was wrong. A conflict system must distinguish we searched and found nothing from the search did not complete. A failed search must report itself as failed. Silence is not clearance.

A workable protocol fits on a single page and survives the busiest week of the year:

• Collect every name before intake: client, counterparty, related entities, directors and ultimate beneficial owners
• Run the search against the firm's full historical register, not against living memory
• Confirm the search actually completed, and treat any incomplete search as a stop, not a pass
• Where a match appears, escalate to a partner, and never clear it at the desk
• Record the waiver decision in full: the match, the reviewer, the reasoning, the date
• Only then issue the matter number and open the workspace

Legal billing looks like one activity: you send a client a document asking for money. The tax authority sees three separate obligations wearing the same coat, and it expects each to be kept in its own book.

The first duty is the VAT invoice. It creates a tax liability. Most VAT regimes require these documents to be numbered sequentially and without gaps, because a missing serial reads as a suppressed sale and an out-of-order serial reads as a document created after the fact. The register itself is evidence.

The second duty is the exempt document. Certain legal services, and most disbursements you advance on a client's behalf, carry no VAT. Billed under a VAT serial, they pollute the taxable base: the return must reconcile every number in the VAT book to output tax, and an exempt line sitting in that sequence forces you to explain a serial that produced no tax. Better it was never in that book.

The third duty is the proforma. A proforma is a request, not a fiscal event. It is a quotation, an advance notice, a way to tell a client what a matter will cost before the work is billable. It must never touch the revenue sequence, because a number in the revenue book is income the moment it is issued. Put a proforma there and you have overstated earnings, and possibly declared tax, on money that may never arrive.

Three duties, three registers. This is not a preference. It is the same discipline a chart of accounts imposes everywhere: unlike things belong in unlike ledgers.

The test for where a document belongs is simple. Ask what the document does, not what it is called.

Each book runs its own count. A proforma numbered PF-0041 has no relationship to VAT invoice V-0287 or exempt invoice E-0113. Three sequences, three audits, no crossing.

Separating the books does not mean sending a client three envelopes. Most legal engagements mix taxable work with exempt pass-through in a single month. The client wants one statement. The tax return wants clean classification. Both can be satisfied if the ruling happens at the line, not at the document.

Per-line VAT means each line on a statement is ruled on its own: taxable at the applicable rate, exempt, or a zero-rated disbursement. The statement totals each class separately and shows the client one figure to pay, while the underlying entries file to the correct book behind the scenes.

Consider a single monthly statement:

Lines 1 and 2 belong to the VAT book and generate output tax. Lines 3 and 4 are pass-through: money you advanced, not income you earned, and taxing them would be an error the client should not pay for. The document is one. The bookkeeping is three-aware. That is the whole point of ruling per line.

The dangerous document is the proforma, because it looks finished. It carries a number, a client, a figure. The temptation is to treat it as an invoice once the client agrees, editing it in place or copying its serial into the revenue book.

Neither is safe. A proforma that becomes an invoice by edit leaves no trace of what it was, and a gapless VAT sequence cannot borrow a number that already lived in another book. The correct motion is conversion: the proforma is closed, and a fresh document is issued into the VAT or exempt book with the next serial in that book's own sequence. The proforma stays on the record as a proforma. The invoice is born as an invoice. Nothing changes identity halfway through its life.

Conversion should also be one-directional and gated. A proforma becoming revenue is a decision about when income is recognized. That decision belongs to someone with authority to make it, not to whoever happened to draft the estimate.

A document file records information. A legal instrument carries consequences: it can be relied upon, disputed, produced in evidence, and audited by someone who was not in the room when it was made. An invoice belongs in the second category. It is the firm's assertion that a sum is owed for named work, addressed to a named party, on a stated date, under a stated tax treatment. When a tax authority, a client's controller, or an opposing expert examines it, they are not reading prose. They are testing whether the assertion holds.

Five properties decide whether it holds. Identity: the invoice carries a serial drawn from the correct book. Integrity: its contents are fixed at the moment of issue. Correctability: errors are repaired by reissue, with the original preserved. Currency honesty: any conversion is disclosed and frozen. Provenance: the record shows who issued it, when, and to whom. Remove any one and the instrument weakens. Remove integrity and it collapses.

An invoice serial is not decoration. In many VAT regimes it is a legal requirement that invoices carry sequential numbers without unexplained gaps, so that a run of invoices can be reconciled and none quietly suppressed. A serial is a promise that this document occupies exactly one position in a known sequence.

The complication is that a firm does not issue one kind of statement. Taxable services, tax-exempt items, and pro forma requests are different instruments with different consequences, and they cannot share a numbering line without confusing all three. The discipline is to keep separate books, each with its own independent sequence.

A single statement to a client may legitimately mix taxable and exempt lines, which is why the tax treatment belongs on the line and not on the whole invoice. But the serial that identifies the document must come from the book that matches its legal nature. A pro forma that borrows a taxable serial has just been recognized as revenue by accident.

The moment an invoice is issued, its contents must stop moving. This is the property clients and auditors care about most, and it is the one most casually violated by systems that treat an invoice as an editable form. If the total, the rate, the recipient, or the line items can change after issue, then no issued invoice can be trusted, including the ones nobody touched, because the reader has no way to tell an untouched invoice from an edited one.

Errors are inevitable, and freezing does not forbid correction. It forbids silent correction. A defensible workflow repairs a mistake by issuing a new, separately numbered document that references and supersedes the first, while the original remains on the record exactly as it was sent. The client's copy and the firm's copy still agree. The correction is visible as a correction, which is precisely what an auditor needs to see.

Cross-border practice bills in more than one currency, and exchange rates move daily. This creates a quiet opportunity for disputes: the firm converts at one rate, the client's accounts department books at another, and months later the two records disagree over a figure that both sides believe they read correctly.

The rule that removes the ambiguity is to bill in the currency of the engagement, freeze the conversion rate at the instant of issue, and print both the original amount and the converted amount on the face of the invoice. The rate is then a fact about that document rather than a live quantity. If the currency moves the next morning, the invoice does not. A reader years later can reproduce the arithmetic exactly, because every number needed to check it is on the page.

Two conventions are common and both are defensible. A continuous sequence never resets, so INV-004871 follows INV-004870 across year boundaries. A yearly reset restarts the count each fiscal year, usually carrying the year in the serial, so 2026-0001 opens the new year. The choice is a matter of firm policy and local filing expectations. What matters is that the firm picks one, states it, and applies it without exception.

The non-negotiable is that gaps must be explainable. A missing serial invites the obvious question: was an invoice issued and then hidden? A sound system can answer it, because a number is consumed only when a document is genuinely issued, and a cancelled or reissued document leaves a visible trail rather than a hole. If your numbering can produce a gap that no one can account for, the whole sequence loses its evidentiary value.

Before an invoice leaves the firm, it should satisfy every line below. Each item maps to one of the five properties.

• Serial from the correct book, sequential and gap-free within that book.
• Issuing firm and issuing lawyer identified, on the firm's letterhead.
• Named recipient, the party actually liable, in that party's language where relevant.
• Date of issue, fixed, and matching the moment the contents froze.
• Per-line tax treatment, so taxable and exempt items are unambiguous.
• Currency and, where converted, the frozen rate, with both amounts shown.
• Clear reference to any superseded invoice when the document is a reissue.
• Totals a reader can reconstruct from the lines without outside information.

An invoice that passes this list can be handed to a stranger years later and defended without the person who issued it in the room. That is the only test that matters.

A proforma is a statement of what work would cost. It is a quote, a fee proposal, a draft for discussion. It is not a demand for payment and it is not a record of anything earned. The distinction is not academic. The moment a document that carries no obligation is treated as if it carried one, three separate records begin to lie at once.

Receivables overstate. Money a client may never owe appears as money the firm is owed. The revenue picture the partners rely on to make decisions is inflated by drafts. Tax exposure appears where none should exist. In most VAT regimes the tax point is triggered by an issued tax invoice, so a proforma sitting in the taxable book invites VAT to be declared on sums that were never invoiced and never collected. And the audit trail blurs. A negotiation draft that looks identical to an issued invoice cannot be told apart later, so the firm loses the ability to prove what it actually charged.

The failure is rarely deliberate. It happens because the proforma and the invoice look alike and behave alike inside a single numbering scheme. A fee proposal is drafted using the same template as a real invoice. It gets a number from the same sequence. It is emailed to the client to move the conversation forward. Weeks later, nobody can say whether serial 2024-0417 was a quote or a bill, because both drew from the same book.

Once that ambiguity exists, every downstream report inherits it. The receivables ledger counts the draft. The VAT summary counts the draft. The auditor, working from the same records, has no basis to exclude it. The only durable cure is structural: proformas must be a separate species of document, with their own book, their own marking, and a wall between them and revenue that no accident can breach.

Four rules hold the line, and each one has to be enforced by the system rather than by memory.

• A separate serial book. Proformas draw from their own sequence, independent of the VAT and exempt invoice books. A proforma serial and an invoice serial can never collide, and a gap in one is never explained by the other.
• Unmistakable marking. Every proforma states on its face that it is a proforma and not a tax invoice. A client, a colleague or an auditor can tell at a glance which document they are holding.
• No contact with revenue. Proformas are excluded from receivables, from the revenue ledger and from every VAT computation until they become something else. They do not inflate the numbers because they are not in the numbers.
• One-way conversion, by an authorized person, on the record. A proforma becomes a real invoice only through a deliberate act by someone with the right to perform it. The conversion runs one direction, and it leaves a trace: what was converted, by whom, and when.

Proformas are not a loophole to be avoided. They are the correct tool for a specific and common set of situations, and using them properly keeps the revenue book honest.

• Fee proposals. Before an engagement is agreed, the client needs a costed statement to consider. That statement is an offer, and it belongs in the proforma book until the client accepts and the work justifies a real invoice.
• Advance and retainer discussions. When a payment on account is being negotiated, a proforma states the amount and terms without prematurely recognizing revenue or triggering a tax point.
• Cross-border quotes. A client in another jurisdiction often needs a formal price before committing, sometimes to arrange their own approvals or currency. A proforma serves this without pretending a taxable supply has occurred.

In each case the pattern is the same. The proforma does the work of communicating a price. The invoice, issued later and deliberately, does the work of recording income.

A bill is not a single event. It is a sequence, and each step in the sequence is a separate fact that either happened or did not. Treat the sequence as four distinct states, and record each one on the server when it occurs.

• Sent. The statement left the firm. This is the one state most firms bother to record, usually as a line in a sent folder, and it is the weakest of the four. It proves you released the document. It proves nothing about what happened after that.
• Viewed. The client opened the statement. A view is the first fact the client cannot later deny without contradicting a log they did not write and cannot reach.
• Acknowledged. The client confirmed receipt by an act of their own. This is consent to the fact of delivery, captured at the moment it was given.
• Paid. Money settled against the statement. Payment is the strongest state, and on a well-built system it arrives already carrying its own proof.

The discipline is to record each state when it happens rather than to infer it later from silence. Silence is not a state. A client who has said nothing has not been shown to have received anything, and a case built on the absence of a complaint collapses the moment the client says the complaint was made and ignored.

Service of process has a settled logic. A document is not served because it was posted. It is served because service was effected and recorded, and a tribunal reads the record, not the intention behind it. Billing deserves the same logic.

The Acknowledge Receipt press is the practice equivalent of signed service. It converts a passive event, an email that may or may not have landed, into an affirmative act by the person who owes the money. They saw the statement. They pressed the button. The system recorded the moment, server side, in the client's own language. From that point the question is no longer whether the bill was delivered. That question is closed.

This changes the character of every later conversation about the invoice. A dispute after acknowledgment is a dispute about the amount, which is a legitimate professional discussion. A dispute about whether the bill was ever received is simply not available.

Payment deserves a record of the same weight as the bill and stronger protection than it. When a payment settles, the system writes a receipt with its own independent serial, such as R-000214, separate from the invoice books, and seals it. Sealed means sealed for everyone: no edit, no deletion, by any role, including the administrator who runs the firm.

This is deliberate. A receipt that an administrator can quietly alter is not evidence of anything. It is a note. The value of a receipt lies precisely in the fact that no one, however senior, can reach back and change what it says. Corrections, when they are genuinely needed, happen the way honest ledgers have always handled them: by a reversal entry that stands beside the original, never by rewriting the original out of existence. The history stays legible. Anyone auditing the account can see what was recorded, what was reversed, and why.

The older model separates the payment from the proof of payment. Money moves through one channel. A receipt is typed later, by hand, in another. The gap between the two is where errors and disputes live: receipts issued for payments that never cleared, payments that cleared but were never receipted, serials assigned out of order by a tired hand at the end of a month.

When the client pays by card in the portal, that gap closes. The settlement of the payment and the writing of the sealed receipt are the same event. The receipt is not a description of something that happened elsewhere. It is the thing that happened, recorded at the instant it happened, with its serial and its seal already in place. The payments ledger behind it is append-only, so the record of what was received only ever grows. Nothing is quietly unwound, and nothing has to be reconstructed from memory afterward.

Consider a fee dispute six months after the work has closed. The client's position is confident and specific: the final statement was never received, so the account was never truly presented, so the fee is contested. Under the old model this becomes an argument between two recollections, and the firm that keeps its records in a sent folder tends to lose arguments between two recollections.

Here the matter workspace holds the sequence. The statement was sent on the fourteenth. It was first viewed on the sixteenth. Acknowledge Receipt was pressed on the seventeenth, by the client, recorded server side. The dispute does not proceed to a contest of memory, because there is no contest to have. The trail is not an assertion the firm makes about the client. It is a record the client's own actions wrote.

A law firm's books are not a shop's books with a different logo. Three structural facts set them apart, and each one is where an inspection tends to start.

First, money arrives before it is earned. A retainer is a payment on account of work not yet done. Until you do the work, it is not your income; it is a balance you owe, as service or as a refund.

Second, money passes through the firm that was never its own. A filing fee, an expert's charge, a courier: paid on the client's behalf and recovered at cost. Treated as revenue, it inflates your turnover and your tax at the same time.

Third, the tax treatment changes line by line. A single statement can carry a taxable service, an exempt item and a disbursement, each under its own rule. A generic ledger that taxes the invoice total gets all three wrong at once.

Everything else in this chapter follows from these three facts.

Treat a retainer as what it is: client money held against future work. On the day it lands it is a liability, not a sale. It becomes revenue only as you earn it, invoice by invoice, and it returns to the client if you do not.

The discipline this requires is separation. The retainer balance is tracked apart from your earned fees, drawn down only when a real invoice is raised against it, and never quietly folded into the month's income because the bank account looks healthy. A retainer spent before it is earned is a debt you have already collected and may still have to repay.

The same discipline protects a proforma. A proforma is a request for payment, not a tax document and not revenue. It should sit outside your books entirely until it is deliberately turned into a real invoice, in one direction, by someone authorized to do it. Anything that lets a draft drift into your revenue by accident is a hole in the books.

Pass disbursements at cost. A payment you made for the client, recovered exactly, adds nothing to your margin and should add nothing to your taxable turnover. Record what you paid, recover what you paid, and keep the underlying receipt against the matter so the pass-through is provable.

Rule VAT per line, not per invoice. Most VAT regimes tax some services, exempt others, and treat certain disbursements as outside the tax altogether. When a statement can carry only one rate, you are forced either to split one piece of work across several invoices or to misstate the tax. A per-line rule lets a taxable fee, an exempt item and a disbursement share one clean statement, each taxed correctly.

Freeze the exchange rate at issue. On a multi-currency invoice the rate that matters is the rate on the day it was raised, and it must not move afterward. If the home-currency value drifts every time the market moves, your ledger will never reconcile and your tax base becomes a matter of opinion. The rate is a fact recorded at issue, not a live quote.

The single most important habit in a firm's bookkeeping is this: you correct by adding, never by deleting.

A payments ledger is the running record of money against your invoices: each settlement, each reversal, each adjustment. When something is wrong, the fix is a new dated entry that offsets the old one, not the quiet edit or removal of the original. A refund is a line. A bounced payment is a line. A miskeyed amount is a correcting line beside the mistake, which stays visible.

The reason is reconciliation. A bank statement is itself append-only: the bank never reaches back and erases a transaction. If your ledger is also append-only, the two reconcile by addition. Every line in the bank has a matching line in your books, and the totals agree because nothing was ever removed on either side. The moment you allow deletion, a balanced set of books stops being evidence of anything. It proves only that someone made the numbers agree.

A receipt sits at the end of this chain. Once a payment settles, the receipt that records it should carry its own serial and be sealed against change or deletion for everyone, including whoever runs the firm.

Your accountant should be able to close your month from three artifacts, handed over without a conversation.

The first is the books export: your invoices for the period, with per-line tax, in a format that opens in a spreadsheet. The second is the payments ledger: every settlement, reversal and adjustment, so cash can be tied to invoices and to the bank. The third is the receipts run: the sealed receipts issued in the period, each with its serial, so money collected has independent proof.

What your accountant should never have to ask for is the connective tissue: which payment cleared which invoice, what a given disbursement was for, why an invoice was reissued, or where a retainer was drawn down. If those answers live only in your head or in an email thread, the month cannot be closed without you, and it cannot be defended without you either. The point of good books is that they explain themselves.

A small firm can close a month in an afternoon with a fixed routine. Run it the same way every time.

• Confirm every invoice raised in the period is issued and serially numbered, with no gaps in the VAT, exempt or proforma books.
• Match every bank deposit to a payment entry, and every payment entry to an invoice; investigate anything unmatched before you continue.
• Confirm each disbursement recovered has its underlying receipt on the matter.
• Check that retainers received were booked as balances held, and that any drawn down were drawn against a real invoice.
• Verify that reversals and refunds appear as new correcting entries, not as edits to old ones.
• Confirm every settled payment produced a sealed receipt with its own serial.
• Reconcile the payments ledger total to the closing bank balance; the difference should be explainable line by line.
• Export the books, the payments ledger and the receipts run, and file the pack for the accountant and the archive.

The docket is the heartbeat of a practice. Hearings, filing deadlines and limitation dates are the points where inattention becomes malpractice, and they cannot rest on one person remembering. A private calendar is a single point of failure with a human attached. When that person is ill, travelling, on leave or gone, the dates leave with them.

One docket for the whole firm means every controlled date lives in the same place, anchored to the matter it belongs to, visible to those entitled to see it. Access still follows firm and role: restricted matters stay invisible to anyone outside their circle. But within that circle no date is anyone's private secret. Cover becomes possible because the record does not walk out of the building at five o'clock.

A firm that works across offices, or a lawyer who travels, learns quickly that a time without a place is ambiguous. A filing cut-off at four in the afternoon is not four in the afternoon everywhere. Read the same entry in two cities and you have two deadlines, one of them wrong.

The docket resolves this by showing each member the date and time in their own local time, computed from a single stored instant. The register holds one truth, and each reader sees it correctly translated. No one performs mental arithmetic across zones at the moment of highest pressure, which is exactly the moment such arithmetic fails.

A notification that arrives after the deadline is not a warning. It is a record of the failure, delivered too late to prevent it. The only useful reminder fires before the event and grows louder as the event approaches.

Escalation matters because a single quiet alert is easy to dismiss and forget. A limitation date should announce itself well out, again as the window narrows, and again in the final days, so that silence is never mistaken for safety. The rule is simple and absolute: reminders fire before, escalate toward, and never after.

Most deadlines can be recovered. A missed limitation date usually cannot. It ends the claim before any court weighs its merits, and no eloquence undoes it. A date that severe should not rest on one person's data entry.

Treat limitation dates as a distinct class. Enter them at intake, while the facts are fresh and the file is open. Have a second person verify the computed date against the underlying facts, because the person who calculated it is the person least likely to see their own error. Then alarm it early, with more margin than an ordinary deadline, so that if the first warning is missed there is still room to act on the second.

Run the docket as a discipline, not a diary. These habits keep it trustworthy:

• Every hearing, filing deadline and limitation date is entered on the firm docket, never a personal calendar.
• Each date is attached to its matter, so its context travels with it.
• Limitation dates are verified by a second person before the file leaves intake.
• Every date carries an owner and, where work is required, a task with a due date.
• Reminders are set to fire before the event and to escalate, with extra margin on limitation dates.
• When a date moves, it is changed once, at the docket, so everyone reads the new date rather than the old one.
• Nothing is closed from memory; it is marked done against the record on the matter.

Most firms do not decide how they communicate. It accumulates. A matter opens in email, migrates to a chat thread the first time someone needs a fast answer, then splinters across personal inboxes as people copy the colleagues they happen to remember. Within weeks the record of what the firm has promised a client is scattered across systems no partner controls and no auditor can reach.

This carries three costs, and none of them appear on an invoice.

• Commitments the firm cannot read. A lawyer tells a client, in a chat message at nine in the evening, that a document will arrive Friday. That promise now binds the firm, but it lives nowhere the firm can see. When the lawyer is on leave, the commitment is invisible until it is breached.
• Updates that consume billable hours. Where are we is the most expensive question in practice. Answering it requires a senior person to stop, reconstruct the state of a matter, and narrate it. The work has real value to the client and none to the firm, because it produces nothing that was not already known.
• Message archaeology before every call. Preparation degrades into search. Before a client call, someone scrolls threads to assemble a timeline that should have existed as a single record from the start.

The cure for where are we is not a faster answer. It is removing the need to ask. When a client can see the current state of the matter without spending a lawyer's time, the conversation changes character. It moves from status to substance.

A client portal makes the standing facts of a matter self-serve: filings, case status, invoices and sealed receipts, each presented in the client's own language. The client checks a filing date the way they check a bank balance, without an email and without waiting for someone to be free. What remains for the call is the work only a lawyer can do: judgment, strategy, the questions that have no form field.

This is not a thinner relationship. It is a better one. Time that was spent reciting the record is returned to the matter itself.

Email will not disappear from practice, and it should not. It carries the correspondence that matters most. The discipline is in the boundary, not the abolition.

The first rule is scoped intake. A mailbox sync should watch only a designated folder and a defined set of allowed senders, never the whole inbox. Nothing files to a matter automatically. A lawyer reads a message and attaches it to a matter deliberately, because relevance is a legal judgment, not a rule an algorithm should be trusted to make on its own.

The second rule is attributable output. When the firm writes to a client or a third party, it should write from the matter, on the firm's letterhead, from the lawyer's own connected mailbox, and the send should be logged to the matter. Outbound correspondence then becomes part of the record instead of a private act that only the sender can recall.

Modern practice is rarely monolingual. A client is entitled to read their own matter in their own language, and that entitlement is not satisfied by translation as an afterthought. It requires the register of law, not the register of the street.

Three demands follow. Documents reach the client in the client's language, not the firm's. Arabic is typeset in true right-to-left, so a bilingual invoice or letter reads correctly rather than as mirrored English. French carries the correct legal terminology, because the wrong word in a legal document is a liability, not a stylistic choice. A portal that shows a client their filings in a language they cannot read has solved nothing.

Every message has a right home. The test is simple: what is the message, and who needs to read it later.

Seniors tend to fail delegation in one of two directions. The first is hoarding: the partner keeps drafting, keeps filing, keeps billing, and the firm's capacity is capped at one person's working hours. The second is abdication: the partner hands out work and loses the thread, so nobody can say later who prepared a document, who released it, or whether anyone checked it at all.

The cure is not more trust or more meetings. It is boundaries that hold on their own. A role should see exactly what the role warrants, no more, and the seams between roles should be recorded as they are crossed. Delegation becomes safe only when work assigned is also work traceable.

Mersal defines four roles: partner, associate, paralegal and billing. Each sees a different slice of the firm, and the slice is not a preference set in the interface. It is enforced by Row-Level Security at the database engine, which filters every query by firm and by role before any data reaches the screen. There is no application-side path around it. A control that is only hidden can be found; a control enforced below the interface cannot be clicked around.

The point of the table is not the labels. It is that each row is a fence the software honors even when someone would prefer it did not.

Preparation and release are different acts, and they should be performed by different people. An associate can build a matter completely: parties, documents, the conflict pre-flight, the draft engagement. What the associate cannot do is activate it. The matter waits in a partner approval queue, and the role gate on activation is enforced by the database, not by convention.

When a partner releases a matter, the release is recorded. This does two things at once. It keeps a junior from opening files the firm has not vetted, and it gives the partner a defensible record that the decision was theirs, on a date, in the ledger. The associate prepares. The partner releases. The record remembers both.

A task without an owner is a wish. Mersal puts tasks on the matter timeline with an owner and a due date, so an assignment is visible in the same place as the correspondence, hearings and documents it relates to. Anyone with access to the matter can see what was asked, of whom, and by when.

This is what turns delegation from a verbal handoff into a governable act. The docket reinforces it: hearings, filing deadlines and limitation dates sit in each member's local time, and reminders fire before the event and escalate as it approaches, never after it has passed. Work is assigned in the open and tracked to its owner.

Some files must not merely be private. They must not exist for most of the firm. A lateral hire's old client now adverse to a current one, a partner's own dispute, a matter where a colleague is conflicted: for these, ordinary permissions are not enough, because a file a person can see they are barred from is a file they know about.

Restricted matters are invisible to anyone outside their circle. Not greyed out, not access-denied, simply absent from every query that a person outside the wall runs. Because the filter lives at the database and not in the interface, the wall does not depend on a screen behaving correctly. An ethical wall that can be seen around is not a wall.

A closed matter is not a deleted matter. Closing means the active work is finished. It does not mean the record can lapse. In many jurisdictions a firm's duty to its client, and to its own defense, begins in earnest at exactly the moment the work stops.

Files are requested years after the last hearing. Tax authorities audit backward. A former client, or their successor in interest, asks for their own papers. Archivage is the discipline of closing a matter so that its record stays whole, legible and retrievable long after everyone who touched it has moved on. The matter is archived. Nothing about it is discarded.

An archive that keeps only the final documents is not an archive. It is a summary, and a summary cannot answer the questions an audit or a dispute will ask. The whole record must survive closing:

• The full timeline: every task, hearing, filing and event in the order it happened, with its owner and its date.
• Documents and their versions: not only what was signed, but what was drafted, revised and superseded.
• The fiscal trail: every invoice, receipt and ledger entry, each serial intact and unbroken.
• Correspondence: what was sent, by whom, and when it was delivered and read.
• The conflict clearance: the pre-flight search that let the matter open, and any partner waiver that accompanied it.

Software changes. Vendors are replaced, formats fall out of use, contracts end. The retention obligation bends to none of this. Many bars require client files to be kept for years after a matter closes. Many tax regimes require fiscal records to be retained for their own, often longer, period.

A record you cannot read in ten years is not retained. This is why export in open, portable formats is a requirement and not a convenience. The firm must be able to lift an entire matter, its documents and its fiscal history out of the system and hold it in forms that need no particular vendor to open. Ownership of the record means the ability to leave with it.

Lawyers move. Partners retire, associates leave, teams reorganize. When a matter lives in one person's inbox and private drive, their departure is a data loss the firm may not notice until it needs the file.

The remedy is structural. The matter must belong to the firm's record, not to any individual's mailbox. Correspondence attached to the matter stays with the matter. Documents filed to the matter stay with the matter. When a lawyer leaves, the successor inherits a complete workspace, not a reconstruction assembled from memory and forwarded emails. Handover stops being a scramble and becomes a reassignment.

A well-kept archive is not only a defense. It is the firm's memory of itself. The argument that prevailed three years ago is precedent for the matter opening today. The clause negotiated for one client is the starting point for the next. The archive is where a firm's institutional knowledge accumulates instead of walking out the door.

It is also the firm's proof of its own history: what it advised, when, for whom, and on what basis. A firm that can produce its record can defend its judgment. A firm that cannot is arguing from memory.

Run the same checks every time a matter closes, so that closing produces an archive and not a gap:

• Confirm the timeline is complete, with no open tasks and no unrecorded hearings or outcomes.
• Confirm every document version is filed to the matter, not left in a personal drive.
• Reconcile the fiscal trail: invoices issued, receipts sealed, ledger balanced, serials unbroken.
• Confirm correspondence is attached, with its delivery and read record intact.
• Confirm the conflict clearance and any partner waiver are on file.
• Reassign ownership away from any departing member to a current custodian.
• Produce a full export in open formats and store it under the firm's retention schedule.
• Mark the matter archived, then verify its record stays readable and exportable.