Two decisions that get treated as one
A courier charge lands on the firm's card. Somebody codes it to a matter, and the job feels finished. It is not: coding it to a matter recorded where the cost belongs, not who bears it.
Those are separate facts. A cost can be unambiguously attached to a matter and still be the firm's to carry — travel to a client meeting is the textbook case, and firms differ on it. Another cost can be recoverable and yet have no matter attached, which is how a genuinely billable expense ends up in general overheads.
Collapsing the two produces a specific failure. The cost is coded, the coding looks complete, and nobody ever asks the second question — so the default answer becomes whatever the accounting system does with an uncoded expense, which is to absorb it.
The flag exists to stop that default from being silent. It is one field and a short reason, sitting on the same row as everything read off the document, and its whole purpose is to make an unanswered question look unanswered.
Not the same thing as dimension tagging
FlowParse already has a feature for attaching a cost to a project, matter, site or fund — it is called dimension tagging, and it answers a different question from this one. Since the two are used together, the boundary is worth being precise about.
| Dimension tagging | Recoverable flagging | |
|---|---|---|
| Question it answers | Which matter does this belong to? | Does the client pay for it? |
| Who usually answers | Whoever books the cost | The fee earner or the engagement terms |
| Changes over time | Rarely — a cost belongs where it belongs | Sometimes, on review before billing |
| Affects the accounts | Where the cost is reported | Whether it becomes revenue |
| Blank means | Unallocated — findable and fixable | Undecided — and therefore unbilled |
The last row is the practical difference. An unallocated cost is a tidiness problem that somebody will notice at a period end. An undecided one is money, and nothing in a normal month will surface it — which is why it gets its own field rather than living as a note in a description.
What the flag carries
| Field | Values | Why it is separate |
|---|---|---|
| Recoverable | Yes · No · Not decided | Three states, because two would hide the third |
| Reason | Free text, one line | The only thing that survives a query a year later |
| Amount to recover | Number, defaults to the document's | Partial and cost-plus recharges both exist |
| Matter reference | Your own reference | Set by tagging; shown here so both are visible at once |
| Decided by | A person | Because the query goes to a person, not to a field |
| Decided on | A date | Distinguishes a decision from a default |
The three-state flag is the design decision that matters most here. A boolean forces every unseen cost into one of the two answers, and whichever way it defaults, the firm loses the ability to tell decisions from omissions. Three states cost nothing and keep the difference.
“Amount to recover” is separate from the document's amount for the same reason: a cost recharged at half or with a handling percentage still has to keep what the supplier actually charged, because that is the figure the evidence will show.
What can be read from the document
Everything a supplier knows is on their invoice, and it is a surprisingly complete picture of the cost itself.
Who charged it, with a registration number where they have one — the reliable identifier, since a name is written three ways across a year.
What it was, in the supplier's own words rather than a paraphrase — which is what the client will be shown if they ask.
When: the document date, and the service date where they differ, which on fee notes is often by weeks.
How much, split into net, VAT by rate, and gross — kept apart so the treatment stays open.
Line detail where the document has lines, which is what turns a bulk statement into individual decisions.
The file and page it came from, which is added rather than read, and is what makes a row defensible.
All six are mechanical. None require knowing anything about your firm, your matters or your engagement terms, which is exactly why they can be taken off a person's desk entirely.
What cannot be read, at any price
Which of your matters it belongs to
A search provider knows a property address, not your file reference. Where the reference does appear it is captured; where it does not, the field stays empty rather than being inferred from a date and a hope.
Whether your client agreed to bear it
That lives in an engagement letter, a scoping call or a client's expectations. None of them are on the invoice.
Whether your firm's policy allows it
Firms differ on travel, printing and out-of-hours costs, and many have never written the policy down. A tool cannot discover a policy that does not exist.
What your regulator requires you to disclose
This varies by profession and jurisdiction and changes. It is a question for your compliance lead, and it is not one a document reader should ever appear to answer.
The temptation with all four is to produce a plausible answer and let the user correct it. That trade is worse than it looks: a field somebody has to verify costs more attention than a field they have to fill, because filling it is a decision they were going to make anyway and verifying it is a decision plus a check.
Defaults, where they are safe
Some suppliers are unambiguous. A court fee provider is recoverable every single time; a coffee shop is not. Requiring a decision on those is friction with no judgement in it, and friction with no judgement is what causes people to stop flagging altogether.
So a supplier can carry a default, applied when its documents are read, and shown as a default rather than as a decision. The row still records that nobody looked at it — which keeps the distinction between an automatic answer and a considered one.
The line for setting a default is narrow on purpose: it belongs to suppliers where the answer has been the same every time for as long as anyone can remember, and nowhere else. Set one on a supplier that is recoverable eighty per cent of the time and the twenty per cent stops being seen, which is a more expensive mistake than the typing it saved.
Defaults are also worth reviewing once a year rather than never. A supplier's role in a firm changes — a printing account that used to be client bundles becomes internal marketing, and the default quietly starts recharging clients for the firm's own brochures.
The moment the flag gets set
There are three plausible moments, and they are not equally good.
When the cost is incurred. The person who ordered the search knows exactly why. This is the cheapest possible moment and the least likely to happen, because they are mid-task and the cost is not yet a document.
When the document is processed. Someone in finance has the invoice in front of them and can see the supplier, the amount and the date — but usually not the context that decides recoverability.
When the bill is prepared. Everything is visible at once and the decision is easy, but it is also the latest possible point: anything missed here is missed permanently, because the bill goes out.
In practice the workable arrangement is the second and third together — finance flags the clear cases and leaves the rest undecided, and the fee earner clears the undecided list as part of preparing the bill. That works only if “undecided” is a real state that produces a real list. With a two-value flag there is nothing to clear, which is why the third state exists.
One document, many matters
The bulk supplier statement is where flagging either earns its place or fails completely. A monthly account from a search provider, a courier or a printing firm covers dozens of jobs across dozens of matters, and arrives as a single document with a single total.
Treated as one thing, it becomes one posting to one code, and every recoverable cost inside it stops being recoverable at that moment. Nobody decided that; the document's shape decided it.
Read as lines, the same document is forty rows, each with its own date, reference and amount, and each able to carry its own matter and its own flag. Two of them might be internal and thirty-eight recoverable — a distinction that is invisible in the total and obvious in the lines.
This is also the case that most rewards the source reference. When a client queries one line of a bill, the answer is a specific line on a specific page of a specific statement, not the whole month's account from that supplier.
The one-line reason
A flag without a reason answers the question for as long as the person who set it remembers why. That is shorter than most people expect and considerably shorter than the life of a matter.
The reason does not need to be careful prose. “Agreed at scoping call”, “outside fixed fee”, “our error, absorbing it”, “client asked for next-day” — four to six words, written once, at the moment the decision is obvious.
It earns its keep in two situations. When a client queries a line months later, it is the difference between an answer and an afternoon. And when the same question comes up on the next matter with the same client, it is how the firm answers it the same way twice instead of differently.
The second is the one firms underrate. Inconsistency between fee earners on what gets recharged is a source of client friction that nobody can trace, because each individual decision was defensible and no two were the same.
A waiver and a leak look identical afterwards
This is the argument for the whole feature, so it is worth stating plainly.
A cost that the firm decided not to recharge and a cost that nobody remembered to bill produce exactly the same accounting outcome: an expense the firm carried. In the ledger they are indistinguishable. In management terms they could not be more different — one is a commercial choice, the other is a process failure — and they have opposite remedies.
A firm that cannot separate them tends to respond to the total by tightening policy: fewer things may be waived, more must be recharged. If the real problem was receipts that never arrived, that response irritates clients and fixes nothing, and the money keeps going.
With the flag in place, the quarter's number splits in two. “We chose to absorb this much” is a business decision to review. “This much was never decided at all” is a process to fix. Only the second one has a fix.
Reviewing before the bill goes
The list that matters at billing time is short and specific: costs on this matter, not yet billed, with the undecided ones at the top.
That ordering is deliberate. The decided ones need a glance; the undecided ones need the fee earner, and they are the only reason to open the list at all. A screen that shows fifty rows in date order buries the six that need a person.
Two other things are worth surfacing at the same moment. Costs dated after the last bill on a matter that has already been billed once — those are the late arrivals that are usually written off by inertia. And any cost with no document behind it, because a bill line whose evidence cannot be produced is the one that will be queried.
None of that requires a practice management system. It requires the costs to exist as rows with dates, flags and a matter reference — which is the entire premise of disbursement tracking.
What you get back
One row per cost, or one row per line where a document has lines, with everything read from the document plus the fields you added: matter, flag, reason, amount to recover.
Excel to review in, CSV to import into whatever holds your matters, JSON for anyone with a process behind it, XML where a system asks for it. Up to 100 files per pass, so a month of cost documents is one table.
The source file and page ride along in every format. It is the column nobody asks for at import time and the only one that matters when a client asks where a number came from.
When two people flag the same cost differently
A cashier marks a £220 courier charge recoverable because that is what the practice normally does. The matter partner marks it absorbed because this particular client was promised no extra charges in a phone call three weeks ago.
Both are behaving correctly with the information they have. The disagreement is not a data-quality problem; it is a fact about the engagement that existed only in a conversation.
Systems handle this badly when they treat the flag as a value that gets overwritten, because the second person's edit erases the evidence that anyone thought otherwise. What survives is a single flag with no history, and the next similar cost repeats the same collision.
The behaviour that works is duller: the later decision wins, the earlier one stays visible, and the reason line carries whichever justification was given. Nobody needs an approval workflow for this — they need to be able to see that two people looked at it and why the answer changed.
The pattern to watch for is repetition. One disagreement is a fact about one client; the same disagreement four times in a month is an engagement term that was never written down, and that is worth fixing at the template rather than at the row.
What the flag looks like in aggregate
Individually the flag is administrative. In bulk it is the most diagnostic field a practice has, because it is the only one that records an intention rather than a fact.
| Pattern | What it is really telling you |
|---|---|
| One cost type almost always absorbed | An unwritten policy. Write it down or start charging for it — either is better than deciding it 300 times a year. |
| One person's rows disproportionately absorbed | Usually not generosity. Usually a different understanding of what the engagement allows. |
| A large undecided pile on one client | Somebody is uncomfortable about the relationship and is deferring rather than deciding. |
| Undecided rows older than sixty days | Waivers that nobody has admitted are waivers yet. |
| Flags set weeks after the cost date | The document is arriving late, and the flag is being set from memory rather than context. |
The first row is the one that pays for the whole exercise. A practice that discovers it absorbs every courier charge has found a real annual number and a decision it can actually make — and it could not have found that from a ledger, because the ledger records the payment and not the intention behind it.
The last row is subtler and worth watching. A flag set at the moment the cost is incurred carries real context; the same flag set five weeks later carries a guess. If most flags are late, the problem is upstream in how documents reach the practice, and no amount of care at the flagging step will repair it.
What it will not do
It does not raise bills, hold a matter ledger, record time, or take any part in client money — that last one is a regulated activity and this is a document reader.
It does not tell you what your profession permits you to recharge, and it should not appear to. Those rules differ by regulator and by country, and a tool that produced an answer would be producing one with somebody else's liability attached.
It does not decide VAT treatment. Net, VAT and gross are kept exactly as the document stated them so that whichever treatment is correct can be applied without reopening anything.
And it does not chase people. A cost that never reached the firm at all is invisible here, which is why the practical companion to this feature is comparing the card statement against the documents — the one check that finds what is missing rather than improving what is present.
The first month
Do not start by flagging everything. Start by reading one month of cost documents into rows and looking at what comes out, because the first useful discovery is usually the shape of the pile rather than any individual decision.
Then set defaults for the two or three suppliers where the answer has never once been in doubt. That covers a surprising share of the volume and leaves a much shorter list of things that actually need thought.
Then flag the rest as they are billed, and let “not decided” accumulate for a month without worrying about it. At the end of that month the undecided list is the most informative thing the firm has produced about its own billing — it is, precisely, the set of costs nobody has an answer for.