74 transactions, one description
We pulled a creator's Apple Cash activity for the first 7 months of this year. 74 transactions. 72 going out, 2 coming back. Net of the returns, $12,601.
Every single outbound row carries the same description in the bank feed:
APPLE CASH SENT MONEY
No name. No memo. No account. The only two facts that survive about any of this money are the date it moved and how much of it moved.
This is business spending. It paid editors, covered meals on shoot days, reimbursed people for things they bought, and settled up after productions. None of that is in the record, because the record has nowhere to put it.
What the shape tells you, and where it stops
You can learn a surprising amount from two columns, and then you hit a wall.
The median send is $100. The largest is $1,500. 22 of the 72 land between $100 and $250, and 15 more between $250 and $500, which together account for most of the money. 46 of the 72 are exact multiples of $25, worth $10,150 in total. Round numbers are how people pay people. The 26 that are not round, worth $2,648, look more like someone covering an actual receipt.
15 separate days carry 2 or more sends. One day in March carries 4. Those cluster days read like shoot days, and shoot days are precisely the spend that should attach cleanly to a production.
Now the wall. Every one of those inferences is a guess. Nothing in the data says who was paid, and nothing says what for. $200 on a Tuesday is an editor, a lunch, a returned deposit, or a friend paying back a concert ticket, and the feed treats all 4 identically.
The question you cannot ask
Here is the part that matters more than the missing write-offs.
If you pay an individual for services and the total crosses a threshold over the course of the year, you owe them a 1099 and you need a W-9 on file to produce it. For payments made in 2026 that federal threshold is $2,000, raised from the $600 figure that had stood since 1954. Payments made in 2025 are still governed by the old number.
Look at the data and the obvious conclusion is that nothing here comes close. The largest single send is $1,500. Nothing crosses the line.
That conclusion is wrong, and the reason it is wrong is the whole point.
The threshold is cumulative per person per year, not per transaction. 6 sends of $350 to one editor is $2,100 and a filing obligation. 6 sends of $350 to 6 different people is nothing at all. Same rows. Same total. Opposite outcomes. And this record cannot tell the two apart, because it does not contain a single name.
So the honest answer to whether this creator owes anyone a 1099 for this money is not yes and it is not no. It is that the question is unanswerable from the record as it exists. That is a worse position than owing one, because at least an obligation you know about can be met.
Why this keeps happening
The temptation is to say the mistake was using a consumer payment rail for business. That is true and it is useless, because it ignores why it happened.
At the moment of spend, on location, the fastest way to move $200 to the person standing in front of you is the one already on your phone. It takes 4 seconds. Anything that takes longer loses, every time, to the thing that takes 4 seconds. That is not carelessness, it is a rational response to the options available, and any advice that begins with "stop doing that" is advice that will be ignored by lunchtime.
The same logic explains why the record is empty rather than partial. Apple Cash is built to move money between people. It is not built to record why, because for its intended purpose the why is obvious to both parties and interesting to neither. The information does not go missing later. It never gets captured at all.
What to do about it
Reconstruct now, while it is recent
7 months of memory is imperfect. 19 months is gone. Work backwards from the cluster days first, since those tie to specific shoots and are the easiest to reconstruct, and start with the largest sends. 20 recovered names out of 74 is not a failure. It is the difference between an unanswerable question and a mostly answered one.
Find the repeat payees first
The filing question turns on people who were paid repeatedly, not on the single largest transaction. If the same editor appears 6 times, that is the name worth recovering before any other. Sort your reconstruction by who shows up most, not by what cost most.
Collect the W-9 at the start, not in January
Asking someone for their tax details when you hire them is routine. Asking in January, after the year has closed, for work they barely remember, is a favor you are now asking of someone who has no reason to grant it. The cost of collecting early is a 2 minute conversation. The cost of collecting late is sometimes not being able to.
What we would not tell anyone to do is abandon the rail. The money moves that way because that way is fastest, and a system that only works when the creator uses the approved payment method is a system that will be abandoned within a week. The capture has to happen where the spending happens, at the speed the spending happens, or a parallel record in a notes app wins and everything drifts back into it.
The part worth being precise about
This money is not lost. That distinction matters and it gets muddled constantly.
Whether an expense is deductible does not depend on which app moved the money. It depends on whether the expense was ordinary and necessary for the business, and on whether you can substantiate it. A payment to an editor is a business expense whether it went by bank transfer, a card, or a person-to-person app. What changes across those rails is not eligibility. It is what evidence stands behind the claim.
A transaction from a connected bank account carries independent confirmation of the amount, the counterparty and the date. A line someone typed carries a person's word and, if you are lucky, a screenshot. Both can be entirely true. Only one is corroborated by a third party. Keeping that distinction attached to each line, rather than as a disclaimer at the top of a summary, is what lets a preparer treat the well-evidenced items as well evidenced instead of discounting everything equally.
So the accurate description of this $12,601 is not that it vanished. It is that it went unclaimed, because nobody labeled it. Unclaimed is recoverable. That is a materially better position than it first appears, and it is worth saying plainly, because the version of this story where the money is simply gone is the version that stops people from going and getting it.
The general case
One creator, one rail, 7 months, $12,601. Scale that thought across a working year, across the other rails that behave the same way, and across the number of people running real businesses off consumer payment apps, and the size of the missing record starts to come into focus.
The spending happened. The work happened. The people got paid. The only thing that failed is the part where any of it got written down in a form that a preparer, a lender, or the creator themselves could act on six months later.
That is not a discipline problem and it will not be solved by telling creators to be more organized. It is a capture problem, and it gets solved at the moment the money moves or it does not get solved at all.
Figures are from a single creator's 2026 year-to-date Apple Cash activity, shared with permission. ARCA is a financial operating system built for full-time creators. Nothing here is tax advice; thresholds and filing obligations vary by situation and are worth confirming with your preparer.