The Form 1 Backlog Broke

What this is

On July 21 this series published The Queue That Split. Its central finding was that Form 1 waits were not frozen but sloping, accruing roughly eleven days for every calendar month that passed, in a line straight enough that an ordinary least-squares fit described it well. That edition closed by telling anyone in the Form 1 queue to plan on the number continuing to move. It noted the climb would stop when capacity or demand changed, but said nothing in the data showed it stopping yet.

Thirteen days later, the data shows it stopping.

This edition is short, and it hedges more than usual, because the whole finding rests on about three weeks of approvals. That is a thinner base than we normally publish on. We are reporting it anyway for one reason: the last edition told readers to expect the opposite, and a dated series that waits for comfortable evidence before correcting itself is not worth running. What follows is the break, the four checks it had to survive, and an honest account of what could still be wrong with it.

All figures come from 3,496 approvals reported by NFA filers, snapshotted August 3, 2026. Community data, not official ATF records. The method note at the end covers what was excluded and why.

The break

Group the approvals by the week ATF cleared them, keep only complete weeks, and the Form 1 Individual median reads like this across the last eleven weeks: 65, 64, 69, 71, 72, 40, 82, 83, 73, 48, 22.

Set the 40 aside for now. It is a collection artifact and gets dealt with below. Without it, the lane holds a band between 64 and 83 days for eight weeks, drifting slowly upward exactly as the July edition described. Then, in the week of July 13, it turns. Three weeks later it has fallen by sixty-one days.

The daily numbers show the same thing with no weekly smoothing. Each day's Form 1 Individual median, in order from July 13: 79, 77, 54, 74, 70, 76, 74, 65, 56, 48, 30, 41, 37, 23, 34, 7, 55, 23, 22, 16, 16, 14. That is a glide rather than a step. No single day carries the move, and no single day could be pulled out to undo it.

Form 1 Trust is the more interesting case, and it is why this edition never once quotes a combined "Form 1" number. Its weekly medians read 66, 69, 74, 76, 71, 83, 83, 86, 73, 84, 50. For ten of those eleven weeks Trust does what the July edition said both Form 1 lanes were doing. It climbs. Only in the final week does it turn, and it turns on fourteen approvals. Form 1 Individual moved first, moved further, and moved on roughly four times the evidence. Whether Trust is following it or just wobbling is not something three weeks and fourteen records can answer.

What the queue was actually working

The mechanism shows up in which filings were getting cleared, week by week.

Through mid-July the Form 1 Individual lane was still working filings placed in March, April and May. In the week of July 13 it cleared nineteen April filings and thirty-two May filings. The week after that, April all but disappears and June becomes the largest group. In the week of July 27, forty-five of the fifty-three approvals were filings placed in July, the same month they were approved.

That is what a backlog draining looks like from the outside. A queue that has been working four-month-old filings starts working same-month filings because the four-month-old ones have run out. Reported Form 1 Individual approvals also roughly doubled over the same stretch, from the low twenties per week through early July to sixty-three, seventy and fifty-three in the three weeks that followed. Form 4 volumes stayed flat.

Four checks

A drop this large on a sample this fresh deserves suspicion before it deserves print. Four things could produce it artificially. We tested each one.

Could it be filings that have not been approved yet? No, and this is the strongest single piece of evidence. Counting only approvals that waited thirty days or less removes the problem completely. A thirty-day wait cleared last week was filed a month ago, so nothing about it depends on filings still sitting in the queue.

Across the first eight weeks shown, Form 1 Individual produced between zero and two such approvals per week, apart from the Virginia week, where it produced twelve. In the week of July 13 it produced nine, then thirteen, then forty-four, out of weekly totals that never topped sixty-eight. Form 1 Trust's equivalent counts are zero in eight of the first ten weeks, then six in the last.

Could it be that slow approvals just get reported later? This one is a real effect, and it runs in exactly the direction that would fake the finding. Among Form 1 Individual records old enough to have fully settled, approvals that waited thirty days or less reach us a median of one day after approval. Those that waited longer take four. So a fresh week shows its fast records first and looks quicker than it will once everything arrives.

The fix is to hold the reporting window constant: measure every week using only records that reached us within three days of approval, so no week gets a longer collection period than any other.

Under that constant window the Form 1 Individual series reads 64, 65, 70, 71, 70, 40, 83, 83, 73, 48, 22. The break is unchanged. Reporting lag is real, it does bias fresh weeks fast, and it is not what produced this.

Could it be a batch of filings from one party? Partly, in one specific week, and it is worth reporting because it shows the check working. The dip in the week of June 22 is not a queue event. Twenty-one of that week's twenty-eight Form 1 Individual approvals came from Virginia, twenty of them reaching us on a single day. Take Virginia out and that week reads 69 days instead of 40, in line with the weeks on either side. Take Virginia out of the whole series and the late-July break is untouched: 65, 65, 71, 71, 73, 69, 82, 83, 73, 48, 22.

The late-July weeks show no such concentration. The week of July 27 spans eighteen states with no state above twelve records, and its approvals reached us across seven separate days.

Could the last week simply be too fresh? Yes, and this is the one caveat we cannot clear. Roughly a third of approvals reach us within a day, half within three days, two-thirds within a week. By that curve the week of July 6 is about 95% reported, the week of July 13 about 80%, the week of July 20 about 68%, and the final week of July 27 only about a third. The 22-day figure for that last week will rise as the rest of its records arrive. The most anyone should take from it is direction.

The claim that does not lean on the freshest data is the week of July 20: 48 days, seventy approvals, roughly two-thirds reported. Against 83 days two weeks earlier, that alone is a thirty-five-day fall on a well-settled base.

Where the lanes stand

Over the last two complete approval weeks, Form 4 Individual sits at a median of 5 days, Form 4 Trust at 31, Form 1 Individual at 33, and Form 1 Trust at 78.

The middle two deserve a closer look. Three weeks earlier, Form 4 Trust stood at 32 days and Form 1 Individual at 59. They are now 31 and 33, with confidence intervals that overlap across most of their range, so on this evidence the two lanes cannot be told apart. We are not claiming Form 1 has become the faster lane. What we can say is that the ten-to-one spread the July edition described has closed to roughly fifteen-to-one at the extremes, and to nothing at all in the middle of the table.

For anyone waiting, the practical reading has changed for Form 1 Individual filers, where recent approvals are clearing far faster than anything else this year would have suggested. It has not changed for Form 1 Trust filers, where the evidence of a turn is one week old and rests on fourteen records.

What would explain this, and why we are not choosing

Several things would produce the pattern above. Processing capacity may have increased. Something in how these filings get handled may have changed. Or the demand surge that followed the $0 tax stamp on January 1 may simply have finished passing through, leaving a queue that had been sized against a backlog with nothing left to work.

That last one fits the shape of the drain. A backlog emptying rather than a queue speeding up would look much the same from where we sit. But fitting is not the same as demonstrated. This dataset observes approvals and never applications. Without the intake side, a capacity story and an exhaustion story are indistinguishable, and we are not going to pick between them on three weeks of output data. The same limit applied when the July edition declined to claim the $0 stamp had caused the divergence, and it applies just as much now that the movement has reversed.

What we cannot see

Three limits bear directly on how much weight this finding can carry.

We cannot separate Form 1 by what is being made. A Form 1 to build a suppressor and a Form 1 to make a short-barreled rifle are not obviously the same piece of work, and nothing in our data tells them apart. If the acceleration is concentrated in one item type rather than spread across all of them, this edition would have no way to know. It also means a filer submitting many near-identical Form 1s, same design, different serial numbers, contributes many records without contributing many independent observations of how the queue behaves.

We can only partly detect batch filings. The Virginia cluster above got caught because it concentrated in one state and one collection day. A single party filing across several states over several weeks would leave no such signature, and we would count it as many unrelated filers. The fields that would settle this are not available to us at the scale required.

We observe a sample, not the queue. These are approvals reported by filers in the NFA community. The people who report are not a random draw from everyone who files, and the sample can shift in composition without the queue changing at all.

What happens next

We will reassess at the end of August, once the July and August cohorts have matured and the reporting lag on the weeks above has fully cleared. That assessment will report what it finds, including, and especially, if it finds the drop did not hold.

Three questions carry forward:

• –Does the Form 1 Individual level settle near the low twenties, or drift back up once the backlog-clearing effect passes and the lane is working only current filings?
• –Does Form 1 Trust follow, on evidence thicker than fourteen records?
• –Does Form 4 Trust keep going? It has climbed slowly and without reversing all year, from 10 days in January to 31 now, and it has drawn the least attention of the four lanes while being the most consistent.

Early evidence that a queue is clearing faster is worth reporting when the same series recently told readers to expect the opposite. It is not yet worth treating as the new normal.

Method and limits

Figures come from nfa_watch.approvals as of August 3, 2026: 3,496 records carrying both a filing date and an approval date. One record with a 977-day wait was excluded as implausible; the next longest in the set is 120 days. Eighty-three records were excluded as duplicates, on the rule that a duplicate requires both an identical (form, owner type, state, filing date, approval date) tuple and an identical masked control number. Our control numbers are truncated for privacy, so matching on the mask alone would have thrown out 536 genuinely distinct filings.

Medians group by approval week, the output side of the pipe, meaning what actually got cleared each week. That is a different view from filing-cohort curves and should not be read as one. Medians are rounded to whole days. Only complete weeks appear in the charts; the partial week after August 2 is excluded. Confidence intervals are bootstrapped from the observed records and describe sampling uncertainty only, not the gap between this sample and the queue as a whole.

Charts on this page are locked to the August 3, 2026 snapshot and will not move. The live versions of these studies, which update continuously, are in the Lab.