What this is
A community-run record of what happens after a 221(g) at a US consulate — how long administrative processing actually runs, what gets asked for, and how cases end. It exists because there is no public data on this and people in the middle of it are guessing.
What this is not
It is not legal advice and cannot predict your case. It is also not a random sample — a six-month wait is memorable and a two-week one is not, so slow cases are over-represented. Treat everything here as a rough signal about a community, not a base rate.
Why there is no "average processing time"
Every forum quotes one. It is almost always wrong, and wrong in a specific, predictable direction.
Suppose 100 people interviewed a year ago. Fifty were approved within two months. The other fifty are still waiting today.
Average the cases that finished and you get about two months — and that number is worthless, because it silently discards the fifty people whose wait is the whole problem. Worse, it gets better the more people are stuck, because anyone still waiting simply is not counted.
The honest answer to "how long?" is not one number. It is "half of this group is still waiting at six months" — which is what the curve on the results page shows. Cases with no decision stay in the denominator until the day they resolve.
P50, P80, P90, P95
The results page gives percentiles — "80% of cases resolved within N days" — but reads them off the survival curve rather than from the list of finished cases. That matters twice over.
Statistically, a percentile taken over finished cases only ignores everyone still waiting, and understates every figure. The damage is worst at exactly the numbers people care about: if a third of cases are still open, a "P90" computed from the rest is not a P90 at all.
For privacy, a raw P95 over forty cases is one specific person's exact wait. Read off the curve it inherits the 30-day resolution and the small-cell cutoff, so it never lands on an individual's number.
Where the data has not gone far enough to reach a percentile, the page says "not reached" rather than estimating one. With many cases still open the high percentiles genuinely are unknown, and a guess dressed up as a statistic would be worse than an honest blank.
Small numbers are shown, not hidden
Many published statistics hide any group smaller than five. This site deliberately does not, and the reasoning decides how much you should trust the early numbers.
That convention exists to defeat linkage attacks — joining an "anonymous" dataset against another one to work out who is who. The famous demonstration used ZIP code, date of birth and sex against voter rolls, and it works because those fields exist in both. Nothing stored here is like that: consulate, visa category, 221(g) reason and outcome appear in no external table anyone can join against. The only attack left is someone who already knows your situation using the page to confirm it, and a threshold of five would not stop that either — they would look at a group of five and still know you were in it.
Hiding those numbers also had a real cost: the site showed nothing at all until ten cases had been submitted, which is exactly when a tracker most needs to look worth contributing to.
Mind the denominator. Early on a "67%" may be two cases out of three. Every percentage here carries the count behind it — read that first, and treat anything under about thirty cases as a hint rather than a finding.
What is still limited, and why
Two limits stay, but for statistical reasons rather than privacy ones — a figure that is meaningless is worse than no figure:
- Durations appear as bands ("121–180 days"), never exact day counts. Thirty days is about the precision self-reported recollection can support.
- The survival curve stops when fewer than five cases are still being tracked. The tail of such a curve is computed from a handful of people, and a single case resolving would swing it wildly. Percentiles read off the truncated part are reported as not reached rather than guessed at.
The real protection here is the data model, not arithmetic — above all the fact that no calendar dates are ever collected. A 221(g) timeline with real dates genuinely is a fingerprint, and many of these cases were posted publicly in forums with those exact dates. Day-offsets and an interview month are not. That, along with no names, no employer, no case number, no IP address and no free-text box, is what keeps this anonymous.
What is collected
Eight multiple-choice answers, and optionally a timeline. There is no free-text box anywhere on this site, deliberately — a comment field is the one place people identify themselves.
Never collected:
- No name, email, or account of any kind
- No employer, attorney, receipt number or case number
- No calendar dates. You enter dates; your browser converts them to a number of days and sends only that, plus the month you interviewed. A timeline with real dates is a near-unique fingerprint — and many of these cases were posted publicly in forums with those exact dates, so storing them would let anyone match the two.
- No IP address. Abuse limiting uses a hash salted with a value that rotates every 24 hours, and those rows are deleted on expiry.
- No analytics, no third-party scripts, no cookies.
Your case stays yours
On submission you get a private link containing a random token — the only way back to your case. Only a one-way hash of it is stored, so it cannot be recovered or looked up, by us or anyone else. Use it to add milestones as things move, or to delete your case.
Deleting removes your case from the live dataset and from every published number. In the interest of being straight with you: the database is backed up privately, so a copy of an already-deleted row can persist in an old backup for a while. Nothing from it is ever published again.
Nothing can remind you to come back. With no email and no account there is no way to reach you — that is the cost of real anonymity, and it is deliberate. The site can only remember the link in your own browser, which disappears if you clear site data or switch devices. Save the link yourself.
Where it runs
On Cloudflare Pages and a Cloudflare D1 database. Cloudflare, as the network carrying the request, sees your IP in transit the way any website's host does — this application simply never writes it down.