Method and past error
Every figure on this site comes from three public sources, read by programs that can be run again, and every forecast is graded against what really happened. This page explains all three and publishes the full result, including where we do badly.
The Visa Bulletin archive
We read the 294 bulletins published from December 2001 to October 2026. The State Department’s main host turns away automated readers and we do not try to get around that: the reading is done on adoptions.state.gov, which serves the same pages from the same department. The reader copes with four eras of the bulletin: the 2002 tables with no China column, rows with typos (“4rd”, “Charge- ability”), the arrival of the two charts in October 2015, and the four names the EB-5 rows have carried. A country without its own column follows the worldwide date, as the bulletin itself says. If a single cell is missing, the site rebuild stops.
The forecast
For someone G days behind the cutoff of their case, we look at every month of the previous 10 years and measure how many months that same cutoff took to move G days forward, retrogressions included. Months whose outcome is not known yet are handled with the Kaplan-Meier estimator, counted neither as a success nor as a failure. From that distribution we take three points: the fastest 20% of cases, the median, and 80%. A first model, which extrapolated the average pace of advance, was dropped at the back-test: its range contained the real date in only 23% of cases.
The back-test and the publication rule
From October 2012 to October 2025, for every case and every month, we make the forecast using only the bulletins known that month, for people 6, 12, 24 and 48 months behind the cutoff, and compare it with the month their date really became current. A forecast is published only if, for that case at that distance, the range held the real date at least 40% of the time, the median error is no more than half the real wait (with a floor of 4 months), and at least 12 outcomes are known. If our model fails but the simple rule “the cutoff keeps the pace of the last 12 months” passes, we show that rule with a band equal to its median error measured over the same period, and say so. If neither passes, the calculator says “no reliable forecast”. More than six years behind is outside what we tested, and no estimate further than ten years ahead is printed: no date is published there either.
| Case | 6 months behind | 12 months behind | 24 months behind | 48 months behind |
|---|---|---|---|---|
| F1 · All other countries | 4 / 5 · 59 % ✓ | 7 / 9 · 54 % ✓ | 13 / 20 · 48 % ✓ | 15 / 38 · 34 % |
| F1 · Mexico | 9 / 9 · 62 % | 12 / 17 · 44 % | 23 / 21 · 18 % | 47 / 33 · 0 % |
| F1 · Philippines | 8 / 18 · 55 % | 11 / 24 · 55 % | 16 / 26 · 38 % | 18 / 43 · 31 % |
| F2A · All other countries | 3 / 4 · 67 % ✓ | 5 / 6 · 76 % ✓ | 8 / 20 · 70 % | too few months |
| F2A · Mexico | 3 / 3 · 84 % ✓ | 5 / 6 · 89 % ✓ | 7 / 18 · 78 % | too few months |
| F2B · All other countries | 3 / 3 · 60 % ✓ | 5 / 7 · 53 % ✓ | 10 / 18 · 31 % | 11 / 18 · 34 % ✓ (12-month pace) |
| F2B · Mexico | 7 / 3 · 51 % ✓ (12-month pace) | 10 / 7 · 44 % ✓ (12-month pace) | 18 / 7 · 21 % ✓ (12-month pace) | 37 / 19 · 0 % ✓ (12-month pace) |
| F2B · Philippines | 4 / 8 · 60 % ✓ | 7 / 12 · 48 % ✓ | 10 / 11 · 47 % ✓ | 13 / 15 · 58 % ✓ |
| F3 · All other countries | 4 / 5 · 63 % ✓ | 5 / 6 · 51 % ✓ | 11 / 13 · 37 % ✓ (12-month pace) | 10 / 21 · 46 % ✓ |
| F3 · Mexico | 9 / 8 · 67 % | 13 / 12 · 50 % | 30 / 18 · 9 % ✓ (12-month pace) | 511 / 28 · 0 % ✓ (12-month pace) |
| F3 · Philippines | 7 / 8 · 53 % | 12 / 12 · 53 % | 13 / 31 · 59 % ✓ | 16 / 66 · 62 % ✓ |
| F4 · All other countries | 5 / 6 · 55 % ✓ | 10 / 14 · 52 % ✓ | 13 / 31 · 43 % ✓ | 23 / 51 · 36 % |
| F4 · India | 5 / 6 · 46 % ✓ | 10 / 18 · 40 % ✓ | 26 / 22 · 24 % ✓ (12-month pace) | 59 / 34 · 38 % |
| F4 · Mexico | 8 / 14 · 49 % ✓ | 21 / 16 · 32 % ✓ (12-month pace) | 44 / 28 · 10 % ✓ (12-month pace) | 25 / 53 · 47 % ✓ |
| F4 · Philippines | 2 / 2 · 58 % ✓ | 4 / 3 · 52 % ✓ | 8 / 5 · 53 % ✓ | 11 / 21 · 60 % ✓ |
| EB-1 · India | 3 / 570 · 44 % ✓ | 5 / 578 · 43 % | 10 / 577 · 19 % | 4 / 594 · 29 % |
| EB-1 · China (mainland born) | 7 / 16 · 30 % | 14 / 20 · 18 % | 9 / 74 · 14 % | too few months |
| EB-2 · All other countries | 6 / 9 · 29 % | 12 / 18 · 38 % | too few months | too few months |
| EB-2 · India | 6 / 17 · 67 % | 12 / 34 · 77 % | 21 / 54 · 75 % | 41 / 82 · 80 % |
| EB-2 · China (mainland born) | 3 / 4 · 72 % ✓ | 6 / 12 · 65 % ✓ | 12 / 19 · 44 % | 7 / 38 · 88 % ✓ |
| EB-3 · All other countries | 3 / 8 · 62 % ✓ | 5 / 8 · 76 % | 6 / 16 · 78 % | too few months |
| EB-3 · India | 6 / 10 · 54 % | 9 / 15 · 47 % | 24 / 25 · 41 % | 38 / 39 · 32 % |
| EB-3 · China (mainland born) | 2 / 7 · 65 % ✓ | 5 / 15 · 65 % | 9 / 20 · 68 % ✓ | 11 / 49 · 76 % |
| EB-3 · Philippines | 3 / 6 · 61 % ✓ | 5 / 8 · 61 % | 9 / 32 · 57 % | 20 / 70 · 70 % |
| EB-3 Other Workers · All other countries | 4 / 5 · 64 % ✓ | 7 / 9 · 53 % | 16 / 16 · 35 % | too few months |
| EB-3 Other Workers · India | 7 / 10 · 57 % | 9 / 15 · 40 % | 31 / 25 · 27 % | 39 / 39 · 10 % |
| EB-3 Other Workers · China (mainland born) | 5 / 7 · 70 % | 7 / 10 · 65 % | 14 / 19 · 77 % | 20 / 32 · 82 % |
| EB-4 · All other countries | 10 / 576 · 11 % | 10 / 590 · 4 % | 20 / 575 · 4 % | too few months |
| EB-5 · India | 3 / 595 · 27 % | 4 / 593 · 26 % | 18 / 588 · 21 % | too few months |
| EB-5 · China (mainland born) | 12 / 21 · 57 % | 24 / 39 · 63 % | 40 / 79 · 75 % | 30 / 183 · 100 % |
USCIS processing times
The official USCIS processing-time tool is closed to automated readers too. We use the open data of ImmigrationTimes.org (CC BY 4.0 license), which keeps the figures USCIS published by office and form since March 2022; latest USCIS publication: September 17, 2026. We select them and change their format. Next to them we show the national average from USCIS’s monthly report to Congress (August 2026), which measures something else: the average time of all cases completed in the month.
Fees
All of them come from USCIS Form G-1055, edition 10/07/26, read in full for the forms on this site, and from the State Department’s visa fee page. They live in a parameter file checked by automated tests; when the edition changes, the site does not rebuild until a person has read them again.
What we do not do
We give no legal advice, review no cases, ask for no email or phone number, and what you type in the calculators never leaves your browser. We are not affiliated with USCIS, the Department of State or any government agency.