I-90 Final-Stage Wait Jumps 25 Days, Triggering a 4.75-Sigma Anomaly Flag

The I-90 median final-stage wait rose from 82 days to 107 days in the week of September 21, a 25-day move that the anomaly detector flagged at 4.75 standard deviations above the 8-week trend line, computed against the prior 8-week trailing regression. That flag is statistically rare. At the same time, the year-ago reading for this same week was 215 days among 8,996 approved cases, so the current level, while elevated relative to recent weeks, remains roughly half what it was a year ago.

Over the 12-week window, medians began at 21 days in early July and climbed, with notable variation along the way. Week 34 (Aug 17) reached 104 days before falling back to a 66-to-82-day range across weeks 35 through 38. This week's reading of 107 days sits above that range and above the 12-week trailing average of 62 days. The data brief does not supply a mechanism for either the earlier week-34 elevation or this week's move; both may reflect shifts in the composition of cases approved in those particular weeks, but a single week is not enough to confirm any one explanation.

The `rfe_fast` anomaly flag fired this week because the group of cases that received and responded to an RFE (Request for Evidence, USCIS's standard follow-up letter asking for missing documents) cleared in a median of 24 days, well below the 56-day median for cases that never received one. Receiving an RFE does not indicate likely denial; most cases that receive one are ultimately approved. The inversion this week may reflect a specific batch of RFE responses that happened to close quickly together, rather than a durable structural shift in how RFE-response cases are handled. One week of data is not enough to distinguish those two possibilities.

This week, cases that required extra documentation cleared more than twice as fast as those that sailed through without it.

If the recent 8-week trend continues, the linear projection puts the median at approximately 74 to 150 days the week of Sep 28 and approximately 90 to 166 days by Oct 19. Those bands are wide by design: the 8-week regression carries an R² of 0.36, too weak to support a single-number forecast, and the residual standard deviation is 19 days, producing a ±38-day band at two standard deviations. One week of anomalous data can tilt a linear regression, so the projection may overstate how persistently this week's reading will carry forward. The slope of +5.4 days per week is real in the data, but the low R² means the trend line is explaining only about a third of week-to-week variation.