I-821D median fell 16 days this week, but the broader trend still points upward

This week's median for the I-821D (Consideration of Deferred Action for Childhood Arrivals) final-stage wait landed at 145 days, down 16 days from 161 days last week. The anomaly detector flagged this as an outlier_week_low at z = -3.51, computed against the prior 4-week trailing mean, meaning the drop is more than three standard deviations below that recent trend line. Against the longer 12-week view, 145 days sits 2 days below the 147-day trailing average, so the week is not a dramatic departure from the broader range.

The single-week move should be read with care. The 12-week pattern is classified as noisy_stable: the 8-week slope is +4.3 days per week, but the R² of 0.13 means the regression explains very little of the week-to-week variance. One week's reading is a poor predictor of the next. The current processing environment also remains far removed from the same week one year ago, when the same-week median was 9 days across 1,935 approved cases. That comparison reflects a fundamentally different volume and routing context rather than a comparable baseline.

Over the past 12 weeks, the I-821D median has moved from 86 days (week of June 29) to 181 days (week of July 27) and back down toward the current 145-day reading. That is a 95-day swing within a single quarter. The 25th percentile (p25) and 75th percentile (p75) have also moved substantially: the p25 ranged from 4 days in one week to 138 days in another, and the p75 ran from 106 days up to 206 days. The width of this band is consistent with the noisy_stable pattern classification the data assigns to this series.

What noisy_stable means practically is that the underlying trend is not cleanly accelerating or decelerating. Weekly readings bounce around a rough center of gravity near 147 days, the 12-week trailing average, but individual weeks depart from that center by tens of days in either direction. The data does not point to a specific operational change that explains the swings. For readers tracking their own cases, this pattern means a single good week or a single slow week carries limited information about what follows.

A 95-day swing in one quarter means any single week's reading tells you where this week landed, not where next week is going.

Cases that went through the RFE (Request for Evidence) process and submitted a response posted a median final-stage wait of just 9 days this week, versus 135 days for cases that proceeded without an RFE. The anomaly detector flagged this as rfe_fast, with a speed ratio of roughly 15 to 1 in favor of the RFE-response group. The inversion is counterintuitive: the RFE process typically adds time, not removes it.

The likely explanation is a cohort (a group of cases) artifact rather than a structural policy shift. The weighted approval breakdown shows only 342 cases at the rfe_response stage across the last 12 weeks. A small group of RFE-response cases whose responses arrived at a moment when adjudicator review moved quickly may have cleared together, pulling the median down sharply for this one week. An RFE is not a denial indicator; most RFE'd I-821D cases are ultimately approved. The data brief does not support predicting that RFE-response cases will continue clearing at this pace.