I-765 Median Falls 50 Days in a Week, but the Gains Are Concentrated

The I-765 median (the 50th percentile, or p50: half of approved cases finished faster, half slower) came in at 112 days for the week of September 7, down from 162 days the prior week. That 50-day, single-week decline fired the outlier_week_low anomaly flag at z = -2.34, computed against the prior 4-week trailing mean. The 112-day reading is below both the 12-week trailing average of 142 days and the same-week-last-year figure of 133 days. This is a notable single-week reading, not a confirmed acceleration; the 12-week series has ranged from 79 to 198 days, and week-to-week volatility is a persistent feature of this form.

The mechanical driver is IOE's volume share. E-filed cases processed through IOE produced 8,589 of this week's 11,251 approvals, or 76%, at a 96-day median. When one channel accounts for three-quarters of a weekly snapshot, the aggregate median naturally tracks that channel's pace. The contrast is visible at the other centers: EAC (the Vermont Service Center) posted a 286-day median across 303 cases this week, and LIN (the Nebraska Service Center) posted a 396-day median across 219 cases. Those centers affected far fewer approvals, so their higher readings barely moved the overall figure.

The center picture is divided. IOE carries a slope of -11.3 days per week (R² = 0.50, moderate confidence), meaning it has been improving at a meaningful pace over the past 8 weeks. SRC (the Texas Service Center) posts a 101-day median this week but carries a slope of +4.8 days per week with R² = 0.72, a statistically confident upward drift; SRC also shows a 46% transferred-case share, meaning its approval population includes a large fraction of cases that originated elsewhere, which makes a direct comparison with IOE's population unreliable. MSC (the National Benefits Center, USCIS's Lee's Summit, MO facility) shows a 151-day median and the most reliably worsening trend in the group: slope +6.2 days per week at R² = 0.75. EAC's 286-day median comes with a slope of +4.3 but an R² of just 0.08, so the direction is not statistically confident; EAC also carries a 42% transferred share. LIN's 396-day median is the highest this week, though its slope of -3.3 days per week at R² = 0.29 is too weak to call a trend. Because transferred-case shares differ across centers and case-mix differs by category and geography, center medians are not directly comparable as a performance ranking.

Cases that received an RFE and submitted a response cleared in a median of 54 days this week, 86 days faster than cases that moved through plain processing (140 days). The rfe_fast anomaly flag fired because this inversion is uncommon: in most weeks, RFE-response cases take longer than non-RFE cases, since the Request for Evidence process adds a documentation exchange before the adjudication (the USCIS case-decision step) can resume. One plausible reading is that a batch of RFE responses submitted weeks earlier cleared together in a single week, temporarily pulling the RFE-response median down. The data brief does not tell us the mechanism with certainty. Receiving an RFE does not by itself signal denial; most cases that receive an RFE are ultimately approved.

The rfe_fast flag is based on one week's data. If the pattern holds over the next two to three weeks, it will carry more weight as a signal. For now, applicants with pending RFE responses should treat this as a single anomalous reading rather than a new baseline. The 37-day median for cases at the 'Response To USCIS' Request For Evidence Was Received' stage in the Stage durations (30-day raw) data, with a 25th percentile (p25) of 16 days and a 75th percentile (p75) of 86 days across 6,279 cases, gives a broader sense of what RFE-response waits have looked like recently.