Forecasts use a deterministic supply-demand model (the cause) rather than extrapolating historical cutoff dates (the symptom). For each EB category and chargeability, the model constructs cumulative visa number demand by priority date month from USCIS pending I-485 inventory, DOS consular issuances and annual Table V usage, recent I-140 approvals, and approved-petition counts beyond the current DOF. The model then simulates the modeled demand against the knowledge-available annual visa limit, taking into account category allocations, per-country ceilings, spillover rules, and monthly pacing. It then applies deterministic rules to approximate VO decisions to advance, hold, or retrogress dates based on demand.

Identifiable parameters are recalculated from published usage and approval ratios; otherwise, defaults and bounds are calibrated from historical sweep, hold, and retrogression episodes and evaluated by backtesting (holding out all data "published in the future"). The displayed pessimistic–optimistic range is a scenario band, not a statistical confidence interval: it combines higher demand, lower supply, and more conservative cutoff control at one bound, and the reverse assumptions at the other. The current model converts principal petition counts into total demand through a net parameter reflecting derivative count, duplicate filings, attrition, route choice (upgrade/downgrades), and cross-chargeability. It does not explicitly resolve person-level petition duplication, category upgrades or downgrades, and uses one parameter (not a combination of category + country-of-chargeability) to model all these behaviors. EB-4/EB-5 vertical spillover is assumed to be negligible in recent FYs. We also do not model time-varying discretionary behavior by VO, such as unusually aggressive or restrained advances intended to elicit demand; thus policy, processing, or filing-behavior shifts may move actual dates outside the band and are not possible to be modeled here.