Original analysis
193, 285, 660: the three answers to how fast America could draft
Published 24 July 2026. Every figure below is from a government document, and every document is cited. Journalists and researchers are welcome to quote this page; the sources are listed in full at the bottom and the underlying data is browsable.
Ask how quickly the United States could restart the draft and you get one number: 193 days. It appears in Government Accountability Office reports, in congressional testimony, on Selective Service's own website. It is repeated as though it were an engineering estimate.
It is not. It is 13 days plus a guess. And the agency that would have to execute it has twice put a different number on the record: 285 days in 2012, and approximately 660 days in a wargame it ran on itself in January 2026. None of the three has been retracted. They measure different things, and almost nobody reporting on this distinguishes between them.
1. Where 193 comes from
Until 1994 the Pentagon's requirement was blunt: first inductees available 13 days after mobilisation notification, 100,000 available at 30 days. In November 1994 an Assistant Secretary of Defense memorandum changed the requirement — not by studying the pipeline, but by adding six months of assumed advance warning in front of the existing numbers.[1]
The arithmetic is exact, and it is provable in two places:
| Milestone | Pre-1994 | Assumed warning | Current |
|---|---|---|---|
| First inductees report | 13 d | + 180 d | 193 d |
| First 100,000 report | 30 d | + 180 d | 210 d |
193 − 13 = 210 − 30 = 180. GAO documented the change in 1997 and again in 2012, and the 2020 national commission's own diagram labels the front block "six-month assumed lead time."[1][2][7] The rationale in 1994 was that the active and reserve forces would suffice for two major regional conflicts, so a draft would only ever follow a long, legible crisis. GAO found in 2012 that the requirement had "not been updated since," that the Department "did not conduct any additional analysis," and that it asserted only that the figure was "still considered realistic."[2][3]
The consequence nobody states. Because 180 of the 193 days are assumed warning, the 193-day requirement is not a claim that Selective Service can deliver inductees 193 days after being told to. It is a claim that it can deliver them 13 days after being told to, provided it was quietly told six months earlier. Those are radically different promises, and only one of them has ever been tested — neither, in fact: the commission found that "no significant tests of the system have been undertaken to assess whether it can fulfill its mission during an emergency."[7]
2. The 13 days have about three days of slack
Here is where the decomposition stops being trivia. The 13-day leg was never relaxed — it survived the 1994 change intact. And a separate body of law constrains it. 32 C.F.R. 1624.4(a): "The date specified to report for induction shall be at least 10 days after the date on which the Order to Report for Induction is issued."[6]
Ten of the thirteen days are therefore consumed by the notice period, before anyone has run a lottery, generated an order, or put anything in an envelope. What remains for the lottery, the selection run, order generation and the mail is on the order of three days.
Selective Service's own plan of record confirms the fit is exactly that tight rather than coincidentally so. The plan schedules first induction notices mailed "not later than day 183" and first inductees reporting "no later than day 193"[8] — a ten-day gap, precisely the statutory floor, with nothing allowed for mail transit. And mail transit time is not covered by the ten days and is not specified anywhere in the released record.
Two further legal facts make the three days of slack thinner still, and both are in the same regulation. Filing a claim for reclassification delays the reporting date until "not earlier than the tenth day after" the claim is finally determined — that is, ten days after the entire appeal chain is exhausted.[6] And the notice goes to "the current address most recently provided by the registrant," which is a registry whose quality is separately measurable and separately falling.
3. 285: the agency's own 2012 correction
In June 2012, GAO recorded Selective Service conceding it could not meet the requirement:
"The Selective Service System officials said that due to reductions in the number of personnel available to set up area offices across the country, it now estimates it could not deliver the first inductees until 285 days after mobilization."[2]
A 92-day shortfall, attributed to a specific cause: standing up the field infrastructure. The same report records the agency saying it "is not currently resourced to meet DOD's requirement… without jeopardizing the fairness and equity of the draft."[2]
The scale explains the shortfall. A mobilised Selective Service is 2,069 local boards, 436 area offices, 96 district appeal boards, 48 alternative service offices and roughly 9,200 paid staff, of whom about 6,500 — 71% — are newly hired federal employees who do not exist in any form during peacetime.[4] The published plan compresses that hiring into days 6–45. Meanwhile the peacetime agency has gone the other way: an authorisation of 267 full-time equivalents in FY1993, down to a planned 101 in FY2026 — a 62% reduction against a requirement that never moved.[5]
Honest limits on the 285: it is a self-assessment with no published methodology, it is fourteen years old, and no later quantified capability estimate exists in the public record — GAO's 2018 follow-on obtained none. Budget and staffing were broadly flat from 2012 to 2019, so the shortfall is presumed to persist, but that is an inference and should be labelled as one.
4. 660: what the agency found when it wargamed itself
On 15 and 27–29 January 2026, Selective Service ran an internal Associate-Director-level tabletop exercise on its own mobilisation. The scenario was a protracted attritional war with a peer adversary in which reserve forces are exhausted for critical specialised skills, Congress amends the Military Selective Service Act to authorise conscription, and the President directs a specialised lottery at M+16.[9] The after-action report's bottom line, verbatim:
"The Response Planning exercise resulted in an estimate of approximately 660 days for the Selective Service System to return to full operations from our existing state of readiness. Forty-four individual actions were identified that the System can complete while in Active Standby status to reduce the projected response time by up to one year."[10]
Take the report at its own word and the arithmetic gives a floor: 660 − 365 = approximately 295 days if all 44 readiness actions were completed in advance. Against the scenario's own requirement of first delivery at M+30, the exercise self-diagnosed a gap of roughly 265 to 630 days.
Three sentences from the same report matter as much as the number:
- "The System currently does not have any planning COAs to guide a return to full operations."[10]
- The exercise identified three courses of action — mass mobilisation, phased expansion, tiered capability — and "no conclusions were reached."[10]
- Listed under required decisions: "What is the true pacing function for mobilization?"[10]
That last line is the reason this project exists. The agency responsible for the draft has stated, in its own after-action report, that it does not know what its own bottleneck is.
Honest limits on the 660: the after-action report is a draft, and the enclosures that build the figure up — "Sequence of Actions and Timing", per directorate and Agency-wide — were not released, so the number cannot be audited. The 44 actions are never enumerated anywhere in the released set. The 295-day floor is our arithmetic on the report's own two figures, not a stated conclusion. Nine rows of specific observations and recommendations are redacted in full under exemption (b)(5).
5. Why the three numbers do not contradict each other
They measure two different clocks, and conflating them is the central analytical error in public discussion of this subject.
Clock A — organisation standup
How long until the agency is capable. Hiring 6,500 people, appointing 11,000 board members, opening 436 area offices, activating 96 appeal boards. This is what 660 measures. It is also what the 92-day slip in the 285 figure was attributed to.
Dominated by recruiting, clearing, training and real estate. Not sensitive to how many inductees you eventually want.
Clock B — pipeline latency
How long an individual takes to traverse the machine once it is running: lottery, order, statutory notice, examination, claim window, board, induction. This is the 13-day leg, and it is what the simulator measures.
Dominated by statutory minimum notice periods and by whichever resource saturates first. Highly sensitive to volume.
The 193 and 285 figures are hybrids: each contains a 180-day assumed-warning prefix (Clock A, assumed away rather than measured) followed by a pipeline leg (Clock B). The 660 figure is pure Clock A, measured from today. Subtracting one from another, as commentary routinely does, produces nonsense.
Stated cleanly, the government's own documents say: if we get six months of unambiguous warning and use it, the pipeline can deliver in about a fortnight; if we do not, becoming capable at all takes about two years, or about ten months with a programme of advance work that has not been funded or, as far as the released record shows, begun.
6. Quantified: their timetable, two sets of parameters
Everything above is document analysis. This section is the part that needed a model, and it is the reason this project exists.
We ran Selective Service's own plan of record through the simulation twice.[15] Same timetable, same milestones, same nine stages. The only difference: the first run uses the government's own planning assumptions, and the second uses corpus-sourced parameters for the same quantities.
| Milestone | Plan day | Their assumptions | Sourced parameters |
|---|---|---|---|
| First inductee reports | 193 | 199 (+6) | 199 (+6) |
| 100,000 inductees reported | 210 | 216 (+6) | 291 (+81) |
| 36,000 medical personnel (HCPDS) | 223 | 229 (+6) | — |
Two things fall out of that table, and they are the whole argument.
- The engine can express their plan. Under the government's own assumptions it hits all three published milestones within six days (+3.1%, +2.9%, +2.7%). The six days are not slop: they are exactly one day per stage traversed, an artefact of the model's rule that no cohort crosses two stages in a single day. Known sign, known size, published rather than smoothed away. So when the model disagrees with the plan elsewhere, it is not because it cannot represent the plan.
- The divergence is throughput, not schedule. Swap in sourced parameters and the first inductee does not move at all — day 199 either way. The 100,000 milestone moves 81 days, from day 216 to day 291. At the plan's own day 210, the sourced run has delivered 12.6% of the 100,000 rather than 67.6%.
And the 81 days localise to a single parameter. Under the Department of Defense's assumption of 18,000 examinations a day, the examination stage never binds at all — 0% of stage-days. Under the measured envelope (65 stations at about 2,249 a day), it binds 78.8% of stage-days, with a peak examination queue of 1.67 million men.
Stated plainly: the difference between the government meeting its own draft milestone and missing it by nearly three months is one number that the Department of Defense itself labels an assumption, has never measured or tested, and which contradicts that same department's stated physical capacity by roughly seven times. That is not a modelling result. It is a documentary one, expressed as a date.
Why this is not circular. Nothing was tuned to make this come out. Every era and plan parameter is derived from a document independent of the series it is scored against, and each override carries its citation and the value it replaced. The same uncalibrated model, run against the actual Vietnam draft, reproduces six years of real induction volume to within +4.7% (1,525,028 simulated against 1,456,389 actually inducted) and Selective Service's own delivered-to-inducted yield to within −3.4% (0.757 against 0.784). It also fails, in ways we publish: monthly timing error is large (weighted absolute error 36.2%, running about 1.9 months late), and a 1980 cold-start registration curve exposed a structural gap in how the model holds a registrant population at all. The failures, in detail →
7. The 180-day assumption has no mechanism behind it
Everything above rests on the warning prefix, so it is worth asking what triggers it. The January 2026 exercise's own scenario spreadsheets carry a column headed "Credible Indicator / Readiness Condition," and the first moderator questions put to players were "What is our definition of a credible indicator? What is the purpose of a credible indicator?"[11]
In other words: as of January 2026, the agency had not defined the indicators that would start its 180-day clock. The kick-off brief lists five categories — congressional and executive oversight, resource and scalability signals, legislative precursor activity, sustained National Security Council engagement, escalating credible media reporting — with part of the second one redacted.[11] Those are indicators of political attention, not of adversary intent. All four of the unredacted categories are things that happen after Washington has already decided something is wrong.
That has a modelling consequence, which the honest thing to do is state rather than paper over: warning time is not derivable from any Selective Service source. In our parameter ledger it is an ASSUMPTION-grade quantity, and on the simulator it is a control you set yourself rather than a number we supply.
8. The special-skills contradiction
One more finding from the 2026 exercise, because it is a clean example of a plan that cannot execute. The scenario has the President directing a specialised lottery at M+16 and Congress requiring first delivery at M+30, against an eventual demand of one million inductees in fifteen months.[9]
Both available branches are dead well before M+30:
- Draft from the existing registry by occupation. Nothing in the released record supports it: the regulation defining registration enumerates nine data elements a registrant records, and occupation is not among them; no schema for the current registrant systems has ever been published. Nineteen target occupations survive in the released documents only inside an information-technology scenario guide.[15]
- Use the standing mechanism. The only one that exists is the Health Care Personnel Delivery System, aimed at health-care occupations, and it registers nobody today. Its own timetable spends days 5–45 on mass registration of the target population before it can hold a lottery, with first health-care inductees at day 193 and 36,000 inductions by about day 223.[12]
A day-16 lottery over occupational categories is not reconcilable with a mass-registration-first design. This is not a criticism of the exercise — the exercise existed to surface exactly this — but it is a hard constraint on any reporting that treats a skills draft as a near-term option. The document-by-document version of this section, including what would change the analysis, is on the special-skills page.
A related correction, since it circulates in the other direction: the 2020 commission did not "vote down" a special-skills draft. Its Final Report records a stated conclusion — "Ultimately, the Commission did not support drafting individuals with critical skills" — with no roll-call, tally, dissent, minority statement or separate view anywhere in the report, the executive summary, the eight staff memoranda or the four hearing transcripts.
9. Meanwhile, 193 has quietly left the budget record
The 193-day requirement appears in the Selective Service congressional budget justifications in an unbroken run from FY2012 through FY2020, and in none since — a text search across every archived justification finds zero hits from FY2021 through FY2027. The current document instead commits the agency to "Maintain the Integrated Master Mobilization Schedule (IMMS)," a document that is not published.[13]
The figure has not been repudiated — it still appears on the agency's public website. But the most-cited number about how fast America could draft has been replaced, in the budget record, by a schedule the public cannot read. That schedule is the single highest-value unreleased document in this subject, alongside Enclosure 4 of the January 2026 after-action report. The year-by-year census, and prepared request letters for both documents, are on the missing-documents page.
What would change these numbers
- Release Enclosure 4. The Agency-wide "Sequence of Actions and Timing" is what the 660-day figure is built from. Without it the number cannot be audited by anyone outside the agency.
- Release the Integrated Master Mobilization Schedule. It is now the operative document and it has replaced a public number with a private one.
- Enumerate the 44 readiness actions. The claim that a year can be cut off is the most consequential quantitative statement in the released set, and it is unaccompanied by a single specific.
- Publish a per-station examination capacity. USMEPCOM's "Maximum Daily Capacity Allocation" appears in no public document, which means the binding constraint in any capacity-limited draft is unmeasurable from outside.
- Define the credible indicators. Until they exist, the 180-day warning prefix that carries 93% of the 193-day requirement is an assumption with no trigger.
To be unambiguous about what this article is not saying. None of the above suggests a draft is planned, imminent, or likely. The Department of Defense stated six separate times in its statutory report to Congress that it "currently has no operational plans that envision mobilization at a level that would require conscription," and there is no published total-volume requirement for an American draft at all.[14] The point of this analysis is narrower and more durable: a public commitment that is repeated for thirty years without being tested, and that the responsible agency has twice contradicted on the record, deserves to be read carefully.
Sources
- GAO/NSIAD-97-225, Selective Service: Cost and Implications of Two Alternatives to the Present System (1997), Background section, paragraph beginning "Until 1994, DOD required the first inductees to be available 13 days after mobilization notification." govinfo
- GAO-12-623, National Security: DOD Should Reevaluate Requirements for the Selective Service System (June 2012): the 1994 memorandum decomposition at PDF p. 8; the 285-day estimate and "not currently resourced" at PDF p. 10 (printed p. 6); "it is unclear whether DOD would need 100,000 inductees in 210 [days]" at PDF p. 17. gao.gov
- GAO-18-226, National Security: Ongoing Review of the Military Selective Service Process Could Benefit from Additional Information (January 2018), Table 1 element 5 (day 193 / day 210 / day 222) and p. 4; board and office counts at PDF p. 13. gao.gov
- GAO-18-226, Table 1 element 3, manpower paragraph: "more than 500 Reserve Force Officers, 1,500 military retirees recalled to duty, 700 state resource volunteers, and 6,500 newly hired federal employees." The ~9,200 total is arithmetic on those four stated categories and is a lower bound.
- Selective Service System, Annual Report to Congress, FY2000, staffing narrative (267 FTE authorised in FY1993; 165 in FY2000) archived; and FY2026 Congressional Budget Justification, personnel table (110 → 101, −8%) sss.gov.
- 32 C.F.R. 1624.4(a) (minimum 10-day induction notice; claim-filing delays the reporting date until not earlier than the tenth day after final determination); 32 C.F.R. 1624.5(a) (notice sent to the most recently provided address); 32 C.F.R. 1624.10(a) (7-day minimum for examination notices). govinfo, 32 CFR part 1624
- NCMNPS, Inspired to Serve: Final Report (March 2020), Figure 11 "The Current 193-Day Total Mobilization Plan," PDF p. 115, labelling the front block "Six-month assumed lead time"; "no significant tests of the system have been undertaken" at PDF p. 114. archived
- Selective Service System, Readiness Plan / Agency Response Plan, Annex C Mobilization Timeline, April 2009: "First induction notices mailed not later than day 183"; first inductees report "no later than day 193"; lottery "on or before Day 45." Released in a FOIA compilation. governmentattic.org
- Selective Service System, 2026 Response Planning ExPlan, 1 October 2025, PDF p. 9 (scenario; M-day; "POTUS directed SSS to hold a specialized lottery on M+16"); and 2026 Response Planning Timeline, Moderator, and Notetaking Guide — IT, PDF p. 2, disclosure blocks 11 ("First delivery required on M+30") and 13 ("DoW requests 1 million inductees in 15 mos."). FOIA release archive
- Selective Service System, Response Planning Exercise After Action Report (draft, redacted), 2026: BLUF at PDF p. 1; "does not have any planning COAs" and the three courses of action with "no conclusions were reached" at PDF p. 2; "What is the true pacing function for mobilization?" at PDF p. 3, item E; unreleased enclosures listed at PDF p. 1. PDF
- Selective Service System, 2026 Response Planning Exercise Phase 1 Kick-off, PDF p. 7 (credible-indicator categories, one bullet redacted under (b)(5)); and the three directorate Timeline, Moderator, and Notetaking Guides, p. 1 (the "Credible Indicator / Readiness Condition" column, and the opening moderator questions). FOIA release archive
- Selective Service System, ARP Annex K Appendix 2, HCPDS Mobilization Timetable, April 2009, PDF p. 8 (Day 5–45 mass registration then lottery) and PDF p. 9 (day 193 and day 223 milestones). PDF
- Selective Service System, Congressional Budget Justifications FY2010–FY2027: "193 days" / "M+193" / "day 193" present every year FY2012 through FY2020; zero hits FY2021–FY2027. (The FY2010 justification is a scanned image and could not be text-searched; FY2011 has no occurrence.) FY2027 justification, Strategic Objective 1.1: "Maintain the Integrated Master Mobilization Schedule (IMMS)." sss.gov/cj
- Department of Defense, Report to Congress on the Registration System (FY2017 NDAA §551), "Assumptions Used by the Department," PDF pp. 33, 36, 37–38 (the no-operational-plans statement, six occurrences); notional mobilisation timeline at PDF pp. 34–36. archived
- 32 C.F.R. 1615.4, "Duty of persons required to register," subsection (a), enumerating the elements a registrant records; Selective Service System, Registrant Integrated Processing System (RIPS) Manual, Version 3, April 2009, registration data elements at PDF p. 58 and the ten-day duty to report a change "in his name, address, or telephone" at PDF p. 60. govinfo, 32 CFR part 1615
- MOBSIM historical backtest suite, run 24 July 2026 against engine v0.1.0
with corpus-sourced era and plan parameter ledgers: three episodes (Vietnam steady
state CY1966–71, the 1980 registration reinstatement, and the plan-of-record
milestone crosscheck), nine runs, 864 replications, conservation error at or below
1e-6 in every run. Every override carries its own citation and the value it replaced;
scores, diagnostics and failures are published together. Full report and the exact
commands to reproduce it are in the project's
validate/directory.
Corrections are welcome and will be published on this page with the date they were made. Contact: contact@draftclock.org. Every number above is also a row in the dataset, with the archived file it was read from.