sigmod-review-process

SIGMOD Review Process

A SIGMOD research submission enters a journal pipeline with a conference cadence: papers submitted to a PACMMOD round are reviewed double-anonymously through Microsoft CMT, authors get a mid-round feedback window, and the first verdict is three-valued — accept, reject, or revise. Understanding which lever moves which outcome is most of review-stage strategy. Mechanics shift by cycle; reconfirm against the live CFP before relying on any date.

Anatomy of a round (SIGMOD 2027 Round 3 as the concrete case)

Jul 10, 2026   Abstract + conflict declarations close
Jul 17, 2026   Paper deadline (11:59 PM AoE)
Sep 10-17      Author feedback phase: reviews visible, short reply window
Oct 19, 2026   Verdict: accept / reject / major or minor revision
Nov 19, 2026   Revised paper + 4-page letter due (revision cases)
Dec 12, 2026   Final accept/reject on revisions

Roughly three months from PDF to first verdict, five to final — faster than most journals, slower than a rebuttal-only conference, and repeatable four times a year.

The three verdicts, decoded

Two consequences deserve emphasis. First, a revision verdict is genuinely winnable — the requirements are written down, and reviewers judge compliance rather than re-litigating taste. Second, the embargo makes "just resubmit next round" impossible, so a rejected SIGMOD paper's next stop is usually a different venue (see sigmod-topic-selection).

Who is reading

The PC is dominated by database systems researchers and industrial engine builders. Predictable evaluation reflexes:

• They benchmark mentally: the first skim looks for the workload table and the hardware paragraph before the design section gets full attention.
• They know the baselines personally — sometimes literally, as authors or maintainers — so an untuned competitor is recognized on sight.
• Architecture claims are audited against systems they have built: "this cannot be lock-free as described" is a common decisive comment.
• Novelty is measured against both recent PACMMOD/PVLDB output and decades of systems history; rediscovering a 1990s technique under a new name is a classic kill.

Feedback phase leverage

The mid-round window is short and asymmetric: reviewers have already scored, and you are trying to change specific minds before the PC discussion. Highest expected value, in order: correcting an outright factual misreading; answering a direct question with a pointer to a specific table; resolving an apparent contradiction between two of your own sections. Near-zero value: arguing significance, promising future work, or thanking reviewers at length inside a tight character budget.

Reading a split panel

• One detailed negative review plus shallow positives → the verdict follows the detailed one; answer it line by line in the feedback phase.
• Universal "interesting but evaluation thin" → expect major revision; start the missing experiments before the verdict arrives.
• A correctness objection from a builder of a rival system → treat as the round's decisive issue even if other scores are high.
• Complaints about clarity of the problem statement → often masks a fit problem; consider whether the framing buried the data-management core.

Round vocabulary, decoded

Terms that trip up newcomers to the PACMMOD model:

• Round — one of the year's submission batches; your paper lives its whole life inside a single round unless rejected.
• Issue — the PACMMOD journal number your round's acceptances publish in; citable independently of the conference.
• Feedback phase — the mid-round reply window; not a rebuttal in the score-flipping sense, and not the revision.
• Revision — a formal verdict with enumerated requirements, not an informal second chance; compliance is what gets re-reviewed.
• Embargo — the 12-month bar on returning to the research track after rejection or late withdrawal.
• Shepherd — when assigned, the PC member who verifies promised camera-ready changes; treat their notes as binding.

Etiquette with teeth

Confidentiality and anonymity bind for the whole round: no tweeting the submission, no updating a named preprint mid-round in a way that links to the anonymous version, no contacting suspected reviewers. Violations here are policy matters for chairs, not style points.

After the final decision

Close the loop regardless of outcome. On acceptance, archive the full review thread beside the paper — the shepherding and ARI stages quote it. On rejection, hold a one-hour post-mortem within a week while memory is fresh: classify each objection as fixable-evidence, fixable-writing, or fit, and let that classification pick the next venue rather than pride or sunk cost. Reviews from a SIGMOD round are unusually actionable input for a PVLDB or ICDE resubmission precisely because the reviewer pools overlap.

Output format

[Round position] submitted / feedback window / awaiting verdict / revising
[Panel read] per-reviewer stance and the decisive reviewer
[Verdict forecast] accept / revision / reject, with the driving issue
[Feedback plan] top corrections worth the character budget
[Revision feasibility] 30-day workload if a revision lands
[Embargo contingency] next venue if rejected