Icde workflow

ICDE Workflow

Use this as the project-management skill for an ICDE research submission. Replace every date with the current edition's official timetable and plan backward from the round you target.

ICDE is a conference, not a journal-of-the-conference: it has no standing editor-in-chief and no article-processing charge. Leadership rotates per edition (General and Program Chairs named on each icdeYYYY.github.io site), the cost model is registration, and accepted papers publish in IEEE Xplore — not in a PACMMOD- or PVLDB-style venue journal. Re-check the current organization page rather than carrying a name forward.

The two-round unit

Unlike SIGMOD's four quarterly PACMMOD rounds or PVLDB's twelve monthly deadlines, ICDE gives you two shots per edition and no more. ICDE 2027's Round 1 paper deadline was June 11, 2026; Round 2 is November 11, 2026 with notification February 10, 2027. Because a Round 1 reject cannot re-enter Round 2, the round choice is a real commitment, not a convenience.

Milestones

• Venue and round fit: confirm the contribution is a data-management primitive and that the chosen round is reachable without rushing the evaluation.
• Evidence lock: freeze the mechanism, workloads, baselines, scale grid, and the claims map.
• Paper deadline (CMT): upload the IEEE-format PDF plus supplemental material; file all co-author conflicts.
• Review release: triage correctness, novelty, evaluation soundness, and reproducibility.
• Revision window (if invited): ICDE 2026 granted invited revisions a 4-week window with a re-review before the final decision; budget for it or forfeit it (revision status for the current edition is 待核实 — confirm).
• Decision: archive reviews, the revision, and the submitted versions.
• Camera-ready + Xplore: IEEE eCF/PDF-eXpress checks, copyright form, final artifact release, registration, and in-person presentation.

Backward plan from the round paper deadline

Weeks out (heuristic)	Systems-paper milestone
10+	System implemented; workloads and baselines chosen
8	Scale grid running; measurement harness logging seeds and variance
6	All main-result runs complete; crossover/cost analysis drafted
4	Full draft sitting in IEEE two-column format
3	Internal mock review by a builder who did not write the system
2	Supplement assembled: run_all.sh, run_small.sh, README, generators
1	CMT profiles + conflicts filed; page count and format sweep
0	PDF and supplement in CMT before the round cutoff

These offsets are planning heuristics — anchor each to the current important-dates page, never to a past edition's calendar.

Revision-bandwidth reasoning

• If your target round can return a Revise & Resubmit, the honest question at submission time is whether the team has four weeks of free bandwidth in the revision window; a revision you cannot execute is a reject in slow motion.
• The two-round structure means a missed Round 1 pushes you to Round 2 of the same edition (five months later) but a Round 1 reject pushes you a whole edition back — price both outcomes before choosing.

Failure modes by stage

• Treating the two rounds as interchangeable and burning Round 1 on an unfinished evaluation, losing the whole edition.
• Leaving IEEE format conversion to the final week, when two-column overflow surfaces late.
• Skipping the builder mock review — the one chance to hear "your baseline is untuned" from a colleague instead of a reviewer.
• Assuming a revision window exists for the current edition without confirming it.

Output format

[Current stage] idea / building / measuring / writing / submission / revision / accepted
[Target round] Round 1 or Round 2 (edition), with next official deadline + source
[Critical path] <three tasks that determine readiness>
[Revision reserve] <weeks of bandwidth available if R&R returns>
[Risk register] <round-choice / format / evaluation / reproducibility / presentation>
[Owner map] <task -> person or role>