Percom workflow

PerCom Workflow

Use this as the project-management skill for a PerCom submission. Replace every date with the current official timetable and work backwards from the HotCRP registration cutoff — which precedes the paper deadline by about a week. PerCom runs a single annual deadline feeding one review round with a bounded rebuttal and an early-rejection gate; there is no journal-style revision, so plan for the paper to be complete at submission.

PerCom is an IEEE Computer Society conference: it has no standing editor-in-chief and no APC in the open-access sense. Rotating leadership is the per-edition General and TPC (Program) Co-Chairs (PerCom 2027 full roster 待核实 as of 2026-07-09), and the cost model is IEEE registration with at least one author presenting in person.

Milestones

• Venue fit: confirm PerCom over ACM UbiComp/IMWUT, MobiCom, SenSys, or IPSN, and confirm the paper is ready this cycle (percom-topic-selection).
• Data collection lock: IRB/consent secured, subjects recruited, sensing protocol frozen — the long pole for a human-subjects paper.
• Evidence lock: freeze claims, cross-subject splits, baselines, metrics, limitations, and the dataset contents. Nothing can be added in the rebuttal.
• Registration: submit real title, abstract, authors, topics by the earlier registration deadline (PerCom 2027: 4 Sep 2026).
• Submission: upload the anonymized PDF and the open-data plan (PerCom 2027: 11 Sep 2026 AoE).
• Early-rejection gate: notification (~20 Nov 2026) either early-rejects or invites a rebuttal.
• Rebuttal (if invited): answer explicit reviewer questions from existing evidence, anonymously (~26 Nov 2026).
• Decision: accept/reject (~18 Dec 2026); archive reviews, rebuttal, and every submitted version.
• Acceptance: prepare the IEEE Xplore camera-ready (PDF eXpress, eCopyright, ORCID), deposit the de-identified dataset, register, and present in person.

Backward plan from the submission deadline

Weeks out (heuristic)	Human-subjects sensing milestone
16+	IRB/consent approved; recruitment and data collection begun (the long pole)
12	Data collection complete; labeling protocol run with annotator agreement
10	Claims fixed; cross-subject (LOSO) splits and baselines chosen
8	Experiments complete; F1/event-level results with per-subject spread computed
6	Analysis done; limitations drafted alongside results; dataset de-identified
4	Full draft in the IEEEtran 9+1 template; artifact assembled and anonymized
3	Internal mock review by a ubicomp reader ("is this within-subject only?")
2	Limitations hardened, related-work delta sharpened, 9-page budget met
1	Double-blind sweep on PDF and dataset link; open-data plan final
0	Registration by its earlier cutoff, then paper + dataset link on HotCRP

These offsets are planning heuristics only — anchor every one to the current Important Dates page, never to a previous cycle's calendar. Note the 16-week IRB/collection pole: human-subjects data cannot be rushed in the final month.

Cycle reality: one shot per year, gated early

PerCom runs one annual research deadline. As of 2026-07-09, PerCom 2026 (Pisa) has concluded, so the live next target is PerCom 2027 (Goa; deadline 11 Sep 2026). Two planning consequences:

• There is no revision cycle. A paper that is not complete — especially one lacking cross-subject evidence — cannot be rescued in the rebuttal; it will likely be early-rejected. Finish the evaluation before submitting.
• Missing the annual deadline costs roughly a year at PerCom; ACM IMWUT has rolling quarterly deadlines and a revise cycle, so a paper that needs another round of experiments may be better routed there than held for a year — factor that into routing rather than idling.

Failure modes by stage

• IRB or recruitment slipping pushes data collection into the final month and forces a within-subject-only evaluation — the classic PerCom early-reject in the making.
• Discovering at week 3 that the split was within-subject is fatal with no time to recollect; design the LOSO split at week 10.
• Leaving the anonymized dataset to the final week is how a testbed name, a re-identifying photo, or an owner string leaks identity under double-blind review.
• Skipping the mock review forfeits the one chance to hear "your evaluation is within-subject" from a friend instead of a reviewer — and there is no revision to fix it.
• Treating the rebuttal as a revision and planning to "add the experiment then" — the process neither expects nor permits it.

Coordination notes

• Assign one owner for IRB/consent and the dataset, and another for the double-blind anonymity sweep; shared ownership is how both slip.
• Archive the exact submitted PDF, dataset, and (later) rebuttal — the discussion must quote them precisely and stay anonymous.

Output format

[Current stage] idea / IRB+collection / evidence / writing / registration / submitted / rebuttal / accepted
[Next official deadline] <date and source, or unknown>
[Critical path] <three tasks that determine readiness -- IRB/collection is usually one>
[Risk register] <IRB / cross-subject evaluation / 9-page budget / anonymity / dataset / presentation>
[Owner map] <task -> person or role>