| name | vldb-topic-selection |
| description | Use when deciding whether a project belongs at VLDB and in which PVLDB category, applying the data-management-primitive test, choosing among Regular, EA&B, Scalable Data Science, and Vision papers, and routing against SIGMOD, ICDE, CIDR, EDBT, PODS, KDD, systems venues, and The VLDB Journal. |
VLDB Topic Selection
Use this before a line is written. Two decisions hide in "let's send it to
VLDB": whether the work is a data-management contribution at all, and which
PVLDB category gives it the friendliest reviewer expectations.
The primitive test
VLDB rewards work whose core object is a data-management primitive:
storage layout, index, query optimization or execution, transaction and
consistency machinery, data integration and cleaning, streaming state, or the
data infrastructure under ML. Two probes:
- Strip the application narrative. Is what remains a reusable mechanism for
managing data at scale? If what remains is a model architecture or an
application result, the primitive is missing.
- Would the evaluation chapter naturally measure throughput, latency,
scalability, or result quality on data systems? If the natural evaluation
is task accuracy alone, an ML or applied venue fits better.
Category routing inside PVLDB
| Your situation | Category | Watch out |
|---|
| New mechanism + built system + systems evidence | Regular Research (12 pp) | The default; full evaluation burden |
| Rigorous measurement of existing systems, no new system | EA&B (12 pp) | Reproducibility evaluation is mandatory; conclusions must generalize |
| Scale-forward data-science pipeline, practice first | Scalable Data Science (8 pp) | Must still show the data-management lesson, not just an application win |
| Argued agenda without a full system yet | Vision (6 pp) | Small budget; needs a genuinely new direction, not a survey |
Category budgets and continuation for the live volume: verify on the
guidelines page before committing (see the source map's 待核实 ledger).
Neighborhood routing
| Signal in the project | Better home |
|---|
| Quarterly-round rhythm preferred; identical scope | SIGMOD (PACMMOD rounds) — the closest sibling; pick by calendar fit and portfolio, not prestige folklore |
| Formal results: complexity, expressiveness, bounds | PODS or ICDT |
| Provocative architecture argument, prototype-grade evidence | CIDR |
| Solid engineering contribution, broader engineering scope | ICDE or EDBT |
| Mining/learning contribution where data infra is incidental | KDD or an ML venue |
| OS/network mechanism that happens to touch storage | SOSP/OSDI, NSDI, EuroSys |
| Outgrown 12 pages; wants archival depth | The VLDB Journal or TODS |
| Deployed production system, lessons-forward | VLDB industrial track (separate call) |
The practical VLDB-vs-SIGMOD tiebreaker in this collection's experience:
PVLDB's monthly gate and three-month revision suit projects whose evidence
matures unpredictably; SIGMOD's fixed rounds suit groups that plan in
quarters. Scope overlap is nearly total.
Commitment checklist
[ ] Primitive named in one sentence, no application words needed
[ ] Category chosen; its page budget fits the evidence plan
[ ] The one plot that would convince a builder is specified
[ ] Nearest three prior systems identified (see vldb-related-work)
[ ] If EA&B: willing and able to hand everything to the repro committee
[ ] Live volume's topics-of-interest list scanned for explicit fit
Re-route triggers mid-project
- The system never gets built → Vision now, or CIDR.
- The interesting output became the measurement study → EA&B, embrace it.
- The contribution drifted into the model, not the data path → ML venue.
- Twelve pages cannot hold the proofs → PODS split or journal lane.
Output format
[Primitive] <one sentence> / absent (re-route)
[Category] regular / EA&B / SDS / vision — with page-budget check
[Venue ranking] <top choice + two alternates, one reason each>
[Convincer plot] <the decisive figure, described>
[Risk] <novelty / evidence scale / fit — the one that kills it>
[Next action] <build, measure, reframe, or switch venue>
