Infrastructure Decisions · judgement tool

Do I need a message queue?

Answer your task duration, in-request external calls, retry needs and stage — get a Not yet / Probably yes / Yes-priority verdict. Every factor tagged as official fact, calculation, StackSays rule or estimate. No universal traffic threshold implied.

A judgement tool, not a tool ranking · thresholds are StackSays rules, not industry standards · official facts verified 2026-07-17 · share the URL to share your answers

How to read this: every factor below is tagged by where it comes from — official fact official/documented · calculation math on your inputs · StackSays rule our decision logic · StackSays estimate our judgement. The verdict cut-points are StackSays rules, not an industry threshold — there is no universal traffic number for these decisions.

Your situation

The recommendation updates as you answer. Share the URL to share it.

Longest task in a requestwork you'd do while the user waitsCalls slow external APIs in-requestpayment, email, LLM, third-party during the responseWork must retry on failuree.g. must-deliver webhooks, payment side-effectsSends email / webhooksoutbound notificationsOn serverless (Vercel/Lambda/Workers)functions with an execution time limitStagehow far along the product is

Recommendation

Not yet

Based on the factors you set — 0 pushing toward adopting, 2 against.

Why it might be premature

-12 Too early — At prototype stage a queue is usually premature — ship first, add it when a real need above appears (our judgement)StackSays estimate
-6 Long in-request work — Sub-second work is fine to do inline — no queue needed for duration aloneStackSays rule

When to reassess

Revisit the moment a task crosses ~10s, you add an in-request external call, or you need guaranteed delivery — those flip this to yes.

Minimal implementation

Start with the smallest thing: your platform's built-in cron, or a hosted HTTP queue (e.g. Upstash QStash) — not a full workflow engine. Adopt durable workflows (Inngest/Temporal) only when you have multi-step orchestration.

Risk of not doing it

Request timeouts, latency coupled to third parties, lost work on failure, and a DB hammered by synchronous spikes.

Cost & complexity once adopted

A hosted queue is low effort (an HTTP call + a handler). A durable workflow engine adds a new billing meter and a learning curve.

This is a judgement tool, not a leaderboard. The StackSays rule and StackSays estimate factors are StackSays logic and opinion; the official fact factors link to official sources below. No industry-standard threshold is implied — the answer is driven by the conditions you set.

Official facts cited

Source	Verified
Vercel function duration limits (official)	2026-07-17
AWS Lambda timeout limit (official)	2026-07-17

Only the factors tagged “official fact” rest on these sources; rules and estimates are StackSays logic and are labeled as such.

Next decision

Where should background jobs run? →Do I need Redis? →