-
Most exchange failures happen in the first 72 hours, not because the platform was not built, but because liquidity was never validated per pair at launch volume.
-
This checklist covers 26 items across four phases: Pre-Build, Build, Validation, and Post-Launch.
-
It includes a self-assessment readiness score, clear 7.2 / 10 before you go live.
Most exchanges that fail do not fail on the engineering. They fail in the first three days, when the order books that looked healthy in testing widen the moment real volume arrives and the market makers start watching execution quality. A 9% spread on a listed pair is not a bug report, it is market makers quietly cancelling agreements and users quietly leaving. The work below is the work that decides whether Day 1 is a launch or a slow-motion withdrawal. Complete it and you will know, with a number, not a feeling. whether your exchange is ready to take its first real trade.
| Phase | Timing | Items | What breaks if you skip it |
|---|---|---|---|
| 1Pre-Build Foundation | 12–16 weeks out | 6 | Wrong jurisdiction, wrong liquidity partner, unrecoverable custody decisions |
| 2Infrastructure Build | 4–8 weeks out | 8 | Cold-start liquidity failure, AML gap at launch, open security surface |
| 3Pre-Launch Validation | 1–3 weeks out | 7 | Day 1 matching-engine failure, regulatory rejection, trust collapse at signup |
| 4Post-Launch Monitoring | Week 1–4 after live | 5 | Market-maker withdrawal, compliance breach, silent user churn |
Pre-Build Foundation
The single most expensive Crypto Exchange mistake is made here — the decisions that are cheapest to fix now and most expensive once build has started.
Pick the jurisdiction before anything else it determines your banking partners, your AML configuration, and which licenses you even need. Get a regulatory counsel opinion in writing for the UAE, not a forum thread.
Shortlist two Market Making APIs providers and run a real integration spike against your stack before you commit. Validate latency, failover, and the actual data shape not the sales deck.
Decide your compliance posture now and choose AML + Identity Verification APIs to match it. Map every flow that touches a user identity or a transaction so the pipeline is designed for AML, not retrofitted.
Model your unit economics at expected volume. Put Market Making APIs cost, compliance cost, and infrastructure cost in one sheet against your fee structure. If the math only works at 10× your launch volume, you do not have a model yet.
Decide custodial vs non-custodial vs MPC, and document the key-management model your auditor will sign off on. This is the most expensive decision to reverse.
Start banking and fiat-ramp conversations in Phase 1, they are the longest pole in the tent. Get a term sheet from a ramp partner that actually serves the UAE before you design the deposit flow.
Infrastructure Build
This is where most platforms that fail, fail. Each item below is something that is far harder to add after beta than before it.
The altcoin book collapse
One exchange launched with two external market makers connected through a single API gateway. BTC and ETH depth looked healthy all through pre-launch testing. By hour six on Day 1, altcoin spreads had widened to 9.4% on six of eight listed pairs. Four market makers cancelled their agreements within 48 hours, citing execution quality.
Stand up Kafka + Kubernetes for scale from day one. Configure auto-scaling, partitioning, and back-pressure handling and load the config into version control so it is reproducible, not tribal knowledge.
Connect Market Making APIs and set per-pair depth targets before beta. Validate each pair independently a healthy BTC book tells you nothing about your thin pairs.
Wire AML + Identity Verification APIs into live transaction monitoring and run a real suspicious-transaction-report dry run end to end. A configured-but-untested AML stack is not a compliant one.
Build the KYC flow and instrument every step. Run it with real external testers and record the drop-off. The number you get is the number you launch with unless you fix it now.
Lock down the matching engine and segregate hot/cold wallet flows, then commission a focused pen-test of the withdrawal path specifically. The withdrawal function is where exchanges lose funds.
Build the reporting pipeline and submit a test file in VARA's required format before you need to. Format rejections are discovered at the deadline by everyone who skips this.
Configure matching priority, self-trade prevention, and the maker-taker fee schedule, then replay historical order flow against it. Confirm the engine behaves under cancel-storms, not just steady state.
Define circuit breakers, rate limits, and a written incident-response runbook with named owners. Rehearse the first 15 minutes of an incident before you have one.
From Troniex’s implementation work with Crypto Exchange operators
Across exchange builds, the integration that breaks most often is not the matching engine, it is the seam between the market-making API gateway and per-pair risk limits. Teams connect one gateway, see healthy majors, and never test what happens when a single MM drops mid-session on a thin pair. Build per-pair failover before launch, not after the first withdrawal.
Pre-Launch Validation
The last gate before a real user touches it. Everything here is about discovering the failure in a drill instead of in production.
| What to validate | Pass threshold | Fail signal | Resolution |
|---|---|---|---|
| Order execution latency at peak | < 25 ms | Queue backup visible | Scale Kafka before re-test |
| KYC completion rate | > 65% of cohort | Drop-off at selfie step | UX fix on friction point |
| AML false-positive rate | < 4% | Compliance team overloaded | Reconfigure ML threshold |
| Uptime under 3× load | > 99.9% | Auto-scaling not triggering | Infrastructure config review |
| Liquidity spread per pair | < 1.5% top 8 pairs | MM withdrawal risk | Add MM before go-live |
| Regulatory report format | Accepted by test | Format rejection | Reporting pipeline reconfig |
Run a sustained load test at 3× projected peak, not average. Watch the matching queue, the database, and the auto-scaler under pressure and capture where the first thing bends.
Commission a third-party audit with an explicit scope document, then remediate every critical and high finding before launch. Get the scope in writing before you get the report.
Run your compliance stack against VARA's published test scenarios and tune thresholds against the results not against defaults shipped by the vendor.
Re-run the onboarding flow with a fresh external cohort and confirm completion clears your benchmark. Fix the friction point you find before, not after, you spend on acquisition.
Confirm depth on every launch pair at peak volume, pair by pair. Add a market maker before go-live for any pair that cannot hold its spread target.
Run a real failover drill: kill the primary, time the recovery, verify data integrity on the other side. An untested BCP is a document, not a plan.
Rehearse Day 1 with your real listing pairs, real market makers connected, and the team on the bridge. Walk the first hour minute by minute so the launch is a repeat, not a premiere.
The load test that passed at 2×
A team load-tested their matching engine at 2× projected peak and called it green. On Day 1 a listing-driven spike hit 2.4× for eleven minutes. The order-matching queue backed up by 11 seconds, fills printed against stale prices, and the support channel filled with screenshots before engineering even saw the alert.
Post-Launch Monitoring
The first 30 days decide whether the launch holds. Item 26 is the retention mechanic unique to this platform.
Review execution quality and per-pair liquidity every day for the first two weeks. The early signal of a failing launch shows here first.
Watch the live onboarding funnel and isolate the real drop-off step within the first 72 hours, while you can still act on it.
Review the AML queue weekly. A false-positive rate creeping up is an early compliance-cost problem you want to catch before it buries the team.
Submit the first regulatory report on time, in the validated format. The first one sets your standing with VARA.
Turn on the retention loop that keeps traders: tiered maker-taker fees and native-token utility (fee discounts, staking). Seed it Week 1 while attention is highest.
Is your Crypto Exchange ready to launch?
Each category fills in automatically as you tick the checklist above — and the verdict updates live. Drag any slider to model a what-if before you commit.
Score your readiness
Tick the checklist above — categories fill in automatically and the verdict updates live.
Get the Crypto Exchange Launch Readiness Report
A documented PDF — your scores, gap analysis, go/no-go decision, and full checklist record.
Where Crypto Exchange launches actually fail
1Per-pair liquidity assumed from majors
BTC and ETH depth gets validated; the long-tail pairs are assumed to behave the same way. They do not. The first thin-book spread blowout is almost always a pair nobody stress-tested individually.
2A single market-making gateway with no failover
Every MM routed through one API gateway means one outage takes the whole book offline. On a launch day that is not a degraded experience, it is a frozen exchange.
3The narrative one
They had tested KYC with internal users who already knew the drill. Six of twelve external testers in the first beta stalled at the selfie step, the camera permission was triggering an OS-level warning that made the app look like it was asking for something shady. Nobody caught it because internal users had already granted the permission months ago. They only found out because one tester texted a screenshot to the founder instead of abandoning silently, the way the other five did.
4Fee logic that does not survive a cancel-storm
Maker-taker logic tested in steady state behaves differently under a wave of cancels and re-quotes. Exchanges discover the edge cases when a market maker disputes a settlement, not during QA.
5No launch-day rehearsal
The team that has never walked through the first hour together discovers their incident roles during the incident. Predictable problems become live ones because nobody rehearsed the boring parts.
Frequently asked questions
Let’s Get in Touch.
Unsure whether your Crypto Exchange is ready to go live? Our infrastructure team will pressure-test your readiness and map the gaps with honest, practical advice — ABSOLUTELY FREE.
Book Your Free Consultation
