Service Level Agreement

Availability

Aleatoric Data defines availability as the proportion of time the platform successfully serves requests within a calendar month. Formally:

$A = \frac{T_{\text{total}} - T_{\text{downtime}}}{T_{\text{total}}} \times 100\%$

where $T_{\text{total}}$ is the total number of minutes in the measurement month, and $T_{\text{downtime}}$ is the cumulative minutes during which the service returned errors to more than 5% of requests within any consecutive 1-minute window.

Availability is measured per region. A regional outage in US-East does not affect the SLA calculation for AP-Tokyo, and vice versa. Clients routing traffic to multiple regions benefit from independent availability guarantees on each.

Uptime Commitments by Tier

Tier	Monthly Uptime Target	Financial Remedy
Free	Best effort	None
Pro	99.5% ($\leq 219$ min downtime/month)	None
Enterprise	99.9% ($\leq 43$ min downtime/month)	Service credits

Free tier operates on shared infrastructure with no uptime guarantee. Service may be interrupted during maintenance windows, capacity events, or upstream outages without notice.

Pro tier targets 99.5% monthly availability. While no financial credits are issued, persistent degradation below this target is escalated internally and addressed with infrastructure changes.

Enterprise tier is backed by a contractual 99.9% SLA with service credit remedies as defined below.

Latency Targets

Latency is measured as the interval from ingest (the moment a data event is received at the Aleatoric gateway from the upstream P2P network or exchange feed) to publish (the moment the corresponding encoded event is delivered to the client’s transport socket).

For a given percentile $p$, the latency target $L_p$ is defined as:

$L_{p} = \inf\!\big\{l : P(L \leq l) \geq p/100\big\}$

That is, $L_p$ is the smallest latency value $l$ such that at least $p\%$ of all measured requests complete within $l$ time units. This is the standard quantile definition applied to the empirical latency distribution over each calendar month.

Protocol Latency Targets

Protocol	$L_{50}$ (median)	$L_{95}$	$L_{99}$	Measurement Point
JSON-RPC	< 50 ms	< 150 ms	< 500 ms	HTTP response body flush
gRPC	< 10 ms	< 50 ms	< 150 ms	Protobuf message write to stream
Unified Stream (SSE)	< 500 $\mu$s	< 2 ms	< 10 ms	SSE data: line written to socket

Notes:

• JSON-RPC latency includes request parsing, handler execution, and response serialization. It is inherently higher than push-based protocols due to the request-response cycle.
• gRPC latency applies to server-streaming RPCs after the initial connection handshake. Unary RPCs have latency characteristics similar to JSON-RPC.
• Unified Stream latency reflects the push pipeline only — from internal event bus to socket write. This is the lowest-latency delivery path available on the platform.
• All latency targets assume the client is connected to the geographically nearest region. Cross-region routing adds network transit time that is outside the SLA measurement boundary.

Incident Severity Classification

Incidents are classified according to their scope and user impact:

Severity	Definition	Example
P1 — Critical	Complete service outage affecting all users in a region	Gateway unreachable, all feeds returning 5xx
P2 — Major	Degraded service with partial functionality loss	One protocol down, elevated error rates > 5%
P3 — Minor	Isolated issue affecting a single feed or non-critical function	Single coin feed stale, dashboard rendering error
P4 — Low	Cosmetic or informational issue with no data impact	Documentation error, status page lag

Incident Response Matrix

Severity	Initial Response	Status Update Cadence	Resolution Target
P1 — Critical	15 minutes	Every 30 minutes	4 hours
P2 — Major	1 hour	Every 2 hours	8 hours
P3 — Minor	4 hours	Daily	24 hours
P4 — Low	1 business day	As needed	5 business days

Enterprise clients receive direct incident communication via their configured channel (Slack Connect, PagerDuty webhook, or email distribution list) for P1 and P2 incidents. All other tiers receive updates via the public status page at status.aleatoric.systems.

Service Credits (Enterprise Only)

When monthly availability in a region falls below the Enterprise SLA target, credits are issued as a percentage of the monthly fee attributable to that region:

Monthly Uptime	Credit (% of monthly fee)
99.0% — 99.9%	10%
95.0% — 99.0%	25%
< 95.0%	50%

Credit terms:

• Credits must be requested in writing within 30 calendar days of the end of the affected month.
• Credits apply to the affected region only. Multi-region deployments are credited proportionally.
• Credits are issued as account balance and cannot be redeemed for cash.
• The maximum aggregate credit for any single month is capped at 50% of that month’s total fee.
• Credits are the sole and exclusive remedy for failure to meet the SLA.

Exclusions

The following are excluded from availability and latency calculations:

• Force majeure events — natural disasters, war, government action, or other events beyond reasonable control.
• Upstream network outages — Hyperliquid validator downtime, consensus failures, L1 chain halts, or exchange API outages at reference venues (Binance, OKX, Bybit) that prevent data ingestion at the source.
• Client-side errors — misconfigured API keys, invalid request parameters, client network issues, or DNS resolution failures.
• Scheduled maintenance — planned maintenance windows announced at least 48 hours in advance via the status page and Enterprise notification channels. Maintenance windows are limited to 4 hours per month.
• Beta features — any endpoint, feed, or protocol explicitly marked as “beta” in the documentation operates on a best-effort basis outside the SLA.
• Rate limiting — 429 responses resulting from clients exceeding their tier’s rate limits are not counted as downtime.

Measurement and Reporting

Availability and latency are measured by Aleatoric’s internal monitoring infrastructure using synthetic probes and real traffic telemetry. Enterprise clients may request monthly SLA compliance reports showing per-region uptime percentages and latency percentile distributions. Reports are delivered within 5 business days of month-end.