ZMP SLAs | Overview and Interpretation

image-20260519-053819.png

Zeta’s Service Level Agreements (SLAs) outline our commitments to availability, performance, and throughput across the platform over a 30-day window. This article summarizes those commitments, explains percentile-based measurements (P50/P95/P99) in customer-friendly terms, and clarifies scope so you can plan integrations and operations with confidence.

For custom requests, please reach out to your Account Manager.

How to Read These SLAs

  • Percentiles (P50/P95/P99): P50 is the median (half of requests are faster, half slower); P95 means 95% of requests complete within the stated time; P99 indicates the near-worst case we design for in normal operations.

  • Latency SLAs apply under noted load conditions (e.g., requests per second or per hour). If traffic exceeds stated rates, latencies may rise until auto-scaling catches up.

  • Availability reflects platform uptime over a monthly window. Scheduled maintenance and events affecting upstream providers are handled according to your contract and notice commitments.

  • Throughput limits (e.g., batch imports/exports, campaign send rates) are aggregated across tenants or across a customer’s workloads as specified in notes.

SLA Summary

All metrics below are stated as service objectives under defined load envelopes. Clarifications indicate scope and operational expectations.

Category

Type

P50

P95

P99

Notes/Clarification Points

Platform Availability

up-time

—

—

99.90%

Often higher (99.99%) for customer-facing
real-time APIs

Profile Read APIs

latency

≤ 200ms

≤ 500ms

≤ 1s

Applicable at peak rate of 500k/hr (150/s)

Profile Write API (Identify)

latency

≤ 500ms

≤ 750ms

≤ 1s

Applicable at peak rate of 10M/hr (3,000/s)

Event Write API (HPM)

latency

≤ 500ms

≤ 750ms

≤ 1s

Where profiles already exist.
Applicable up to volumes of 500k/hr (150/s)

Event Write API (Non-HPM)

latency

≤ 30s

≤ 1m

≤ 2m

Applicable at peak rate of 10M/hr (3,000/s)

Experience Non-Triggering
Decision Latency (downstream,
incl. Data Cloud re-
identification)

latency

≤ 1m

≤ 3m

≤ 5m

Applicable at peak rate of 10M/hr (3,000/s)

Event Read API

latency

≤ 1s

≤ 1.5s

≤ 2s

Applicable at peak rate of 500k/hr (150/s)

Resource Read

latency

≤ 200ms

≤ 500ms

≤ 1s

Applicable at peak rate of 500k/hr (150/s)

Resource Write

latency

≤ 500ms

≤ 750ms

≤ 1s

Applicable at peak rate of 10M/hr (3,000/s)

Data Imports Batch
(Profiles, Events,
Resources)

volume

—

—

  • Low - 1M/hr

  • Medium - 2M/hr

  • High - 5M/hr

  • Super - 10M+/hr

Events are behavioral data
from varied sources
(web, mobile, other).
Volumes are aggregate
across imports.

Data Exports Batch
(Profiles, Events)

volume

—

—

  • Low - 1M/hr

  • Medium - 2M/hr

  • High - 5M/hr

  • Super - 10M+/hr

Exports deliver events and/or
users to
client-specified locations.
Volumes are aggregate
across exports.

Data Exports (Real-Time)
- Webhook

volume

 

 

  • Low - 1M/hr

  • Medium - 2M/hr

  • High - 5M/hr

  • Super - 10M+/hr

Exports deliver events and/or users to
client-specified locations.
Volumes are aggregate
across exports.

Campaign Batch Send

volume

—

—

  • Low - 2.5M/hr

  • Medium - 5M/hr

  • High - 18M/hr

  • Super - 40M+/hr

Audience pre-processing and
full message pre-building enable
fastest scheduled send rates.
Measured against hours exceeding
SLA volume.

For any questions, contact support: hubsupport@zetaglobal.com

Zeta Global - Proprietary & Confidential Information: Date Last Updated -