Admin

Connectors live
4 / 73 degraded or absent
Join coverage
97.7%Weighted mean across user-joinable sources · floor 95%
Below the floor
1Sources whose metrics are degraded to partial
Behind their latency target
1Freshness breach — alertable in its own right

Connectors

RSHFT

Redshift

Owner · AWG
StateLive
Last sync12m ago
Target60m
24h volume4.2m
Join coverage100%
Volume vs. expected102.0% of 4.1m

Commercial system of record. Join key is dim_user → AUMS UUID, so coverage is definitionally complete.

Access · Read-only user on replica · Redshift Data API
AWG

AWG platform API

Owner · AWG
StateLive
Last sync4m ago
Target15m
24h volume812.9k
Join coverage99.8%
Volume vs. expected101.6% of 800k

Entitlement handshake and subscription lifecycle transitions.

Access · REST, service account
FIRE

Firebase

Owner · SimpleStream
StateLive
Last sync38m ago
Target60m
24h volume21.4m
Join coverage96.1%
Volume vs. expected102.1% of 21m

user_id property carries the AUMS UUID on 96.1% of events. The 3.9% gap is pre-auth sessions, which cannot resolve by design.

Access · BigQuery daily + streaming exportSchema drift · user_properties.plan_tier added 26 Aug
NPAW

NPAW

Owner · SimpleStream
StatePartial
Last sync26m ago
Target30m
24h volume9.7m
Join coverage94.2%
Volume vs. expected86.3% of 11.2m

Highest-risk dependency in the project. Volume is 14% below expected because Tizen and tvOS payloads are not yet in the S3 feed; QoE metrics for those platforms are labelled partial everywhere they appear.

Access · Interim S3 raw-data delivery — Data API access still open (§16 Q4)
DDOG

Datadog

Owner · SimpleStream
StateNot connected
Last syncnever
Target15m
24h volume
Join coveragenot user-level

No metric in this tool depends solely on Datadog. APM traces and RUM sessions are absent from every reading until this connects.

Access · Awaiting API key — instrumentation at very early stages
CW

CloudWatch CDN

Owner · AWG
StateLagging
Last sync1h 34m ago
Target60m
24h volume58.2m
Join coveragenot user-level
Volume vs. expected97.0% of 60m

Edge logs carry a request id, not an AUMS UUID. Treated as a cohort-level signal only — never attributed to an individual user.

Access · S3 standard access logs (batch)
ZD

Zendesk

Owner · Sony L1
StateLive
Last sync2m ago
Target5m
24h volume1.3k
Join coverage98.4%
Volume vs. expected107.0% of 1.2k

Requester email resolves through dim_user. The 1.6% gap is tickets raised from addresses with no matching subscription.

Access · Zendesk API + webhooks via Arcadian

Open data-health issues

  • NPAW

    Interim S3 raw-data delivery — Data API access still open (§16 Q4)

    Partial
  • Datadog

    Awaiting API key — instrumentation at very early stages

    Not connected
  • CloudWatch CDN

    S3 standard access logs (batch)

    Lagging
  • Firebase

    user_properties.plan_tier added 26 Aug

    Schema drift

What a degraded source does to a reading

  • PartialThe metric is computed over the population that did resolve and labelled partial everywhere it appears. It is never silently averaged with the missing group.
  • AbsentNo metric in the tool depends solely on Datadog, so its absence removes APM and RUM detail without invalidating any headline figure.
  • Cohort-onlyCloudWatch carries a request id rather than an AUMS UUID, so edge findings are reported as cohort signals and never attributed to a person.
  • AlertableCoverage and freshness failures raise alerts of their own — 1 is open now against the NPAW feed.

Ingestion run history

10 runs · 2 failed · 4.9m loaded · 11,166 rejected
RunSourceStatusStartedDurationRows loadedRejected
IR-88412ZDSuccess4m ago2m18
IR-88411AWGSuccess9m ago5m34,120
IR-88409RSHFTSuccess18m ago6m174,280
IR-88407NPAWPartial32m ago6m402,680
IR-88404FIRESuccess44m ago6m894,1121,284
IR-88394RSHFTSuccess1h 18m ago7m168,940
IR-88398CWSuccess1h 40m ago6m2,425,500
IR-88396NPAWFailed2h 32m ago3m0
IR-88390FIREPartial3h 44m ago10m812,0049,882
IR-88387CWFailed5h 40m ago4m0
  • IR-88407Loaded 5 of 7 platform partitions. Tizen and tvOS partitions absent from the S3 prefix — not an error, the interim feed does not carry them.
  • IR-884041,284 events rejected: user_properties.plan_tier appeared on 26 Aug and is not in the target schema. Rejects are quarantined, not dropped.
  • IR-88398Batch delivery. 94 minutes behind a 60-minute target — the lag is the delivery cadence, not a fault.
  • IR-88396S3 GetObject 403 on prefix npaw/2026/08/28/11. Credential rotated by NPAW without notice; retried successfully at 12:14.
  • IR-88390Elevated reject rate during the 3.4.1 rollout — 9,882 events carried an app_version string the dimension did not yet contain.
  • IR-88387Manifest for the 08:00 batch never arrived. Backfilled by IR-88398.

Retention against the history the detectors need

  • RSHFTRedshiftretains 730dneeds 90d640d headroom
  • AWGAWG platform APIretains 365dneeds 90d275d headroom
  • FIREFirebaseretains 60dneeds 30d30d headroom
  • NPAWNPAWretains 30dneeds 28d2d headroom
  • DDOGDatadogretains 15dneeds no detectors
  • CWCloudWatch CDNretains 30dneeds 28d2d headroom
  • ZDZendeskretains 1095dneeds 30d1065d headroom