Evaluate the failure class.
Keep the viewer claim honest.
A bounded evaluation for a streaming organization that wants to examine one reliability failure class with agreed evidence, permissions, and viewer-impact checks—without treating a controlled evaluation as a production guarantee.
Claim status labels: see definitions.
10 business days from evidence-ready kickoff
One agreed failure class, one evidence-ready window, one readout.
Infrastructure-green alone never closes the viewer lane.
01 · qualify
The engagement starts with a bounded operating question.
This is for a named streaming team that can authorize evidence and agree what a useful verdict means. It is not an informal trial or a request for blanket access.
Streaming organization
Your team operates or owns an authorized streaming reliability workflow that can be discussed and evaluated.
Named technical owner
One person can coordinate the scenario, evidence handoff, access questions, and readout decisions.
One agreed failure scenario
The evaluation starts with a failure class that can be described clearly enough to define evidence and a viewer cohort.
Permission to provide evidence
Your organization can authorize the least-privilege evidence access or approved fixture needed for the scenario.
Agreed evaluation window
The team can reserve a defined evidence-ready window for kickoff, review, and readout.
Read-only by default
Evidence access begins with the least privilege needed to read approved source events, player telemetry, logs, metrics, or timelines.
Customer-controlled credentials
Your team creates, scopes, rotates, and revokes any credentials or tokens used for the handoff. Never send secrets through the public intake form.
No blanket production write access
The program does not require broad production mutation rights. Any controlled action must use the existing approval-gated recovery lifecycle or a declared fixture.
Public intake is qualification only
The form stores buyer and scenario context. It is not an evidence uploader, secret exchange, or public incident registry.
Admin review is PII-gated
Submission details and review notes are available only through the authenticated admin workflow; raw message content stays out of list/detail JSON.
Private evidence stays scoped
Private source and inbound routes retain their session or token gates and tenant scoping. Evidence is not publicly listed.
No autonomous remediation promise
The program may document governed recovery verification, but it does not promise unattended production changes or a guaranteed viewer outcome.
02 · evidence
Use current typed paths, not a speculative upload portal.
The evaluation inventory is built from the private evidence and integration surfaces that already exist. Every row below says whether the path is supported, conditional, or out of scope.
Typed vendor and source events
Agreed source events such as CloudFront, CloudWatch, CloudTrail, and other currently supported vendor paths can anchor the timeline when configured and authorized.
Player-side CMCD
Player-side CMCD can be correlated with infrastructure evidence where the affected cohort and time window are available.
Logs, metrics, and timelines
Relevant logs, metrics, incident context, action/approval/verification rows, and the agreed scenario timeline are collected when the current source path records them.
Viewer-recovery status
Viewer recovery is evaluated only when downstream player evidence supports it. Infrastructure-green alone does not close the viewer lane.
Unsupported evidence
Arbitrary file uploads, unverified customer results, unsupported vendor integrations, and claims of viewer recovery from infrastructure-green alone are excluded from this program.
03 · scenarios
Choose a scenario the evidence can actually answer.
The examples below are current source-backed or controlled patterns. Synthetic Incident Lab material helps align vocabulary, but it is never presented as a customer result.
Compare healthy-looking upstream or infrastructure signals with downstream viewer impact and identify the evidence gap between them.
Source events plus affected-cohort player telemetry and a documented timeline.
Trace a controlled edge or origin-shield condition through source evidence, governed action records, and viewer-side verification.
CloudFront or related source evidence, action/approval rows where applicable, and a viewer-impact signal.
Correlate CMCD from an affected cohort with the agreed infrastructure window to separate an operational signal from a viewer outcome.
CMCD rows with a clear time window, cohort context, and enough post-action data to assess stability.
Synthetic examples can help align on a scenario and vocabulary, but they are illustrative and cannot be presented as customer results or production readiness.
Illustrative fixture data only; not a substitute for customer-authorized evidence.
04 · verdict
Success means infrastructure evidence and viewer evidence agree.
The readout distinguishes an operational improvement from a viewer-recovery conclusion, checks recurrence, and names gaps instead of smoothing them over.
Original failure class is evidenced
The evaluation records infrastructure/source evidence that matches the agreed failure class instead of substituting a different signal.
Affected cohort is evaluated downstream
The evaluation includes viewer-side evidence for the affected cohort whenever the current source path can provide it.
Recurrence and stability are checked
The readout records whether the signal recurred and whether the agreed post-action window remained stable.
viewer_confirmed may close the viewer lane
Use viewer_confirmed only when downstream viewer evidence meets the current recovery policy for the scenario.
infra_only remains open
Infrastructure improvement without downstream viewer confirmation is reported as infra_only and remains an open or qualified gap.
insufficient_data remains open
Missing or inadequate post-action viewer evidence is recorded as insufficient_data and is not converted into a success claim.
Five-minute stability rule is conditional
Reuse the existing five-minute post-action stability rule only when the selected recovery policy supports it; do not generalize it to every scenario.
05 · delivery
Know what arrives at the end of the window.
The program produces a bounded evaluation record, not a benchmark score or an automatic production rollout.
Evaluation brief
Agreed scope, owner, scenario, permissions, evidence window, and decision questions.
Evidence inventory
A record of received source, player, timeline, action, approval, and verification evidence plus gaps.
Scenario timeline
The ordered failure, investigation, action, verification, and viewer-impact events available for the evaluation.
Findings and limitations
What the evidence supports, what it does not support, and which claims remain unverified.
Viewer-recovery verdict
A viewer_confirmed, infra_only, or insufficient_data outcome where the current evidence supports one.
Blocker and gap register
Open evidence, access, owner, and next-step blockers recorded for follow-through.
Readout and paid-path recommendation
A human readout with a bounded recommendation for a fixed-fee review or retained reliability practice.
Evidence received or insufficiency recorded
The agreed evidence is present, or the missing evidence and consequence are explicitly recorded.
Scenario verdicts documented
Each agreed scenario question has a finding, limitation, or unresolved gap.
Unsupported claims separated
Illustrative examples, unverified customer results, and unsupported capabilities are kept separate from findings.
Owner and next step recorded
Open blockers have an owner and next action, even when the evaluation stops without a paid engagement.
Final readout delivered
The agreed stakeholders receive the final readout and any bounded paid-engagement recommendation.
06 · buyer questions
The evidence questions buyers ask before kickoff.
These answers keep the evaluation boundary visible: what can be assessed, what the reference window means, and where illustrative material stops being evidence.
07 · commercial terms
A short, human-led operating window.
This is the evaluation offer’s operating term covering kickoff, evidence handoff, review, and readout. It is commercial policy, not a verified product capability or SLA.
Contact channel: A named Streamwake contact through the existing contact form and the agreed human conversation. No 24/7 response or undocumented vendor-side access is promised.
Paid path: A successful or appropriately bounded evaluation can lead to a scoped human conversation about a fixed-fee review or retained reliability practice. There is no self-serve checkout or automatic conversion.
08 · start qualification
Bring the buyer, the boundary, and the evidence question.
Tell us who owns the workflow, which scenario matters, what read-only boundary is acceptable, and what a useful viewer-recovery signal would look like. Do not submit secrets.
Public intake stores qualification context only.
Private evidence handoff follows the agreed authenticated or token-scoped path.
A person reviews the request and scopes the next conversation.