Data cleaning
AuraScore 89/100

OTT Telemetry and Viewer Session Scrubbing Architecture

Design a data cleaning framework to eliminate noise, bot traffic, and telemetry anomalies from high-volume streaming video event logs.

Use this template when raw client-side player logs produce skewed watch-time figures or corrupted session metrics due to network drops, beacon spam, or web scrapers. It outputs a rigorous telemetry sanitation framework for downstream analytics pipelines.

Template

Role: Principal Media Telemetry Analyst specializing in OTT quality-of-experience (QoE) instrumentation and streaming data pipelines.

Context

  • Streaming Tier: {{streaming_service_tier}}
  • Ingestion Stream Format: {{telemetry_ingest_format}}
  • Heartbeat Interval: {{heartbeat_interval_seconds}}
  • Bot Filtering Thresholds: {{bot_detection_criteria}}
  • Inactive Session Cutoff: {{session_timeout_threshold}}
  • Downstream Analytics Target: {{downstream_analytics_engine}}

Task

Construct a robust telemetry data cleaning and sessionization framework that purges synthetic traffic, interpolates missing heartbeats, reconciles out-of-order event sequences, and produces trusted viewer session tables in {{downstream_analytics_engine}} for {{streaming_service_tier}}.

Method

  1. Analyze the structure of raw event payloads from {{telemetry_ingest_format}} to identify malformed JSON, corrupted client headers, and invalid timestamp encodings.
  2. Define stateless payload validation rules to discard unparseable packets and log malformed telemetry rates.
  3. Formulate a multi-variable bot filtering filter leveraging {{bot_detection_criteria}} to flag and isolate non-human playback activity.
  4. Design sessionization reconstruction logic using {{session_timeout_threshold}} and {{heartbeat_interval_seconds}} to stitch fragmented playback events into contiguous viewing intervals.
  5. Implement a chronological ordering and deduplication window to resolve network retry duplication and clock-skew discrepancies from mobile and smart TV clients.
  6. Develop an imputation and interpolation method for dropped intermediate heartbeats that prevents artificial inflation of Quality of Experience (QoE) rebuffering metrics.
  7. Create derived cleaning flags to separate legitimate short-form browsing from abandoned streams or playback initiation errors.

Constraints

  • MUST maintain an immutable audit trail of filtered events with specific error reason codes.
  • MUST NOT recalculate playback duration by simple timestamp delta if intermediate pause or seek events are missing.
  • Cleaning latency MUST remain compatible with the streaming buffer window of {{downstream_analytics_engine}}.
  • Session stitching algorithms MUST handle out-of-order events arriving up to 10 minutes late.

Output format

Provide the architectural framework formatted with these mandatory sections:

  1. Raw Stream Ingestion Filtering Rules (table of payload validity checks)
  2. Synthetic & Bot Traffic Cleansing Protocol (detection heuristics and exclusion criteria)
  3. Session Reconstruction & Deduplication Algorithm (step-by-step logic and time-window rules)
  4. Metric Imputation & Anomaly Remediation Standard (handling missing pings and clock drift)
  5. Cleaned Schema Definition for Downstream Delivery (target column specification) Limit total length to between 650 and 950 words.

Self-review

  • Are the deduplication rules explicitly accounting for mobile network retry behaviors?
  • Does the framework reference all variables: {{streaming_service_tier}}, {{telemetry_ingest_format}}, {{heartbeat_interval_seconds}}, {{bot_detection_criteria}}, {{session_timeout_threshold}}, and {{downstream_analytics_engine}}?
  • Are the cleaning steps directly focused on QoE and telemetry rather than static catalog data?
AuraScore breakdown
89/100Provisional
Instruction clarity15/15 · Strong

Explicit role, a named task, and discrete steps the model can follow.

Context architecture12/12 · Strong

Background, inputs and variables the model needs before it starts.

Constraint engineering12/12 · Strong

Hard boundaries — what the model must and must not do.

Output specification14/14 · Strong

A named, field-level shape for the response.

Reasoning structure10/10 · Strong

Ordered work items that force analysis before an answer.

Model compatibility10/10 · Strong

Length and structure that travel across frontier models.

Token efficiency5/10 · Thin

Signal density — instruction weight without padding.

Reusability7/7 · Strong

Documented variables so the scaffold adapts to new inputs.

Robustness3/5 · Adequate

Quality bar, assumptions and behaviour when inputs are thin.

Observed performance1/5 · Thin

How much real usage the template has behind it.

data-analytics
data-cleaning
media-entertainment
telemetry-cleaning
ott-analytics
stream-processing