# Metric Triage Agent

*/Opportunities/Metric_Triage_Agent*

## Opportunity Overview

**Wedge**: The initial beachhead targets investigating conversion rate drops in product-led growth SaaS companies. This niche provides a highly quantifiable pain point with immediate revenue implications and standardized event tracking via tools like Amplitude. Expansion proceeds sequentially from product funnel metrics to financial metrics like billing failures, and finally to operational metrics like infrastructure costs.
**Timing**: Large language models now reliably execute multi-step reasoning over structured data, translating anomalous alerts into dynamic SQL queries. The standardization of the modern data stack around Snowflake and dbt provides the necessary semantic layer for an agent to navigate corporate data autonomously.
**Why This I C P**: Mid-market B2B SaaS data teams operate with lean analyst ratios relative to their data volume. They experience acute bottlenecks when executives demand immediate explanations for daily revenue or engagement fluctuations, making them eager adopters of automation.
**Size Of Prize**: 40,000 mid-market and enterprise B2B software companies globally multiply by a $30,000 annual spend on data automation software to yield a $1.2B addressable market.
**Gap Narrative**: Data analysts and growth teams manually query databases for hours to diagnose unexpected metric fluctuations. Existing business intelligence dashboards only surface the anomaly, forcing humans to write repetitive SQL to find the root cause. This agent investigates the metric drop autonomously by correlating it with underlying dimensional changes, recent deployments, and system logs.
**Defensibility**: Defensibility builds through a proprietary knowledge graph of the company's business logic. As the agent learns how a specific company defines its metrics, handles edge cases, and maps raw tables to business concepts, switching to a generic alternative requires rebuilding months of institutional context from scratch.
**Why This Thesis**: An agent approach aligns with the dynamic nature of root cause analysis. The system formulates a hypothesis, queries the data warehouse, evaluates the result, and iterates until it isolates the specific segment or event responsible for the metric change.

## Opportunity Linked Thesis

**Thesis**: [Agent](/Theses/Agent)

## Opportunity Linked I C P

**Icp**: [SaaS Provider](/CompanyTypes/SaaS_Provider)

## Opportunity Market Sizing

_Illustrative — target and order-of-magnitude estimate figures, not an achieved track record (this Thing is concept-stage)._

**S A M**: ~$300M-500M mid-market to enterprise B2B SaaS providers
**S O M**: ~$10M-25M
**T A M**: ~30k-40k global SaaS and digital-native companies × ~$25k-30k/yr ≈ ~$750M-1.2B
**Growth Rate**: ~15-20%/yr, driven by the exponential growth of observability data and the increasing cost of technical labor
**Paid Comparable Spend**: ~$60k-120k/yr per company in fractional Site Reliability Engineering labor, data engineering time, and legacy observability alert add-ons

## Opportunity Incumbents

- [Datadog Watchdog](/Products/Datadog_Watchdog) — Tool
- [Sisu Decision Intelligence](/Products/Sisu_Decision_Intelligence) — Tool
- [PagerDuty Event Intelligence](/Products/PagerDuty_Event_Intelligence) — Tool
- [Custom Python Scripts](/Products/Custom_Python_Scripts) — DIY
- [Manual Excel Analysis](/Products/Manual_Excel_Analysis) — Spreadsheet
- [ThoughtSpot SpotIQ](/Products/ThoughtSpot_SpotIQ) — Tool

## Opportunity Win Conditions

**Kill Thresholds**:
- Agent requires manual SRE correction on >25% of generated summaries
- Human-in-the-loop escalation remains >80% after 60 days of deployment
- Time-to-first-value for accurate triage exceeds 14 days
- Inference costs for log processing exceed $500 per customer per month
**Leading Metrics**:
- Mean time to automated diagnostic summary
- Percentage of alerts suppressed or resolved without human escalation
- On-call engineer acceptance rate of agent-proposed root causes
- Daily active triage volume per connected data integration
**What Proves Right**: Engineering teams route P3 and P4 alerts directly to the Metric Triage Agent instead of an on-call human rotation. The agent resolves or accurately categorizes the root cause without human intervention in at least 40 percent of cases within the first 30 days of deployment. Customers expand their usage tiers to process greater event volumes because the agent consistently suppresses noisy downstream alerts.
**What Proves Wrong**: Site reliability engineers configure the agent but continue to manually review raw logs because they lack trust in the diagnostic summaries. The false-positive suppression rate remains below 10 percent, meaning the agent acts as another alert forwarder rather than a functional triage layer. Data integration takes longer than three weeks due to bespoke pipeline requirements, stalling adoption before the customer experiences the first automated triage.

## Opportunity Build Profile

**Hardest Part**: Generating syntactically correct and semantically valid exploratory SQL queries against a highly custom, undocumented customer data warehouse without human intervention. The agent must accurately navigate complex schemas to isolate root causes without hallucinating table relationships that do not exist.
**Min Viable Scope**: Restrict v1 to diagnosing downward spikes in conversion rates for e-commerce companies using Snowflake and dbt, outputting a slack alert with the most correlated dimensional slices. Completely exclude automated remediation, complex forecasting models, and non-warehouse integrations like Datadog or Zendesk.
**Cold Start Problem**: The agent lacks context on a specific company's idiosyncratic metric definitions and messy schema layouts out of the box. Break this by requiring the first design partners to integrate via an existing structured semantic layer like LookML or dbt metrics, bypassing raw table chaos entirely.
**Time To First Value**: 1 to 2 weeks (requires connecting read-only warehouse credentials, mapping the primary KPI, and waiting for the first natural anomaly to trigger an automated investigation)
**Data Moat Available**: true
**Technical Difficulty**: High

## Neighborhood

### Where the gap lives

- [Improvement Identification Cycle Time](/Metrics/Improvement_Identification_Cycle_Time) — latent gap · Metrics

### Incumbent in

- [PagerDuty AIOps](/Products/PagerDuty_AIOps) — incumbent in · Products
- [Bespoke Python Scripts](/Products/Bespoke_Python_Scripts) — incumbent in · Products
- [Manual Excel Analysis](/Products/Manual_Excel_Analysis) — incumbent in · Products
- [ThoughtSpot SpotIQ](/Products/ThoughtSpot_SpotIQ) — incumbent in · Products
- [Datadog Watchdog](/Products/Datadog_Watchdog) — incumbent in · Products
- [Sisu Decision Intelligence](/Products/Sisu_Decision_Intelligence) — incumbent in · Products

### Applies thesis

- [SaaS Provider](/CompanyTypes/SaaS_Provider) — applies thesis · CompanyTypes

### Embodies

- [Agent](/Theses/Agent) — embodies · Theses

### Similar Opportunities

- [Data Pipeline Repair](/Opportunities/Data_Pipeline_Repair) — similar · Opportunities
- [Usage Telemetry Agent](/Opportunities/Usage_Telemetry_Agent) — similar · Opportunities
- [Root Cause Investigator](/Opportunities/Root_Cause_Investigator) — similar · Opportunities
- [Root Cause Analyst](/Skills/Complex_Problem_Solving/Opportunities/Root_Cause_Analyst) — similar · Opportunities
- [Automated Fault Triage](/Opportunities/Automated_Fault_Triage) — similar · Opportunities
- [Flight Risk Intelligence](/Departments/Example_Four/Opportunities/Flight_Risk_Intelligence) — similar · Opportunities
- [Turnkey Account Expansion](/Opportunities/Turnkey_Account_Expansion) — similar · Opportunities
- [Root Cause Analyst](/Opportunities/Root_Cause_Analyst) — similar · Opportunities
- [Account Preservation Engine](/Opportunities/Account_Preservation_Engine) — similar · Opportunities
- [Autonomous SRE Responder](/Opportunities/Autonomous_SRE_Responder) — similar · Opportunities
- [Pre-Run Anomaly Detection](/Opportunities/Pre-Run_Anomaly_Detection) — similar · Opportunities
- [B2B Outreach Agent](/Opportunities/B2B_Outreach_Agent) — similar · Opportunities
- [Retention Cortex](/Opportunities/Retention_Cortex) — similar · Opportunities
- [AI Incident Triage](/Opportunities/AI_Incident_Triage) — similar · Opportunities
- [Pre-Churn Telemetry Analyst](/Opportunities/Pre-Churn_Telemetry_Analyst) — similar · Opportunities
- [Incident Triage Agent](/Opportunities/Incident_Triage_Agent) — similar · Opportunities
- [Algorithm Medic](/Opportunities/Algorithm_Medic) — similar · Opportunities
- [Outage Detection Automation](/Opportunities/Outage_Detection_Automation) — similar · Opportunities
- [Account Preservation Engine](/Occupations/Management_Occupations/Opportunities/Account_Preservation_Engine) — similar · Opportunities
- [Signal Node](/Opportunities/Signal_Node) — similar · Opportunities