TargetBoard Documentation

TargetBoard employs five levels of filtering. Filters are cumulative, meaning if multiple levels (e.g., account, folder, board, and card) have filters applied, the resulting data is the intersection of all those filters.

Filter Levels (from broadest to most specific):

1. Account Level: Global data exclusion.
2. Metric Level: Reusable metric definitions.
3. Folder Level: Applies to all boards and cards within a folder.
4. Board Level: Applies to all cards on a specific board.
5. Card Level: Applies to a single KPI card.

TargetBoard offers two types of boards for organizing metrics and KPIs:

• Managed Boards: Curated and maintained by the TargetBoard team based on industry best practices (e.g., AI Adoption, Productivity, Sales Performance). They are not directly editable but provide a quick start. They are ideal for users who want a low-maintenance setup or standard industry breakdowns.
• Custom Boards: Built from scratch to meet specific requirements. Users can also create a Custom Board by duplicating a Managed Board and then editing the duplicate to serve as a template.

Metrics are measures used to track processes or goals and are located in the Catalog. They are categorized into three types:

1. Core Metrics: Top-level metrics determined by connected integrations (e.g., Jira, GitHub, Zendesk) and the specific use cases defined in the account agreement.
2. Custom Metrics: Top-level metrics built to specific organizational requirements or non-industry standard definitions (e.g., specific DORA metrics). These are created via a Customer Success Manager.
3. Filtered Metrics: Subsets of Core Metrics created by applying specific filters (e.g., 'Backend Team Cycle Time'). These appear indented under their parent Core Metric in the Catalog and are used to avoid repeating the same filter configurations across multiple boards.

TargetBoard Agents are AI-powered operational analysts that monitor company data to detect risks, bottlenecks, trends, and opportunities. They replace the need for manual dashboard reviews by proactively surfacing actionable insights.

Agents focus on specific operational domains such as:

• Software quality
• Velocity (e.g., monitoring cycle time, throughput, blocked work, PR review time)
• AI Impact
• Predictability
• Sustainability

Users can define the data scope for an Agent (e.g., specific teams or projects) and provide feedback to refine the Agent's focus over time.

Signals are the specific insights generated by Agents. A Signal is created when an Agent detects a meaningful change, risk, anomaly, or opportunity. Each Signal provides:

• What happened and why it matters
• Supporting metrics
• The originating Agent
• Severity level
• Creation timestamp

The AI Maturity Scoring Model measures 'how well' developers use AI by converting weekly behavioral signals into a 0–100 maturity score.

Key mechanics include:

• Data Collection: Every developer is tracked weekly from their first AI tool use onward. Silent weeks are treated as signals, not missing data.
• Calculation: Fourteen behavioral signals are measured, normalized to a 0–1 scale, averaged into three group sub-scores, and combined into a weighted composite score.
• Smoothing: The final maturity level is based on a four-week rolling average to prevent erratic level jumps, while the weekly score reacts immediately.
• Scoring Methods:
  • Absolute: Scored against a configured target and capped at 100%.
  • Relative: Ranked against all other active users in the same week to measure relative maturity rather than raw volume.

TargetBoard attributes AI-assisted engineering work by correlating two categories of operational metadata rather than using invasive monitoring. This allows organizations to measure the business impact of AI tools (e.g., effect on PR activity, cycle time, and throughput) without accessing proprietary source code, developer prompts, or local workstation activity.

TargetBoard's attribution model specifically avoids:

• Installing agents on developer machines or monitoring keystrokes.
• Reading source code or accessing proprietary IP.
• Reading prompts, AI conversations, or capturing IDE activity.
• Collecting files or attachments from repositories.