> ## Documentation Index > Fetch the complete documentation index at: https://docs.dacard.ai/llms.txt > Use this file to discover all available pages before exploring further. # Intelligence ROI > How we measure decision quality, track compound intelligence, and trace the receipts on operational ROI. Intelligence ROI traces the loop: agents surface findings, findings drive moves, moves shift dimension scores. The view shows how much your team is moving and how often the loop closes. # Intelligence ROI We go past measuring activity. We trace whether your product team is getting **smarter** over time by tracking decision quality, recommendation accuracy, and the compound intelligence flywheel. Hit **Intelligence > ROI** to access this view. ## What we measure ROI runs on the compound flywheel: agents surface findings, findings drive moves, moves shift dimension scores, dimension lifts compound into a higher composite. | Metric | Read | | ----------------------- | ----------------------------------------------------------------------------------- | | **Actions dispatched** | Total moves sent by agents (Linear issues, Slack messages, re-scores) in the period | | **Dimensions improved** | Count of dimensions that scored higher on re-score after a move was dispatched | | **Score delta** | Composite score move across all products in the period | | **Loop closure rate** | Percentage of dispatched moves with a verified score lift linked to them | | **Velocity** | Score points gained per week, averaged over the last 30 days | ## Decision intelligence score The **decision intelligence score** (0-100) is a longitudinal read that ties your team's launch decisions to actual outcomes. Where dimension scores measure current capability, the DI score measures whether acting on recommendations actually moves your operations. ### Three components | Component | Weight | Read | | ------------------------- | ------ | -------------------------------------------------------------------------------------------------------------------- | | **Outcome quality** | 40% | Weighted hit rate of recommendations. Larger lifts count more than small ones. | | **Decision velocity** | 30% | How fast your team acts on recommendations. Faster move equals higher score. | | **Learning acceleration** | 30% | Whether your recent hit rate is improving over your historical average. Positive trend means the system is learning. | ### Confidence levels | Cycles completed | Confidence | Read | | ---------------- | ---------- | ---------------------------------------------------------- | | \< 5 | Low | Early stage. Score is directional but needs more data. | | 5-9 | Medium | Developing. Patterns are emerging. | | 10+ | High | Reliable. Score reflects real decision-quality trajectory. | The DI score compounds: the more cycles your team completes (Score, Connect, Correlate, Act, Learn), the more reliable the read. ## How loop closure works A loop is **closed** when: 1. An agent dispatched a move targeting a specific dimension 2. A re-score happened after the move was dispatched 3. That dimension's score lifted on the re-score Loops that are still pending (move dispatched but no re-score yet) read as **open** in the flywheel view. To maximize loop closure, re-score products after your team addresses a recommendation. Even a quick URL re-score captures recent moves. ## Time ranges Use the period selector to read ROI over: * **Last 7 days** for recent agent activity * **Last 30 days** for the monthly view (aligned with billing and credit resets) * **Last 90 days** for the quarterly view (planning, investor updates) * **All time** for the full account history ## Knowledge graph Behind the scenes, we build a **decision graph** connecting signals, scores, decisions, and outcomes across cycles. The graph powers compound intelligence queries: * **What caused this score move?** Trace from an outcome back through the recommendation, the agent that produced it, and the signal that informed it. * **Which recommendations worked?** Follow paths from recommendations through to measured outcomes. * **How did human calls affect outcomes?** Every approval, rejection, and Teach DAC reaction lands as a node in the graph. ### Node types | Node | What it represents | | ------------------ | ------------------------------------------------------------------ | | **Signal** | An adapter event (PR merged, issue completed, deploy shipped) | | **Score snapshot** | A point-in-time dimension or composite score | | **Recommendation** | A coaching move surfaced by the intelligence engine | | **Outcome** | A measured score move after a recommendation was acted on | | **Human decision** | An approval, rejection, Teach DAC reaction, or outcome attribution | The graph grows with every cycle. Over time, it becomes a queryable history of your team's decision patterns and their consequences. ## Peer benchmarking When enough teams have scored, we compute **anonymized peer benchmarks** by segment (company stage, team size, industry). Your coaching observations carry benchmark comparisons: > "Your score of 2.1 reads below the peer median of 3.2 (25th percentile across 150 teams). Closing this gap could move your composite materially." Benchmarks compute nightly from the aggregate scoring database. Your individual scores are never shared or identifiable. ## Product breakdown The ROI view breaks down by product, showing: * Which products have the most active flywheel loops * Which products have the highest score velocity * Which products have open (unverified) loops that need a re-score ## Reading the charts **Score velocity chart:** weekly composite deltas per product. Upward trajectory means your team is moving faster than the platform is raising the bar. **Actions dispatched by type:** breakdown of Linear issues, Slack messages, re-scores, and coaching recommendations. Skew toward re-scores suggests agents are catching regressions; skew toward Linear issues suggests proactive movement. **Dimension heat map:** which of the 27 dimensions are lifting most consistently across all products. Strong dimensions (4/4 across multiple products) signal systemic organizational strength. ## Measuring compound ROI The compound flywheel on the Agent Studio dashboard traces real operational impact. | Metric | Read | | ---------------------- | -------------------------------------------------------- | | **Completed cycles** | Full Score, Connect, Correlate, Act, Learn revolutions | | **Accuracy trend** | Recommendation hit rate per period (monthly) | | **Intelligence depth** | How much the system's recommendation precision has moved | | **Score velocity** | Composite move per month | ### The flywheel Each revolution makes the next sharper. 1. **Score** assess current maturity 2. **Connect** adapter signals ground-truth the assessment 3. **Correlate** agents surface cross-function patterns 4. **Act** recommendations become moves (Linear tickets, coaching) 5. **Learn** outcomes are measured and attributed Teams with 10+ completed cycles see materially better recommendation accuracy than teams in their first cycle. That's the compound moat: a competitor can copy the framework, but they can't copy 12 months of your team's decision intelligence data. ## Use in investor updates The ROI view is built to be screenshot-ready for investor and board reporting. The headline metrics (score delta, loop closure rate, velocity) quantify your team's continuous improvement cadence in a format that lands beyond subjective self-assessment. ## Related Configure agents and manage the compound flywheel. Financial attribution and evidence citations in coaching. Read your position relative to peers. Signal-score blending and the 27-dimension methodology.