> ## Documentation Index > Fetch the complete documentation index at: https://docs.charle.agency/llms.txt > Use this file to discover all available pages before exploring further. # Client Intelligence > Understanding client deliverability scores and how they improve scheduling accuracy CharleOS automatically learns from historical task completion data to improve scheduling accuracy for each client. The Client Intelligence system tracks how actual work times compare to quoted estimates, creating per-client scores that help PMs schedule more accurately. ## What is Client Intelligence? Client Intelligence provides data-driven insights into how each client's work typically progresses: * **Development Score**: How development tasks perform against estimates * **Design Score**: How design tasks perform against estimates * **Help Desk Score**: How help desk tickets perform against estimates * **QA Score**: How many QA cycles a client typically requires These scores help the system suggest more accurate time estimates based on each client's historical patterns. ## Understanding Scores ### Score Values A score of **1.00** means the client's work matches estimates perfectly: | Score | Meaning | Example | | -------- | ---------- | ------------------------------------------ | | **1.00** | On target | Work matches quoted time exactly | | **1.15** | 15% slower | A 4-hour quote typically takes \~4.6 hours | | **0.90** | 10% faster | A 4-hour quote typically takes \~3.6 hours | ### Confidence Levels Scores become more reliable with more data: | Confidence | Sample Size | Reliability | | ---------- | ------------------ | ------------------------------------------------- | | **Low** | Fewer than 5 tasks | Use with caution - fall back to maximum estimates | | **Medium** | 5-15 tasks | Reasonable confidence - can use for scheduling | | **High** | 15+ tasks | Reliable data - use for accurate scheduling | The system uses a **weighted average** with more recent tasks having greater influence. This means scores naturally adapt as client patterns change over time. ## How Scores Affect Scheduling ### Two-Layer Planning Model CharleOS uses a two-layer planning model: **Uses: Quoted Averages** When PMs plan which tasks fit in a month's budget, they use average estimates from t-shirt sizes. This determines what the client "pays for". **Uses: Score-Adjusted Estimates** When scheduling actual work, the system uses score-adjusted estimates or maximums. This ensures realistic capacity planning. ### Example A Medium task (3-6 hours, average 4.5): * **Budget**: PM allocates 4.5 hours from client's monthly hours * **Scheduling**: For a client with score 1.2, system suggests 5.4 hours for developer scheduling This protects both the client (they pay the quoted average) and the team (they have realistic time allocations). ## Viewing Client Intelligence Client Intelligence is displayed on the **Intelligence** tab of each client's detail page. ### What You'll See * **Development Score Card**: Score value with confidence badge and sample size * **Design Score Card**: Score value with confidence badge and sample size * **QA Score Card**: Average QA cycles with first-pass rate and sample size * **Score History**: Chart showing score trends over time with a filterable time range (3M/6M/12M/All), plus an event log of score changes * **Outlier Tasks**: Table of tasks excluded from calculations, with the ability for admins/managers to include or exclude them ### Interpreting the Data Score below 1.0 (e.g., 0.85) **What it means:** * This client's work typically completes faster than estimated * Could indicate efficient processes or clear requirements * Good candidate for tighter scheduling **Action:** * Can schedule closer to average estimates * May indicate estimation is too conservative for this client Score near 1.0 (0.95 - 1.05) **What it means:** * Estimates are accurate for this client * Good estimation calibration **Action:** * Continue with standard estimates * Monitor for changes Score above 1.0 (e.g., 1.25) **What it means:** * This client's work typically takes longer than estimated * Could indicate complex requirements, feedback cycles, or scope changes **Action:** * Schedule using larger buffers * Consider if estimates need adjustment for this client type * Review what's causing overruns ## Outlier Tasks Tasks that deviate significantly (more than 130%) from quoted estimates are flagged as **outliers**. These are excluded from score calculations to prevent anomalies from skewing the data. ### Why Tasks Become Outliers Common reasons for outlier status: * Unexpected scope expansion * External dependencies or blockers * Initial estimate was significantly wrong * Client-side delays ### Managing Outliers Outliers are displayed in the Client Intelligence section. Admins can: * Review why the task was flagged * Include/exclude outliers from calculations if needed * Use the data to improve future estimating ## Help Desk Scoring Help desk tickets have their own scoring system, separate from task scores: * Based on estimated time vs actual time logged * Helps AI suggest more accurate time estimates for new tickets * Improves PM scheduling efficiency When a help desk ticket is created, Alan automatically suggests a time estimate based on the ticket content and client history. This suggestion appears in the ticket details for the PM. ## QA Scoring QA Scoring tracks how many QA cycles a client's work typically requires: * **First-Pass Rate**: Percentage of tasks that pass QA on first review * **Average Cycles**: Mean number of review cycles needed * **QA Score**: Multiplier for adjusting QA time estimates ### How It Affects Scheduling 1. **T-Shirt Suggestions**: High QA scores may trigger recommendations to size up 2. **Subtask Estimates**: QA subtasks get adjusted time based on QA score 3. **AI Context**: QA metrics are included in AI suggestion reasoning A QA score of 1.8 means the client typically needs \~80% more QA time than average. The system adjusts subtask suggestions accordingly. For detailed information, see the [QA Scoring](/user-guide/clients/qa-scoring) documentation. ## Related Documentation Detailed guide to QA scoring and its effects How scores fit into the overall billing system How score-adjusted estimates appear in scheduling Understanding task size estimates Help desk ticket workflow and AI suggestions