Project Master

Project Master is the core screen for creating individual sales deals and adjusting Bayesian inference parameters to match the unique characteristics of each deal.

When a project is created, the company's master parameters (Stage SV/T/k, Signal Impact) are automatically cloned, and can then be independently tuned per project.

Location: Sidebar → Sys Module Master → Project

Screen Layout

Project Master uses a 2-panel layout.

Left Panel: Project List

Header Features

• All Sheets Button — Exports Project + Stage + Signal + Member as a multi-sheet Excel file.
• Search Box — Keyword search by project name, code, or customer name.
• Archetype Filter — Dropdown for Aggressive / Conservative / Standard or all.

Table Columns

Inline Filters

• Name — Search by project name
• Customer — Search by customer name
• Active — All / Y / N dropdown

Right Panel: 5-Tab Structure

When a project is selected, 5 tabs are displayed (only the General tab is shown for new entries).

General Tab Details

Deal Information

Bayesian Prior (Alpha / Beta)

This area is read-only and displays the company's default α/β values.

⚠️ The default α and β values are managed in Company Settings. They cannot be changed per project and are applied uniformly across all projects.

Silence Penalty

When there is no activity (meetings, calls, etc.) for a certain period, a penalty is automatically accumulated on β.

When elapsed days > Contact Gap:
  Penalty Count = ((elapsed days - Gap) / Interval) + 1
  Total Penalty = Count × Unit Penalty
  Unit Penalty = 30% of Weak Negation Impact

💡 Example: Gap = 14 days, Interval = 7 days, and 28 days since last activity:
Penalty Count = ((28 - 14) / 7) + 1 = 3 times
→ β receives Weak Negation Impact × 30% × 3 additions

💡 Feel free to adjust Contact Gap and Penalty Interval to match the characteristics of your deal. Most deals have relatively short cycles, so setting Gap to 7–10 days or shorter can help you effectively utilize Bayesian logic even for small projects. For long-term projects, set it longer as appropriate. These two values do not directly affect the Bayesian scale, so you can change them without concern.

Stages Tab Details ⚙️

⚠️ Project-Specific Physics: Values adjusted here apply only to this project. They do not affect company master values.

When a project is created, the company's active stages are automatically cloned. The following 3 parameters can be independently tuned per stage.

Terminal stages (Won/Lost) do not require Bayesian tuning and are not shown in this tab.

🚀 Auto Parameter Learning: Auto-Tuner

EXAWin has a built-in Bayesian Auto-Tuner. When a sufficient number of completed projects (Won/Lost) accumulate, the system analyzes past real-world data and automatically calculates and suggests optimal T, k, and Impact parameters.

Like a sports coach analyzing last season's game records to adjust next season's tactics — the data tells you which settings best distinguished Won deals from Lost deals.

💡 Therefore, when manually adjusting the parameters below, make fine adjustments near the baseline values. Setting values that significantly deviate from the scale may conflict with the optimal values learned by the Auto-Tuner, reducing the overall system's discrimination accuracy. Baseline values are a safe starting point until the Auto-Tuner begins learning.

Parameter Descriptions

SWV (Stage Weighted Value) — Auto-Calculated

When SV is entered, SWV is automatically calculated in real-time:

SWV = 1 + ln(SV)

SWV is the weight used in actual Bayesian updates. When SV = 1, SWV = 1.0 (neutral); SV > 1 means SWV > 1 (amplification); 0 < SV < 1 means SWV < 1 (attenuation).

🔴 Default Parameters by Stage (Theory Reference)

Below are the default parameters per stage derived through simulation. Use these as baseline when tuning per project.

⚠️ The above values are baseline values derived through simulation in the Theory: Parameter Calibration and Auto-Optimization. Adjustments that significantly deviate from the range without justification may distort posterior probability.

🔴 T (Threshold) Adjustment Principles

T defines the threshold for "what minimum % confidence is needed for this deal to pass this stage?"

T is a function of the company and project context. For example, one company might decide "we proceed only with 70%+ confidence," while a startup might judge "50% is enough." T is thus both a technical value and a reflection of management philosophy.

• Recommended range: 0.50 – 0.70 — Realistic and aligned with human intuition for most sales situations
• Default: Increases as stages progress (Discovery 0.35 → Negotiation 0.55)
• Caution: T > 0.70 is excessively conservative and risks dropping promising deals

🔴 k (Velocity) Adjustment Principles

k determines the sensitivity of discrimination near T. Higher k means a sharper "pass/fail" boundary.

• Early stages (Discovery–Qualification): k = 5–7 — Be generous during exploration
• Late stages (Proposal–Negotiation): k = 11–12 — No ambiguity at cost investment stages
• Caution: k = 12 is the theoretical maximum. k > 12 makes the sigmoid essentially a step function, turning discrimination into binary classification. The Auto-Tuner also respects this upper limit.

Impacts Tab Details ⚙️

⚠️ Project-Specific Impact: Adjusts the influence of specific signals for this project.

When a project is created, all company Impact Types are automatically cloned with their default Impact Scores.

What is Impact Score?

The strength of evidence that each signal contributes to Bayesian updates.
Signals selected during activity recording are aggregated using the Compound Score method:

Activity Compound Score = MAX(signal Impact) + Σ(remaining signals' Impact × Dampening)

The strongest signal is reflected at 100%, while remaining signals are reflected at the Dampening ratio (default 25%). This is an evidence dampening mechanism that prevents probability from unrealistically surging even when many signals are selected in one meeting.

Impact Type Classification

🔴 Impact Value Standards (f-Coupling Principle)

Impact values are defined as fractions of Prior strength (S = α + β) following the f-Coupling principle:

Impact = f × S    (S = α + β, default S = 10)

Below are the default Impact values from the theory documentation:

⚠️ The above values are baseline values derived from the Theory: f-Coupling Principle. Changing these values without justification may distort posterior probability.

EPR Guardrails

EPR (Evidence-Prior Ratio) limits the maximum influence of a single signal:

EPR = SWV_max × Impact / S

Regular signals:  EPR ≤ 0.5  (50% or less of Prior)
Game Changer:     EPR ≤ 2.0  (200% or less of Prior)

⚠️ If EPR exceeds the allowed range, saving is rejected at the master level. This is a code-level safeguard to prevent a single signal from excessively overwhelming the Prior.

Members Tab Details

Specify which users can access the project.

Auto-Inclusion Rules

• Admin and Super User roles are always auto-included — cannot be unchecked.
• Regular users (Manager, User) can be selected/deselected via checkboxes.
• Select All checkbox for bulk select/deselect.

What Access Permission Means

Only designated members can:

• View the project in the Activity Board.
• Record activities for the project.
• View Bayesian analysis results.

Revenue Tab Details

Manage the expected revenue for the sales deal.

Operations

Creating a Project

1. Confirm the right form is in the "New Sales Project" state.
2. Enter Project Name and Customer (required).
3. Select/enter Main Contact, Owner, and notes.
4. Adjust Silence Penalty values as needed.
5. Click the Save & Tune Bayesian Parameters button.

🔄 Automatic master parameter cloning executes upon save:

• All active company stages → ProjectStage (including SV/SWV/T/k)
• All company Impact Types → ProjectImpact (including Impact Score)
• Admin/Super User → Auto member registration

Parameter Tuning

1. Click a project in the left list.
2. Adjust SV, T, k for each stage in the Stages tab.
3. Adjust Impact Score for each signal in the Impacts tab.
4. Click the Save & Tune Bayesian Parameters button.

Deleting a Project

Clicking Delete checks the following conditions:

⚠️ Delete Restriction: Deletion is not possible if even 1 activity record exists.
To close a deal, use Active OFF to deactivate, or record a terminal stage (Won/Lost) in the Activity Board.

Form Reset

Clicking the ↺ button at the top right:

• Resets the form to the "New Sales Project" state.
• Hides tabs and shows only the General tab.
• Returns all fields to default values.

Project Code Auto-Generation

Permission System

Parameter Inheritance Hierarchy (Overall Architecture)

Company
├── base_alpha / base_beta ← Applied to all projects
├── SalesStage (master): SV, SWV, T, k
│    └── ProjectStage (cloned per project): SV, T, k individually tunable
├── ImpactType (master): impact_base_val
│    └── ProjectImpact (cloned per project): custom_impact_score individually tunable
└── Project
     ├── silence_gap_days / silence_interval_days
     ├── Activities → BayesianUpdate (cumulative α/β calculation)
     └── Members (access permissions)

Important Notes

1. Project-specific parameter tuning only affects that project. To change master values, use Stage Master or Signal Master.
2. SV must be greater than 0. A value of 0 or less causes errors in SWV calculation.
3. T cannot exceed the 0.01 – 0.99 range. At 0, confidence is always 1; at 1, confidence is always 0, making it meaningless.
4. Impact Score is automatically limited by the EPR principle. Refer to the range allowed by the master when setting.
5. Clean up activity records before deleting a project. Projects with activities cannot be deleted; deactivation is recommended.
6. Admin/Super User are automatic members. Their access cannot be intentionally removed.
7. Adjust silence penalty settings to match deal characteristics. Long deals should have longer Gaps (30+ days), short deals shorter (7 days).