Activity War ROOM

Activity War ROOM is EXAWin's core operational screen — a CRM command center where all sales activities (meetings, calls, emails, etc.) are recorded and Bayesian analysis results are viewed in real-time.

Every time an activity is recorded, the Bayesian engine automatically recalculates P(Win), enabling sales representatives to make data-driven decisions.

Location: Sidebar → Sales → Activity Board

Screen Layout

Activity War ROOM uses a 3-panel CRM layout.

Left Panel: Project List

Header Features

• Filter Dropdown — Filter projects by All / Running / Closed.
• ⚙ Settings Button — Configure initial load filters (Status filter + show only projects created in the last N months).
• Search Box — Keyword search by project name. Filters instantly as you type.

Project Cards

Each project is displayed as a card under the customer name. Card information includes:

Access Permission Display

• Only projects where you are designated as a member can be clicked to enter.
• Projects without permission are displayed in a disabled state with a 🔒 icon.

Closed Project Colors

Center Panel: Activity Input + Timeline

When a project is selected, the center panel activates. The activity input form is at the top and the activity timeline is below.

Project Header

• Project Name + Archetype badge (Aggressive / Conservative / Standard / Custom)
• Won/Lost Badge — Automatically displayed for closed projects
• + New Button — Initializes a new activity input form. Disabled for closed projects.
• Activity Count display
• 💬 New Badge — Displayed with animation when there are unread new comments

Activity Input Form Details

Row 1: Basic Information

⚠️ Stage Selection Notes:

• Past stages that have already been passed are grayed out and cannot be selected.
• Selecting a terminal stage (Won 🏆 / Lost 📌) transitions the project to closed status. No new activities can be recorded after this activity.
• Terminal stage activities do not require signal selection, and Bayesian updates are skipped.

Row 2: Attendees

Row 3: Signal Selection (Bayesian Key)

Signals are sales signals observed during the activity. This selection is the key input for Bayesian updates.

• Searchable multi-select — Search signals in the dropdown and multi-select via checkboxes.
• Selected signals appear as tags at the top. Click × on a tag to deselect.
• Each signal shows its direction (α/β) and Impact Score.

Positive (α) and negative (β) signals are separately aggregated to calculate the Compound Score:

Compound = MAX(scores) + Σ(remaining scores) × 0.25

Example: 3 positive signals selected with Impact Scores of [0.5, 0.3, 0.2] → Compound α = 0.5 + (0.3 + 0.2) × 0.25 = 0.625

💡 The strongest signal becomes the representative value, but the remaining signals contribute with a 25% synergy effect (Dampening = 0.25). This acknowledges the cumulative effect of multiple signals while preventing probability from unrealistically surging or plummeting from a single meeting.

Two badges displayed in real-time at the top:

• α — Positive signals' Compound Score (emerald)
• β — Negative signals' Compound Score (rose)

⚠️ For non-terminal stages, signals are required (red * indicator).

Row 4: Summary

Summarize the activity in one line (required). This text is displayed as the activity's title in the timeline.

Example: "Budget review meeting with CTO, technical demo conducted"

Row 5: Rich Text Editor (3 Tabs)

Record activity details across 3 tabs. Each tab uses a Quill.js-based rich text editor supporting bold/italic/lists/numbered lists/headings/links. Images can be copied and pasted into the body.

💡 Click the expand icon (⤢) at the top right to expand/collapse the editor height.

Row 6: File Attachments

• Click the 📎 Attachments button to open the attachment area.
• Click or drag files to upload.
• Supported formats: PDF, DOC, XLS, image files, etc. (max 10MB).
• Selected files are displayed as previews.

Save

Clicking the "Save activity" button:

1. The activity is saved in Draft status.
2. Signal data is preserved as a JSON snapshot.
3. For non-terminal stages, Bayesian update is automatically executed.
4. An audit log is recorded.

Activity Timeline

Saved activities are displayed in the timeline in reverse chronological order. Each activity is a card with a connecting line on the left providing visual flow.

Timeline Icons

Each activity type has a different circle icon color:

Activities with Bayesian data display a meeting sequence number (#1, #2, ...) inside the icon.

Activity Card Layout

Bayesian Badges (Right side of each activity card)

Activities with Bayesian updates display the following badges:

Status Management (Confirm & Lock / Unlock)

• Confirm & Lock button — Confirms a Draft activity. Editing and deletion are blocked after confirmation.
• Unlock button — Locked → Draft transition. Only admins or project Owners can unlock. Unlocking returns the activity to Draft status, making it editable/deletable again.
• Edit → Save automatically applies Confirm & Lock.

Activity View/Edit Process

Clicking an activity's title (summary) loads the activity data into the edit form. Behavior varies depending on project and activity status.

Active Projects

All past activity titles can be clicked for viewing and editing.

💡 Clicking a Locked activity loads data into the edit form, but fields are disabled. First click the Unlock button to transition to Draft, then re-click the title to edit.

Closed Projects

In closed projects, only the last closing activity can be edited; previous activities are read-only.

⚠️ If the closing activity is Locked, first Unlock → Draft transition, then re-click the title to edit.

Comments & Likes

Each activity card has built-in SNS-style social features:

• ❤️ Like — Reflected instantly via AJAX. Toggle on/off.
• 💬 Comment — Enter text and submit with Enter or the send button. AJAX-processed.
• ↩ Reply — Reply to each comment.
• Delete Comment — Only the author or admin can delete (× button appears on hover).

💡 Comment and like counts are aggregated and displayed on the left panel project cards as well.

Activity Edit Form

Clicking an activity's summary (title) in the timeline loads the activity data into the top input form.

Editable Fields

⚠️ Clicking Save Changes after editing triggers automatic Bayesian recalculation from the affected meeting sequence if signals were changed. Confirm & Lock is automatically applied upon save.

Right Panel: Customer/Analytics Info

📐 Displayed on XL (1280px) resolution and above.

Customer Section

Project Section

Bayesian Analytics Section

This area is the most critical differentiator of this screen. It shows Bayesian analysis results for the selected project in real-time.

Deal Temperature

Emoji and label change based on P(Win) value:

P(Win) — Posterior Probability

• Current P(Win) % displayed as a large number
• Change from previous (▲/▼ %p)
• Gradient progress bar

95% Credible Interval

Guidance messages based on evidence volume:

• n < 5: ⚠️ "Extremely insufficient evidence — prediction cannot yet be trusted"
• n < 20: 📊 "Evidence accumulating — accuracy improves as meeting data builds"
• n < 50: ✅ "Sufficient evidence — this prediction can be trusted"
• n ≥ 50: 🎯 "Abundant evidence — highly reliable prediction"

Silence Penalty

When there is no activity for a certain period (default 14 days) after the last meeting, a penalty automatically accumulates on β. This panel only appears when a penalty exists.

Unit Penalty = Weak Negation Impact × 0.3
Penalty Count = floor((elapsed days - gap days) / interval days) + 1
Total Penalty = Unit Penalty × Count → added to β

Display: [Count] times × [Unit Penalty] = β+[Total Penalty]
Example: 28 days silence, gap=14, interval=7 → (28-14)/7+1 = 3 times
    3 times × 0.036 = β+0.108

⚠️ The silence penalty is added to the previous β first before the α/β update proceeds when the next activity is recorded. Gaps between past meetings are also retroactively applied during recalculation.

💡 ⓘ Help Integration: Click the ⓘ icon next to each metric to open EXAWin help, with links to Docs for more detailed information.

Momentum P(Win)

A time-decay based P(Win) that gives higher weight to recent activities.

Momentum α = prior_α + Σ(each meeting's positive × SWV × e^(-λ × days_ago))
Momentum β = prior_β + Σ(each meeting's negative × SWV × e^(-λ × days_ago) + silence × e^(-λ × days_ago))
Momentum P(Win) = Momentum α / (Momentum α + Momentum β)

• Time decay coefficient: λ = 0.01 (older meetings have exponentially decreasing weight)
• Change is displayed when the difference from standard P(Win) is ±0.5%p or more.

α / β Current Values

Impedance

How much P(Win) dominates the current stage's Threshold (T) — a Sigmoid-based confidence metric.

Impedance = 1 / (1 + e^(-k × (P(Win) - T)))

• If P(Win) > T, impedance exceeds 50% → Decision transition point breached
• T (threshold) and k (slope) values are displayed together.
• Visualized as a violet/purple gradient gauge.

P(Win) Trend Chart

When there are 2+ Bayesian updates, an SVG time-series chart is displayed:

• Blue solid line — P(Win) trend per meeting
• Light blue shading — 95% CI range band
• Red dashed line — Threshold (T) baseline
• Data points — Each meeting sequence (#1, #2, ...)

Closed Projects

Projects where a Won or Lost stage has been recorded transition to closed status.

Activity Deletion

Only Draft status activities can be deleted. Click the 🗑 icon at the bottom right of the timeline card.

On deletion:

1. Processed as soft delete (deleted_at recorded).
2. If other activities remain in the same project, Bayesian is automatically recalculated (full recalculation from #1).

⚠️ Locked (confirmed) activities cannot be deleted. Unlock first before deleting.

Operations

Activity Recording Flow

1. Select a project from the left.
2. Click the "+ New" button at the center top.
3. Select Activity Type, Date & Time, Stage (required).
4. Search and multi-select signals (required except for terminal stages).
5. Enter Summary.
6. Record details in Meeting notes / Action items / Issues tabs.
7. Attach files if needed.
8. Click "Save activity".

Post-Save Auto Processing

Activity Save
├── Save as Draft status in DB
├── Signal → activity_signals relationship save
├── Signal → signal_data (JSON snapshot) preservation
├── If not a terminal stage:
│    └── BayesianUpdate.calculate_and_create!
│         ├── 1. Compound Score calculation (MAX + Σ(rest) × 0.25)
│         ├── 2. Silence penalty → added to β_prev
│         ├── 3. α_new = α_prev + SWV × Compound_positive
│         ├── 4. β_new = β_prev + SWV × Compound_negative
│         ├── 5. P(Win) = α_new / (α_new + β_new)
│         ├── 6. Impedance = Sigmoid(P(Win), T, k)
│         └── 7. 95% CI = Normal approximation (z=1.96)
├── Attachment file save
└── Audit Log recording

💡 SWV (Stage Weighted Value): The weight set for each stage, multiplied by the signal's Compound Score and reflected in α/β. Later stage signals (higher SWV) have greater impact than early stage signals (lower SWV).

Manual Bayesian Recalculation

Click the 🔄 Recalculate button in the right panel "Bayesian analytics" header to recalculate all Bayesian updates from #1. Use this after parameter changes.

Project Closing (Won/Lost)

1. Select 🏆 Won or 📌 Lost from the Stage dropdown.
2. Enter Summary and save.
3. Terminal stage: no signal selection needed, no Bayesian update.
4. Project transitions to closed status.

Permission System

Data Relationships

Active Project
├── Activity Record
│    ├── Type: Meeting / Video / Email / Phone / SNS
│    ├── Signals (N:N) → ActivitySignal → SalesSignal
│    │    └── Compound Score = MAX + Σ(rest) × 0.25
│    ├── Bayesian Update
│    │    ├── α_new = α_prev + SWV × Compound_positive
│    │    ├── β_new = β_prev + SWV × Compound_negative + Silence Penalty
│    │    ├── P(Win) = α / (α + β)
│    │    ├── Impedance = 1/(1+e^(-k(P-T)))
│    │    └── 95% CI (Normal approximation, z=1.96)
│    ├── Comments → Replies
│    ├── Likes
│    └── Attachments (ActivityAttachment)
├── Project Stages (ProjectStage)
│    └── SV, SWV, T, k parameters
├── Project Impacts (ProjectImpact)
│    └── custom_impact_score (f-coupling principle)
└── Members (ProjectMember)

Important Notes

1. Signal selection is the key to Bayesian analysis. Accurately select the signals observed during meetings. Repeatedly selecting only excessively positive or negative signals will reduce analysis accuracy.
2. Recording a terminal stage (Won/Lost) closes the project. If you accidentally recorded Won/Lost, edit the closing activity and change the stage to the immediately previous stage to reopen the project.
3. Multiple people can record activities for the same project simultaneously. Each recorder's name is displayed in the timeline.
4. Editing is not possible after Confirm & Lock. Only confirm when the content is certain. If locked by mistake, request an Unlock from the Admin or Owner.
5. Silence penalty accumulates automatically. P(Win) decreases if there is no activity for an extended period. Regular activity recording is recommended.
6. Bayesian is recalculated after deletion. Deleting a middle activity may change P(Win) for all subsequent meetings.
7. Projects without access permission are visible but not accessible. Request the project manager to add you as a member.