Inventory Copilot — Hackathon Deep Dive

Pitch: Turn MBB's items page from a dumb list into a smart inventory advisor that predicts stockouts, surfaces dead capital, and closes the loop with one-click restocking.

---

The Problem

Every Indian SMB owner has had this happen:

A customer walks in and asks for their usual. You go to the shelf — it's empty.
You lost the sale. The customer is annoyed. You didn't even see it coming.

The current MBB experience:

• You manually set a "minimum quantity" per item
• MBB shows "16 items below minimum stock"
• That's it. You still don't know which items to buy how much of and when

What's broken:

1. The threshold is static and manual — owners rarely set it correctly
2. It tells you what IS, not what WILL BE
3. There's no action — you see the alert, then have to figure out what to do next

The real question an owner needs answered:

"Which items will I run out of, when, how much should I order, and from whom?"

---

The Solution — Inventory Copilot

Three things, in sequence:

1. Predict

Compute sales velocity from actual transaction history → predict days to stockout per item.
No manual setup. Works on day 1 with existing data.

2. Surface

Show the prediction as urgency tiers on the items page — critical, at-risk, dead stock.
One number: Inventory Health Score.

3. Act

One-click restock: pre-fills a PO with last supplier, last price, suggested quantity.
Closes the loop from alert → order in under 60 seconds.

---

Why This Wins

Dimension	Detail
Global proof	Inventory Planner acquired by Cin7 (~$200M). Charges $99–299/month just for this.
India gap	Zero Indian billing app (Vyapar, OkCredit, Khatabook) has demand forecasting.
Data ready	get_item_timeline MCP exists. 1.1B item rows + 491M voucher rows in Snowflake.
No setup needed	Works from existing sales data. No onboarding.
Instant value	Owner opens items page → immediately sees "3 items run out this week"
Monetization	₹499/month easily justifiable. Stockout = lost sale = real money.
Founder criteria	Instant 10x value ✅ Show-off/FOMO ✅ Psychological safety ✅ No setup ✅

---

What Exists Today vs After

Items Page — Before

┌─────────────────────────────────────────────┐
│  Stock Value: ₹4.2L  │  Low Stock: 16  │  Expiring: 4  │
└─────────────────────────────────────────────┘

┌──────────────┬────────┬──────────────┬───────────────┐
│ Item Name    │ Stock  │ Selling Price │ Purchase Price │
├──────────────┼────────┼──────────────┼───────────────┤
│ Atta 10kg    │   8    │    ₹340      │    ₹280       │
│ Dal Moong    │   5    │    ₹120      │    ₹95        │
│ Basmati Rice │  12    │    ₹210      │    ₹165       │
│ Samsung 65"  │   2    │  ₹68,000     │  ₹55,000      │
└──────────────┴────────┴──────────────┴───────────────┘

What the owner is thinking: "Which of these 200 items do I need to worry about right now?"
What MBB tells them: Nothing useful.

---

Items Page — After

┌────────────────────────────────────────────────────────────────────┐
│  🔴 3 Critical (< 3 days)  │  🟡 7 At Risk (3–7 days)  │  💤 18 Dead Stock (₹32K)  │
│  Inventory Health Score: 67 / 100                                  │
└────────────────────────────────────────────────────────────────────┘

💡 AI Summary: "3 items run out this week — together ₹18,400 in blocked sales.
   Atta 10kg is your fastest mover at 4.2 units/day. 18 items (₹32K) haven't
   moved in 60+ days — consider a clearance push."

                              [Restock All Critical ↗]

┌──────────────┬────────┬──────────────┬──────────┬───────────┬──────────┐
│ Item Name    │ Stock  │ Velocity/day │ Days Left │ Margin %  │ Action   │
├──────────────┼────────┼──────────────┼──────────┼───────────┼──────────┤
│ Atta 10kg    │   8    │   4.2 /day   │ 🔴 1.9d  │   17.6%   │ [Restock]│
│ Dal Moong    │   5    │   1.8 /day   │ 🔴 2.7d  │   20.8%   │ [Restock]│
│ Basmati Rice │  12    │   1.6 /day   │ 🟡 7.5d  │   21.4%   │ [Restock]│
│ Samsung 65"  │   2    │   0.0 /day   │ ⚫ Dead  │   19.1%   │ [Clear]  │
└──────────────┴────────┴──────────────┴──────────┴───────────┴──────────┘

What the owner knows immediately: Exactly which 3 items to restock today.

---

Demo Script (60 seconds, jaw-drop moment)

1. Open Items page → intelligence banner loads with "3 Critical" in red
2. Click "3 Critical" → filtered table shows only the at-risk items
3. Click "Restock All Critical" → Restock Assistant modal opens
4. Modal shows: pre-filled PO — Atta 10kg, 50 units, Agarwal Traders, ₹28/kg
5. Confirm → PO created / WhatsApp message to supplier drafted
6. Done. Alert to order, in under 60 seconds.

Before this feature: 10-minute manual process of checking each item, calling supplier.
After: 60-second end-to-end flow.

---

Features — Full List

Layer 1 — Core Intelligence

Sales Velocity

• avg_daily_qty_sold = total units sold in last 30 days ÷ 30
• Zero-sale days are included in the denominator (honest velocity)
• Displayed as: 4.2 units/day or 0.3 units/day

Days to Stockout Prediction

• days_to_stockout = current_stock ÷ avg_daily_velocity
• Urgency tiers:
  • 🔴 Critical — < 3 days
  • 🟡 At Risk — 3 to 7 days
  • 🟢 Safe — > 7 days
  • ⚫ Dead Stock — velocity = 0, item not selling

Dead Stock Identification

• Items with zero sales in last 30 / 60 / 90 days (configurable)
• Per item: days since last sale, current stock, capital locked (qty × purchase_price)
• Aggregate: "₹41,200 locked in 23 unsold items"

Margin per Item

• margin % = (selling_price − purchase_price) ÷ selling_price × 100
• Both prices already exist in item schema — zero backend change
• Alert threshold: items with margin < 10%

---

Layer 2 — Action Loop

One-Click Restock

• "Restock" button on each at-risk item row
• Pre-fills PO with:
  • Last supplier (from purchase invoice history)
  • Last purchase price
  • Suggested quantity: avg_daily_velocity × 15 days − current_stock
• User edits if needed → confirms → PO created

Batch Restock

• "Restock All Critical" — creates POs for all 🔴 items in one action
• Review screen: see all items together before confirming
• Estimated total restock cost shown upfront

Suggested Reorder Quantity

• Formula: (avg_daily_velocity × target_cover_days) − current_stock
• Default target: 15 days of cover
• User can change target days in settings

Supplier Memory

• Pulls last 3 purchase transactions for the item
• Shows: supplier name, price, date
• Pre-selects the most recent for the PO draft

---

Layer 3 — Business Intelligence

Inventory Health Score (0–100)

Single composite score for the business's overall inventory health.

Component	Weight	What it measures
Stockout Risk	40%	How many items are in critical/at-risk zone
Dead Stock Ratio	30%	Locked capital as % of total stock value
Margin Health	20%	Average margin across all items
Turnover Rate	10%	How fast inventory converts to sales

Displayed as: 67 / 100 with a color band (red / amber / green).

Purchase Price Trend

• Track last 5 purchase prices per item over time
• Alert: "Cement 50kg price up 12.3% in 3 months"
• Margin erosion alert: "Cost went up but selling price unchanged → margin: 22% → 9%"

Top Sellers / Slow Movers Tabs

• Top 10 items by: revenue, quantity sold, margin
• Bottom 10 items: lowest velocity, lowest margin, most overstock
• Simple sort — no AI required

Overstock Alert

• Items where current stock > 90 days of sales velocity
• "You have 180 days of Ariel 1kg at current pace. Slow down purchases."
• Capital efficiency angle: "₹8,400 tied up unnecessarily"

---

Layer 4 — AI Narrative

Smart Inventory Summary Card

Uses the existing /web/genai/llm/json endpoint — no new model, no new service.

Input (built from computed metrics):

{
  "critical_items": [
    { "name": "Atta 10kg", "days_left": 1.9, "velocity": 4.2 },
    { "name": "Dal Moong", "days_left": 2.7, "velocity": 1.8 }
  ],
  "dead_stock_count": 18,
  "dead_stock_value": 32400,
  "top_seller": { "name": "Atta 10kg", "velocity": 4.2 },
  "health_score": 67
}

Output (LLM-generated, 3 bullet points):

"3 items run out this week — ₹18,400 in blocked sales if you don't restock."
"₹32,000 sitting in 18 items that haven't sold in 60+ days — consider a clearance push."
"Your best mover (Atta 10kg) accounts for 34% of your daily inventory movement."

This is the only AI in the feature. Everything else is pure math.

---

Technical Architecture

Data Sources (all already exist)

Data needed	Source	Status
Current stock per item	/web/items list API	✅ exists
Sales history per item	get_item_timeline MCP	✅ exists
Purchase price per item	Item schema purchaseInfo.price	✅ exists
Selling price per item	Item schema salesInfo.price	✅ exists
Item stats (counts)	/web/items/stats	✅ exists
LLM narrative	/web/genai/llm/json	✅ exists
Snowflake query path	flo-analytics-llm	✅ exists

What Needs to Be Built

Backend: /web/items/intelligence (new endpoint)

Input: company_id, days=30 (lookback window)

Processing: One query joining stock levels + sales history per item:

SELECT
  ii.id,
  ii.name,
  ii.quantity AS current_stock,
  SUM(CASE WHEN vi.itemable_type = 'Invoice' AND v.voucher_type = 1
       THEN ABS(vi.quantity) ELSE 0 END) / :days AS avg_daily_velocity,
  MAX(CASE WHEN vi.itemable_type = 'Invoice' THEN v.voucher_date END) AS last_sold_date
FROM inventory_items ii
LEFT JOIN items vi ON vi.itemable_id = ii.id  -- item timeline join
LEFT JOIN vouchers v ON v.voucherable_id = vi.itemable_id
  AND v.voucher_date >= CURRENT_DATE - :days
WHERE ii.company_id = :company_id
GROUP BY ii.id, ii.name, ii.quantity

Output:

{
  "items": [
    {
      "id": "abc123",
      "name": "Atta 10kg",
      "current_stock": 8,
      "avg_daily_velocity": 4.2,
      "days_to_stockout": 1.9,
      "last_sold_date": "2026-05-04",
      "urgency": "critical",
      "dead_stock": false
    }
  ],
  "summary": {
    "critical_count": 3,
    "at_risk_count": 7,
    "dead_stock_count": 18,
    "dead_stock_value": 32400,
    "health_score": 67
  }
}

Cache: 1-hour TTL per company.

---

Frontend: New Components

1. InventoryIntelligenceCardsComponent

• Replaces / extends the current item-index-data-cards
• Shows: 🔴 Critical, 🟡 At Risk, 💤 Dead Stock (₹value), Health Score
• Each card is clickable → filters the table to that tier

2. New columns in ItemIndexTableConfig

{ key: 'velocity',    label: 'Sales/day' },
{ key: 'daysLeft',   label: 'Days Left',  sortable: true },
{ key: 'urgency',    label: '' }   // color indicator column

3. InventoryAISummaryCardComponent

• Calls /web/genai/llm/json with inventory metrics
• Streams the 3-bullet summary
• Dismissible, cached per session

4. RestockAssistantModalComponent

• Opens when user clicks "Restock" on a critical item
• Fetches last purchase history for that item
• Shows: supplier, last price, suggested quantity, estimated cost
• CTA: "Create Purchase Order" or "Send WhatsApp to Supplier"

5. BatchRestockComponent

• "Restock All Critical" flow
• Summary review: N items, total estimated cost ₹X
• Creates all POs in one API call

---

Minimum Viable Demo (hackathon day)

If short on time, this is the absolute minimum to demo and still win:

1. Intelligence cards — 4 new stat cards (critical, at-risk, dead stock, health score) calling the new /web/items/intelligence endpoint
2. "Days Left" column — add velocity + days_to_stockout columns to existing items table
3. Filter by urgency — clicking a card filters the table (same pattern as existing low_stock filter)
4. AI summary card — 3-line LLM summary using existing genai endpoint

That alone is a fundamentally different product. Restock assistant is v2.

---

Pitch for Judges

The problem in one sentence: Indian SMB owners lose lakhs every year to stockouts they never saw coming and dead stock they never notice.

The solution in one sentence: Inventory Copilot turns MBB's item list into a predictive advisor — it tells you what will run out, when, and exactly what to order.

Why it's not just a feature, it's a product: Inventory Planner, a SaaS that does only this, was acquired for ~$200M. Every serious inventory management tool globally charges $100–300/month for demand forecasting. India has zero equivalent in the SMB billing space. MBB can own this category.

Monetization: ₹499/month add-on. The first stockout it prevents pays for 6 months. Owners will pay without hesitation.

Why only MBB can build this: We have 10M+ businesses' transaction data. Our velocity predictions get better with every user. No competitor starting today has this training set.

---

Competitive Comparison

Feature	MBB Today	MBB with Inventory Copilot	Vyapar	QuickBooks India	Inventory Planner
Low stock alert	Manual threshold	✅ Velocity-based prediction	Manual threshold	Manual threshold	✅ AI forecast
Days to stockout	❌	✅ Per item	❌	❌	✅
Dead stock ID	❌	✅ With ₹value	❌	❌	✅
Reorder quantity	❌	✅ Auto-suggested	❌	Basic	✅
One-click PO	❌	✅	❌	Basic	❌
AI summary	❌	✅ LLM narrative	❌	❌	❌
Price	Included	₹499/month add-on	Included	₹X	$99–299/month

---

Open Questions to Resolve

1. Backend access — Can we ship a new /web/items/intelligence endpoint on the main Rails/Python backend? Or do we route via genai/Snowflake?
2. Velocity window — 30 days default? Or 7/14/30/60 user-selectable?
3. PO flow — Does a Purchase Order creation API exist? Or do we open the purchase invoice creation flow with pre-filled data?
4. Restock via WhatsApp — Does the WhatsApp sharing infrastructure accept arbitrary text messages (not just invoice PDFs)?
5. Health score formula — Agree on weights before building.
6. Paywall placement — Is this a new micro-app at ₹199–499/month, or bundled into an existing plan upgrade?