Dashboard User Guide

Goal of this chapter: Step-by-step instructions on how marketing specialists, account managers, and leadership can use the Streamlit web dashboard to find opportunities and make fast decisions.

1. Accessing the Dashboard

The Restaurant Prioritization dashboard is hosted internally on our Kubernetes cluster and accessible via company VPN:

Internal URL: http://restaurant-prioritization.internal.hungryhub.com
Supported Browsers: Google Chrome, Mozilla Firefox, Safari, Microsoft Edge.
Authentication: Secured via Google Workspace Single Sign-On (SSO).

2. Navigating the 4 Dashboard Tabs

  ┌───────────────────────┬───────────────────────┬─────────────────────────┬──────────────────────┐
  │  Restaurant Rankings  │  Restaurant Deep Dive │   Marketing Simulator   │  Operations & System │
  └───────────────────────┴───────────────────────┴─────────────────────────┴──────────────────────┘

Tab 1: Restaurant Rankings (The Master Table)

Best for: Weekly marketing campaign planning and ad budget allocation.

Default Sort: Automatically sorted by Marketing Priority (highest-opportunity restaurants appear at the top).
Core Table Columns:
- Restaurant Name & ID: Partner identification.
- Quadrant Segment: e.g., Hidden Gem, Core Cash Cow, High Risk, Deprioritize.
- Status & Location: Bookable status, City, and Cuisine category.
- Confirmed GMV (4w): Clean sales generated over the last 4 weeks.
- Recent Marketing: Facebook ad spend (฿), KOL visits, and Push Notifications (Sent Count).
- Badges: 💎 Hidden Gem, ⚠️ Churn Risk, 🚀 Rising Star, 🛑 Over-Promoted.
Quick Action: Click “Export to CSV” to download the filtered target list directly for uploading to Facebook Ads Manager or emailing food bloggers.

Tab 2: Restaurant Deep Dive (Partner Diagnosis)

Best for: Account Managers preparing for a quarterly review or partner check-in.

Search Bar: Type any restaurant name or partner ID to inspect.
Visuals Provided:
- 12-Month GMV & Covers Chart: Historical trend showing peak seasons, drops, and recovery periods.
- Marketing Timeline: Exactly when Facebook ads ran, which influencers visited, and what push notifications were broadcasted.
- Diner Feedback Gauge: Average star ratings, review velocity, and positive review percentage.
- AI Prediction Breakdown: Plain English explanation showing why the model rated this restaurant high or low.

Tab 3: Marketing Simulator (What-If Sandbox)

Best for: Marketing Directors testing budget re-allocation scenarios.

Interactive Sliders:
- “What if we increase ad spend on Quadrant 1 (Hidden Gems) by ฿50,000?”
- “What if we reduce ad spend on Quadrant 4 (Deprioritized) by 50%?”
Instant Forecast: The simulator uses model coefficients to predict the resulting change in overall platform bookings and GMV before spending a single Baht.

Tab 4: Operations & System Health

Best for: Data Engineers, ML Engineers, and SREs.

Displays database connection status to ClickHouse.
Shows timestamp of the latest model run and table exchange.
Tracks memory consumption, cache hit ratios, and API query latency.

The left-hand sidebar is divided into 3 organized filter sections:

graph TD
    subgraph SidebarFilters["3 Sidebar Filter Sections"]
        S1["<b>1. Primary Filters</b><br/>• Bookable Only (Default: Active)<br/>• City / Location (e.g., Bangkok)<br/>• Cuisine Type (e.g., Buffet, Japanese)"]
        S2["<b>2. Portfolio Segmentation</b><br/>• Business Quadrants (Q1 to Q4)<br/>• Strategy Badges (Hidden Gem, Churn Risk)<br/>• Pricing Tier (฿, ฿฿, ฿฿฿)"]
        S3["<b>3. Marketing Exposure Controls</b><br/>• Max Ad Spend Slider<br/>• Minimum Review Rating (e.g., ≥4.2★)<br/>• Days Since Last Marketing"]
    end

Pro-Tips for Common Workflows:

Workflow A: “Find this week’s hidden gems for Facebook Ads”

In Primary Filters, keep Bookable Only = Yes and select City = Bangkok.
In Portfolio Segmentation, select Quadrant = High Performance, Under-Promoted.
In Marketing Controls, set Max FB Ad Spend (Last 30d) = ฿0.
The resulting list gives you the top restaurants ready for paid campaigns.

Workflow B: “Find partners at risk of churn for Account Outreach”

In Portfolio Segmentation, select Badge = Churn Risk or Quadrant = High Risk / Needs Attention.
Sort the table by GMV Drop (%).
Export the list and share it with Account Managers for immediate partner calls.

4. Reading Marketing Channel Metrics Correctly

When reviewing the marketing history columns:

Facebook Ads (spend_thb): Shows actual cash spent in Thai Baht (e.g., ฿4,250).
Influencer Marketing (kol_thb): Shows talent fees paid to food bloggers (e.g., ฿8,000).
App Push Notifications (push_sent): Displayed as Sent Counts (e.g., “14 Sent”).
Email Newsletters (email_sent): Displayed as Sent Counts (e.g., “3 Sent”).

Tip

Never confuse Sent Count with cost. A push notification showing “14 Sent” means 14 broadcast messages were delivered to users, with zero direct ad cost.

5. Next Steps

To understand how the pipeline runs on Kubernetes and how to handle system alerts, proceed to 06. Operations & SRE Runbook.

HungryHub Knowledge Base