Overview

BlendNet (piloted as Plagnet) is a last-mile distribution platform that improves reliable access to digital services in low-connectivity environments. It combines cloud services + an intelligent edge hub (placed with trusted last-mile intermediaries like kirana stores) to distribute digital content and services to consumers through local access.

🔗 Microsoft Research Project Page

Motivation

Even as connectivity expands, many users still face unreliable networks and affordability constraints, and also barriers to adoption (language, cultural trust, and payments). BlendNet's core insight is that local intermediaries (retailers) can bridge these barriers and become distribution points for digital services—extending beyond fintech into entertainment, education, skilling, healthcare, and more.

Product

• Intermediary-led acquisition and trust: retailer referrals as the primary growth lever; assisted onboarding and discovery
• Cloud + edge architecture: intelligent cloud services push content to edge hubs; consumer app fetches via local Wi-Fi hotspot from the hub
• Retailer dashboard: referral tracking + transaction workflows to support assisted selling (and measure retailer activity)
• Consumer mobile app: content discovery, engagement flows, and telemetry to track sessions/DAUs and conversion intent
• Secure content delivery: DRM-protected playback with device registration and license keys

Technical Stack

• Cloud microservices: provisioning, content workflow, retailer/user management, performance monitoring
• Edge hub: set-top hub device + router hotspot; device management via Azure IoT Edge
• Distribution refresh: bulk content pushed to hubs through satellite uplinks in the field trial
• Apps: consumer + retailer mobile applications (instrumented for engagement + stability)

Deployment

• Large-scale field trial (India): deployed across 258 retailers
• Territory coverage: trial executed across multiple territories with retailer hub placements and phased ramp-up
• Primary use-case: entertainment packs (regional + long-form content), sold via assisted retail model

Demos

• ACM COMPASS 2024 Presentation: Conference talk presenting the BlendNet paper.
  👉 Watch Presentation

• 2020 Pilot Video: Early field pilot demonstrating the platform in action.
  👉 Watch Video

• 2022 Pilot Video: Scaled field trial with retailer-led distribution.
  👉 Watch Video

Results

Quantitative

• Retail footprint: 258 retailers onboarded for the pilot
• User acquisition: 68,248 users joined during the field trial; 1000+ users/day, peaking at 2,955 new users on 29 Sep 2022
• Retailer-driven growth: 95% of users were referred by retailers; tracked via referral codes
• Referral volume: 64,651 retailer referrals recorded
• Content delivery: 232.89 GB delivered to hub devices; 5,240 downloads completed; 98.7% API success rate
• Retailer activation: active retailers grew from 15% to 68% by end of trial; daily transacting retailers grew from 10 to 227
• User engagement: daily unique users rose from 51 to 1,872; sessions grew from 666/day to 14,000+/day

Qualitative

• Intermediary model works: retailers materially reduced adoption barriers (discovery, trust, local-language/cultural context, and assisted selling)
• Hardware/network constraints surfaced early: download speeds degraded with concurrency, highlighting the need for better network management/hardware scaling
• Operational reliability improved with iteration: crash rates and download errors reduced after updates and field support, enabled by feedback loops with retailers