Custom IoT Solutions: Costs, Architecture, and How to Know If It's Worth It
The Internet of Things market will reach $1.68 trillion by 2030, with 29 billion connected devices generating operational data. 92% of enterprises report positive ROI from IoT investments, mostly in predictive maintenance, energy optimization, remote monitoring, and supply chain automation.
An IoT system has five layers: physical devices, connectivity, data processing, analytics, and visualization. Together they collect, transmit, analyze, and act on operational data. How those layers connect, and where the processing happens, decides how the system performs, what it costs, and how far it scales.
Custom versus commercial platforms
A custom IoT system is built around one organization's requirements. Commercial platforms offer standard features, deploy faster, and cost less up front. With custom development you own the features, the data, and the intellectual property. With a commercial platform you're tied to the vendor's roadmap and pay recurring license fees, which at scale often add up to more than a custom build.
Custom development makes sense when your own data analysis is a major source of value, when you have specialized compliance requirements, or when the system itself sets you apart from competitors. GE built a custom industrial IoT system that analyzes data from over 1 million sensors and cut downtime by 10%. Walmart's custom inventory management system reduced out-of-stock items by 20%.
Architecture layers
The physical layer (sensors and actuators) collects data from the environment and carries out actions. Sensor quality limits everything downstream.
The connectivity layer moves data between devices and processing systems. MQTT has become the standard protocol, and it uses 170 times less power than HTTP on 3G networks. The network also has to handle traffic from what can be millions of concurrent connections.
The processing layer stores and prepares data using time-series databases, data lakes, and stream processing services. How you build this layer decides response speed and how far the system can scale.
The analytics layer uses machine learning for descriptive, diagnostic, predictive, and prescriptive analytics. This is where raw sensor data becomes something a business can act on.
The visualization layer turns processed data into something people can read: business intelligence tools, custom interfaces, and mobile apps. Good visualization shortens the time between something happening and someone noticing it.
Edge versus cloud processing
Where data gets processed is the biggest architectural decision in an IoT deployment. Cloud computing centralizes processing in remote data centers, which gives you near-unlimited scale and advanced analytics services. It also adds 50-200+ milliseconds of latency, significant bandwidth costs, and a dependence on connectivity.
Edge computing processes data on the device or on a nearby edge server. Decisions happen in milliseconds. Only alerts or summaries go to the cloud, which cuts bandwidth by 90%+ and costs by 60%. The system keeps running during network outages.
Autonomous vehicles make critical decisions locally, because split-second obstacle avoidance can't wait on a round trip to the cloud. Manufacturing robots need response times under 10ms. Smart security cameras analyze video locally and send only alerts to the cloud.
Most real deployments are hybrid. A wind turbine runs vibration analysis at the edge and shuts itself down if it detects dangerous conditions, while the cloud stores historical data and trains predictive maintenance models on data from the whole fleet.
Industry applications
Healthcare IoT can cut operational costs by 50%. GOJO deployed 20,000+ internet-connected hand hygiene dispensers and saw a 60% reduction in hospital infections. The global healthcare IoT market is growing from $108.60 billion in 2024 to $167.70 billion by 2028.
In manufacturing, predictive maintenance has saved companies up to $25 million by reducing downtime. Stanley Black & Decker's smart factory increased router production by 24% after it got better visibility into its equipment. 66% of manufacturers say IIoT is essential to their success.
Smart buildings cut energy costs by 30-70%. The Burj Khalifa uses IoT sensors across its 163 floors and has reduced energy consumption by 40%. The Edge office building in Amsterdam reached a 70% reduction with 28,000 sensors that track occupancy and air quality.
Precision agriculture raises yields while using fewer resources. California vineyards reduced water use by 20% with IoT-driven precision irrigation. Brazilian coffee plantations cut pesticide use by 30% with drone monitoring.
Logistics and supply chain IoT spending grew from $10 billion in 2015 to $40 billion in 2020. DHL used IoT sensors to distribute over 1 billion COVID-19 vaccine doses, monitoring temperatures as low as -70°C.
Platform selection criteria
The IoT platform market was worth $11.10 billion in 2023 and is projected to reach $27.15 billion by 2030. AWS IoT Core has 13 dedicated services, with edge computing through AWS IoT Greengrass. Microsoft has invested $5 billion in Azure IoT, which integrates directly with Microsoft 365 and Dynamics 365. Google Cloud is strongest at data analytics, through BigQuery and its machine learning tools.
When choosing a platform, weigh security (end-to-end encryption, device authentication, ISO 27001/SOC 2 compliance), scalability (millions of devices, auto-scaling, cost efficiency), and device management (OTA updates, remote configuration, bulk provisioning).
Security implementation
80% of IoT breaches come from poor network integration. The OWASP IoT Top 10 lists the most serious vulnerabilities, including weak passwords, insecure network services, and the lack of a secure update mechanism. At minimum, use TLS 1.2+ for data in transit, AES-256 encryption for data at rest, and cryptographically signed over-the-air firmware updates.
HIPAA-compliant healthcare IoT requires unique user identification, audit logging, and transmission security. GDPR requires data minimization, privacy by design, and impact assessments for high-risk processing. Penalties reach 4% of global revenue or €20 million.
Mutual TLS uses certificates for authentication, which removes password-related vulnerabilities. Over-the-air update systems need SHA-256 integrity checks, rollback protection, and a fail-safe if an update goes wrong.
Need help planning your IoT implementation? Email us at hello@detroitcomputing.com.