We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 - 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!
The 20 to 50 billion new devices coming online over the next few years — estimates vary — represent a massive opportunity for businesses, and not just technology businesses. Smarter devices and troves of sensor data can improve existing business models and lead to the creation of entirely new ones.
The device population explosion will lead to a data explosion that will spark the creation of businesses based on scalable, low-churn models in the coming year. Many of these businesses start with a hardware purchase, but none end there. For instance:
- Services that improve with individual usage. A security camera that learns to distinguish household members from strangers is much more likely to deliver context-sensitive alerts/recordings worth paying for, versus paying for hundreds of recordings of the mailman delivering a letter. Same goes for a fridge that can tell over time which brands to automatically replenish versus one-off items. The dossiers that devices build for their users are forms of lock-in. Nobody wants to retrain their devices.
- Services that improve with penetration density. There’s a network effect to smart devices. Shared learning (as in the self-driving car algorithms at Tesla and Google) and shared real-time data (like the Ring doorbell Neighborhoods feature, or the Tile Community Network) can make the value of a device on a network far greater than it is by itself. And again, this network effect leads to hardware and service lock-in.
- Devices that require higher-margin replacement parts. Think of a smart connected toothbrush (like the Oral-B Bluetooth connected model) that promotes frequent and proper brushing and orders brush replacements for you, or a smart water filtration system that orders its own parts and maintenance, or a smart thermostat that orders air filters. To succeed, these models depend on grabbing the consumer’s attention with ancillary data and analytics (tooth brushing reports and feedback, for example) to keep reinforcing the need for regular attention and replenishment.
In contrast, devices that try to force a subscription for a minimal level of functionality, or with services that do not improve over time as competitors catch up, don’t stand a chance. The Internet of Things is going to kill static subscription services.
But we are going to run into roadblocks unless we quickly embrace more modern development methodologies, security standards, and monitoring tools. There’s also the problem of a very small number of developers with experience in IoT devices and data structures, compared to the hordes of cloud and mobile developers.
The easiest way to get developers working on IoT applications would be to give them familiar tools that abstract away the physical and low-level complexity of IoT.
Enterprise developers are already using Linux container approaches (see Docker, CoreOS, and Apache Mesos). Containers will become important for the development of IoT systems as well. They open the door to virtualization, and bring benefits like failure containment and application isolation. These protections will make IoT systems more reliable, as well as easier to build.
I predict significant strategic announcements in 2017 around support for container-based architectures for IoT from the likes of GE, Samsung, and Intel.
Inevitably, IoT devices will get more powerful due to the falling costs of full system-on-chip architecture. It makes sense that they be treated as full-fledged, cloud-enabled network computing nodes. The more powerful they are, the easier it will be for developers to take their familiarity with existing developmental methodologies (including containers) and apply those skills directly to these devices. Developing the data architectures and analysis is a different skill set, but at least on the hardware side the future doesn’t look too alien.
Security will improve alongside the drive to known development models, but we can also look forward to new standards that will constrain network traffic depending on what a device attests that it needs. Work that the Docker community is doing to implement signed images and attestation on the cloud side can transfer to the IoT domain.
And for industrial IoT applications, we’ll see continued decentralization of intelligence: Nodes won’t connect to the cloud, but to edge or intermediate points. Mobile data bottlenecks, security concerns and physical limitations, coupled with the need for fast reactions between sensors and actuators, (e.g. automated factory sensors that predict failures locally and take preventative actions, like the platform Alluvium is building) all point in this direction. I expect to see machine-learning driven decision-making applications close to the edge (perhaps even on-chip), with gateway devices acting as relays to the cloud for backup, updating, network-effect sharing, and diagnostics.
While these improvements would speed the connected device explosion, their interoperability and updateability are nowhere near where they need to be. Getting the development part of IoT right is a good first step.
Andreas Stavropoulos is a partner at venture capital firm DFJ. (Disclosure: DFJ has invested in Tesla, Resin.io, and Ring, companies relevant to this discussion.)
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn more about membership.