Data is cheap but infrastructure is expensive

J Matthew Pryor

J Matthew Pryor / January 31, 2022

3 min read––– views

Dedicated IoT networks have terrible economics and don't scale.

There have been many waves of enthusiasm and investment around pervasive IoT networks. Connected devices that will stream data from every part of the globe and enable businesses to make better-informed decisions and drive out waste. This has been promised over and over in agriculture, always just on the cusp of delivering real value.

Just as they confuse the user with the beneficiary, people also confuse the cost of installation with the cost of operation, and in the end, the cost of operation wins.

There's no point connecting devices if you're not connecting people.

The recent announcement of French company Sigfox filing for bankruptcy protection is no surprise at all.

The field of dreams nature of dedicated network infrastructure for device-only networks is bound to come unstuck - especially in sparsely populated areas. The business model relies on the enticingly low cost of participation but demands huge uptake and low churn to absorb the constantly high cost of operation.

Data is only useful if people can see it, and in regional areas, people need to be connected to see their data.

Infrastructure costs must be amortized as widely as possible.

For remotely connected device applications it is tempting to solve for power first.

Sigfox and LoRA are specifically designed for low-power and low-cost participation. This comes at the heavy price of needing entirely proprietary infrastructure. Installation costs can be controlled (to some degree) but operating costs are hard to control. Dedicated infrastructure needs dedicated support and with low margins and high visit costs, things can go south in a hurry.

Technologies like 4G/5G and Wifi are not low power but are pervasively supported, and can be shared by devices and people, creating the ability to share operating costs and motivate action to lower support costs.

Device-only networks are a premature optimization, and won't serve the needs of digitally nature agriculture.

Focus on the goal. There is no point in having more data if people can't get access to it

Optimize for operating cost. Maintenance cost will always trump installation cost over the long haul

Just because you can doesn't mean you should. Directly sensed data might be more accurate, but the cost of the additional accuracy often outstrips the additional value delivered.

New sources of data are critical to improving performance and sustainability in agriculture, but the operating cost of getting the data have to be matched with the value delivered and the costs that the market will bear.