/var/blog
Industry and Technology Insights and reflections by Mike Bushong
/var/blog
IoT and Changing Business Models
09.12.17

If you believe IoT enthusiasts, basically everything becomes a sensor that produces data that can be used for some application. The most straightforward business models are to either sell the sensor or sell the application. Indeed, we have seen this in everything from wearables to industrial IoT. 
 
But as we generate all of this data, we will see the continued emergence of data as the product. Dare I call this Data-as-a-Service?
 
Data, data, everywhere
Obviously, the base premise of IoT is that distributed sensors become data sources. That data can be very basic, as with simple counters. It can be rich with context, using metadata to tag sources and times, for instance. It can be correlated across multiple sensors. 
 
Whatever the format, the point is that the data is going to be consumed, typically by some cloud resource, for repurposing as part of an application that provides some form of visibility and/or automated response. 
 
Historical business models
While there are all kinds of different businesses, we tend to see a foundational set of business models that get stamped out over and over again. If you doubt this, consider when you last heard some company describe themselves as the Uber of whatever industry. 
 
Now, those models might vary. Some companies will directly monetize a product or service (this is basically every company you know). Other companies monetize distribution (think: anything streaming). Some companies monetize a brand (franchises, for example). And of course, a lot of companies monetize their users (social media or anything related to advertising).
 
But with the emergence of all of this data, we will certainly see more companies who choose to monetize the data. And IoT will only fuel these possibilities.
 
Give away the razor

 

razor.pngThe most cliche business model discussion of all time is the razor blade analogy. You’ve heard it before, but in case you haven’t, it’s basically the idea that a company will give a razor away because it baits you into having to buy replacement razor blades for the life of the razor. Indeed, the price of the razor pales in comparison to the lifetime cost of the blades. 
 
That analogy can easily be extended to IoT. If the data proves valuable, then the cost of the sensor might be negligible compared to the lifetime value of what the sensor provides. Minimally, this has a pricing impact on the sensors themselves. If the data effectively subsidizes the sensor cost, then the unit cost can actually dip below the cost of goods and services (COGS). Essentially, it becomes a loss leader.
 
Accelerating IoT adoption
If the price of the sensors is essentially subsidized by services riding on top, then these sensors become more economically viable in mass deployments. When individual decisions are being made on a sensor by sensor basis, this might not matter much. But when you consider the cost of blanketing a city with millions of these things, the lower cost could actually trigger a spike in demand.
 
And, of course, the acceleration of IoT sensor deployment would mean that there is even more data that is available for potential monetization. Depending on how that data is used, this could mean an even more attractive product (i.e., higher pricing or more demand), which further subsidizes the sensors. This kind of virtuous cycle would accelerate adoption still faster. 
 
How do you monetize data?
If all that data is out there, the first question is what do you use it for?
 
There are all kinds of easy use cases to pick off. Data from cars can be used to train autonomous driving models (machine learning stuff). You can imagine Google, for instance, licensing their technology to the vehicle drivers under the premise that the software that powers the cars is higher-margin than trying to enter the costly vehicle manufacturing space. And it doesn’t hurt Google if they run their software locally, using yet another surface to monetize ads.
 
In fact, any applications that depend on training data for their machine learning algorithms would benefit from a data source. In some cases, if the application owns the point of collection and the application, there is no additional business model (GE monitoring their jet engines doesn’t create an additional data business model). But if someone other than the application owner owns the point of collection, they have a fairly captive customer.
 
You can imagine security cameras in retail stores tracking foot traffic. If you mashed up computer vision applications so they understood what products are placed where, you would essentially get a customer map to help optimize merchandise placement. Someone like Walmart could then basically charge the suppliers for better locales within the store, which would subsidize the cost of goods, allowing them to price lower than, say, Amazon.
 
If you expand the notion of sensor a bit, it’s easy to imagine location services in branch connectivity devices. The CPE (or the APs within the branch) would basically know who is in the office and what rooms they are in. This data can be monetized by carriers as a form of remote facilities management that acts as an upsell on basic connectivity managed services. 
 
Beyond mere collection and direct monetization, there is an additional opportunity to provide comparative services. A company might be interested, for example, in how they are performing relative to some peer group. In this case, they opt to share their data and pay a subscription fee in exchange for performance benchmarking. And the more companies that opt in, the more valuable the data set, so this creates another virtuous cycle.
 
Implications on the technology
If data truly emerges as a business model, this has implications on the devices themselves. In some cases, it means that either the sensors or natural aggregation points (IoT gateways) will have to support meta data. Tagging the source or the time to some collected data is critical. And anything that has tight thresholds requires some thought into things like time stamping and clock drift over distributed areas. 
 
If the data is sensitive, there is an obvious requirement around security. That might lead to encryption on either the sensor or the gateway device, or it could require basic policy decisions to send Traffic A to Resource A and Traffic B to Resource B. For instance, all traffic from a particular set of sensors might be sent to some collector that resides on customer premises for data protection or sovereignty reasons. 
 
Of course, any requirements like these place additional load on either the sensor or the gateway. Whether it's IPSec encryption or running a collection agent on the device, the additional load might require a more powerful CPU. Or it might open the door to disaggregated system design where the control plane runs remotely and the gateway is basically just a dumb, distributed line card with basic encryption capability and some resources for an agent or two.
 
The bottom line
There are certainly straightforward paths to monetization with IoT. But if combatants limit themselves to conventional warfare, I predict that they will be disrupted by someone who sees value beyond just the applications. And the best part of using data as a key part of your business proposition is that the greater the data set, the more valuable the product… and the wider the competitive moat. 
 
This means that companies building out products and services ought to be doing more than just looking to get data. They ought to be explicitly architecting solutions and setting pricing to accelerate data acquisition. In other words, they ought be using IoT to define their business models, not merely accentuating them.

Top Kudoed Authors
User Kudos Count
4