part (a): Did we get IoT right?
Recently I visited a small business owner who runs a modern manufacturing shop with robots. When we met, he asked, “why would I want to get a complex, cloud-based IoT system just to know if my robots are running? For this price, I can hire a (low-waged) employee to see if my robots are moving – if they stop moving, he can just text me!”
The business owner is absolutely correct.
If the only metric we are tracking is for a few machines to discover whether they are turned ON/OFF, I doubt we have a sustainable business model with IoT. Most of the devices and sensors worth monitoring already come with simple alerting tools. If they do not come with such controls, they can either be monitored manually, or their simple statuses can be inferred. Do you really need an app to let you know if your car stopped running? Seriously?
We (then) started talking about how impressive it is to see a US-based manufacturer successfully compete in a free market, and the owner started beaming on how his factory excels in creating great quality plastic mold injections, and how they have to go beyond manufacturing products and deliver additional services to provide a higher value and justify the prices they charge. The supply chain is no longer as simple as picking up raw materials, building a product, and then shipping it (see upcoming post on building new revenue channels with IoT).
There was something worth digging here...
For small business and large enterprises, running a modern manufacturing business involves defining and tracking KPIs at a greater fidelity, ideally in real time. For example:
· what is my resource consumption for a raw material in real-time?
· how may defect parts am I producing, and how does it compare to historical averages? How accurate is that information, and how current is it?
· how is my plant performing versus my other plans in South East? What is the cause for my San Antonio plant not performing as well as my Austin plant?
We stopped focusing on the “I” in IoT (connected devices) and started focusing on the new data sets we would be now collecting…
We went to the drawing board and started jotting down the metrics the owner would like to track from her IoT application if the technology and the economics of getting that data permitted.
Unlike manual observation, your IoT application should richly harvest the device meta-data to capture the nuances of its behavior. In this way, we can recommend scheduling to service the machine, reducing downtime. In this way, we can start tracking production trends that may adversely affect the supply chain before the consequences become very punitive.
Unlike traditional monitoring tools that come with your devices/sensors, your IoT application should have the capability to store large volumes of historical data, and be able to derive meaningful insights without requiring significant “data science” talent. For example, we can start comparing the real-time performance of a line with its historical behavior. We can compare the performance of one site with another site (VP-level insight). We can track product lineage! For example, we can track if extenuating environmental conditions existed when a particular batch of product was manufactured. Your IoT system should then be able to identify outliers in product or manufacturing behavior that it is observing. In addition, by correlating (historically measured) environmental conditions to product characteristics, we can improve product recipes.
When we set out to build IoT Applications at Oracle, we focused not just on connecting the devices, but on improving business outcomes such as the ones listed above. So far, the response from the market has been amazing!
Soon, we developed a list of metrics (not just production related, but also related to order fulfillment), technical capabilities such as customer portals and mobile applications to track orders and their current stage in the manufacturing process.
For many of these new capabilities, the owner felt he could now create “new products” and build new revenue channels. He was excited. … I was also excited. We both had arrived at a realization – we now understood that IoT was not just about connecting the devices and tracking their status on a dashboard.
Part (b) Introduction to Metcalfe’s law
On May 22, 1973, Bob Metcalfe circulated a memo called “Alto Ethernet”. Besides hailing the invention of Ethernet, the memo also defined the Metcalfe’s law, thus stated:
Metcalfe's law states that the value of a telecommunications network is proportional to the square of the number of connected users of the system (n2).
Since then, the law has been generalized to infer that the value of any network grows proportional to the square of connected nodes. For example, a social network with 1,000 users is 100,000 times more valuable than a social network with 10 users. That was not much of a difference in the number of users, was it?
Many suspect that the same holds true for IoT systems. To develop an IoT strategy, understanding this law is vital. This law also provides a framework to quickly develop new services for IoT (to be discussed more in the next section).
Consider an example where a business owner wants to track the production process on a manufacturing line. The obvious sensors to track will the ones on the robots in the production line measuring, for example, pick and place rates:
· We can track how fast the line in the plant is producing products
· We can track if a line is operational or not in real-time
Umm. OK. Are you excited? Do you believe that this is a transformational technology? No?… good! You are correct. The LOB owner always had this information. We just made the presentation of the data slicker, that’s all.
Now let’s get Metcalfe’s law operational… let us put sensors in all lines and in all plants. Besides tracking the pick-and-place rates, we also track the production shift.
Now, we can start getting additional insights:
· How are different lines in my plant operating?
· Is there a shift supervisor under which the production efficiency improves?
· Why is one plant always performing lower than others in the last week of February? (hint: it is in New Orleans – heh, heh)
· Compared to other lines and historical averages, which manufacturing step is a bottleneck (we may find that that particular robot does not get serviced as often)
These insights combine the power of real-time data access with IoT and advanced data analytics. The LOB owner did not previously have access to this information. This is the new value.
Are we getting excited? Maybe? Let’s try harder…
Now, let’s add different types of sensors. Let us add RTLS tags, RFID readers, and develop geofences with Oracle IoT application inside the plant to track production stages. Now, I can track KPIs such as these in real-time on a modern interface:
· What line is processing a given customer order?
· Did the raw material to service a given order arrive on time?
· In how much time will my customer order be manufactured?
· Did I make a profit servicing a given customer order? (this is a board-level insight!)
In addition, I can create a notification service and an app that lets the customer track her order in real-time – this improves customer experience, no?. We still have not considered new services that can be created once we have this information. For example, we can charge different warranty rates based on the complexity of manufacturing the product etc.
So, as you can see, the value of an IoT system increases with more sensors. Sensors are now cheap, and they are easy to assemble. By putting lots and lots of them, we create incremental value at a “square rate”.
There, Dr. Metcalfe, did I use the term right?