Data and IoT: The Secret Sauce

Data and IoT: Discussion about the Internet of Things often centers around sensors and hardware, the additions to our physical environments. They may be built into street lights and bus stops, or in factories and on assembly lines – or even inside shipping containers. But in fact, the ability to connect the dots will be even more important for the future of IoT.

In its first wave, the “I” in IoT has been mainly used to describe devices that can connect to other devices: we can now free up our central heating from a smartphone so we won’t freeze when we come home, we can tell our TV set-top boxes to record the football game we would miss because of a business dinner, and we can tell the postman via a video doorbell where to leave that parcel. More interestingly, and in the longer term more relevant, will be what connectivity also brings – masses and masses of data, all nicely time-stamped and associated with a specific process in a specific device.

Data and IoT: Deeper insights

In short, machines, robots, and sensors have become data nodes and the true opportunities can now be unleashed. Few important use cases would be possible without meaningful processing of data and increasingly this involves artificially intelligent systems because they can amplify the usefulness of a solution by instilling deep insights, not apparent to humans, into structured data sets. As a result, Data and IoT becomes the focal point of a number of meta-trends of the near future, namely ubiquitous connectivity and ubiquitous computing power, decentralized, in the cloud and with data-processing capabilities that puts everything previously known to shame.

In this way, a simple tracker beacon system in a store becomes an analytics tool that delivers rich information about customer journeys and habits, the attractiveness of certain product groups, and a plethora of other information, for a hardware cost of a few hundred dollars. The combination of structured data sets with the in sights of skillfully trained AI opens additional value streams not other wise accessible.

Wipers make weather forecasts: By capturing the data of auto-wipers in cars, manufacturers could create the timeliest weather forecasting service ever.

Another simple example is the amalgamation of all data on auto wipers in cars (wiper on = rain). A car manufacturer that captured and processed that data could be come the timeliest weather fore casting service. Remember that The Weather Company was sold to IBM – which is housing it under its Watson AI group – for a reported $2 billion in 2016.

Look at the infrastructure
These developments could have a bigger impact on industry than any incremental steps of automation because they allow value to be unlocked in parts of organizations that were not deemed valuable in the past. For example, Volkswagen would arguably not have attributed a ten-figure valuation to its wiper division.

To be able to surface these values, it is necessary for organizations to look at their data and IoT infrastructure and bring it into a shape where it can easily be accessed and processed. Arcane legacy systems will hamper the efficient deployment of IoT solutions as value can only be partially unlocked.

Machines, robots, and sensors have become data nodes and the true opportunities can now be unleashed.

This requires the building of a data science function that can dive deep into the leading edge of AI systems. The building blocks for such systems are now widely and easily available but the minutiae in their deployment is varied (do Monte Carlo systems perform better than multi armed bandits? No, I wouldn’t know either…) and the integration of such functions are not trivial. When well implemented, they will boost almost every organization’s ability to extract, analyze, and action data sets to improve performance on all facets of the value chain: faster and better product development, deeper customer understanding, more focused product innovation cycles, higher productivity.

Now everybody’s doing it! All major players are now offering a new wave of enterprise services that come with a specific sensitivity as they handle the very core of a company’s data infrastructure.

We are thus looking at a new wave of enterprise services that come with a specific sensitivity as they handle the very core of a company’s data infrastructure. This will likely take the shape of a layered cake: the various data layers residing in the company’s domain, whilst the processing of (often anonymized) data sets taking place in standardized AI frameworks hosted in the cloud. All major players are now offering suites of services and tools to handle the key elements of this, including Google (TensorFlow), Microsoft (Cortana), IBM (Watson), and Amazon (Lex and Polly via AWS). In the middle, we will often find specialized frameworks that focus on specific tasks and processes found in particular industries and market verticals (some of which are shown below). These come in at different points of the process – some come in at the time of data capture, some do the heavy lifting of the data collected, and others focus on the data output, for example in the way of virtual assistants.

Bringing dark to light

In industry, the first two will often be the most significant ones because they deploy relatively inexpensive tools to capture, structure, and analyze existing data and processes to propose improvements in processes along the entire value chain without the need for spending big on capital expenditure. The raw data has existed for a long time but it was largely unstructured and thus “dark.” It is the combination of data capture by deploying sensors, the amalgamation of the various data sources available in and to the company, and, last but absolutely not least, the analysis of ongoing comparisons and the iterative improvement AI affords, that is bound to make the biggest difference to the world we know. It lights up the data we have to 1hand by providing structured outputs that humans can process and understand. This is the true superpower that the “I” in IoT holds for us.

Leaders of the Data Revolution

Elon Musk, in his missionary Master Plan Part Deux, explained in 2016 how Tesla uses “fleet learning”, whereby it amalgamates every mile driven in a Tesla to inform its AI in order to improve autonomous driving capabilities. Two years ago, he said Tesla vehicles collected data for three million miles daily and the number of cars has increased annually. This would mean that Tesla now has data from about five billion miles driven to train its self-driving AI. Although it is rumored that Tesla uses components in its air-suspension system from Bilstein, the same manufacturer as Mercedes, and both have on-board GPS systems, Mercedes has not yet made the software connection that currently allows Tesla vehicles to automatically adjust ride height based on previously encountered road conditions.

READ MORE ARTICLES