IoT Digitisation (finding the secret sauce, business needs to insight delivery)
The advent of IoT Digitisation heralds the possibility of creating entirely new ways to build a direct and personal relationship with customers on one hand, and have the greatest level of individual product tracking on the other. The executive C-Suite understands the opportunity, but they are wholly un-informed on how to go about setting the data value initiative up for success. From their perspective, some think it is the same as a software product selection process. That is to evaluate different suppliers and pick the best one based on cost, terms, functionality and other key metrics. But, it’s more than just selecting a software product.
There is no doubting that the Technical Landscape of IoT and Big Data architectures that support storing of IoT data is a crowded one. The complexity of the technical landscape as well as the predominance of open source software has led to companies that take the leap into the digitisation / big data / IoT world having the challenge of building an extensive software development team. This is often a more than subtle transition to a software development world from their core business. So, the question of how best to manage this transition arises and a question of the differences between time to value versus effort to value in deriving the desired value from their data arises.
Having worked with many customers in the last 5 years in the analytics space, the two biggest determinants of a successful path to deriving value from data that I have seen are appropriate executive sponsorship and internal / external collaboration frameworks. The executive sponsorship is needed to ‘knock people’s heads together’ as an old Cork colleague of mine once said. The two heads being the IT department and the business function. The collaboration requirement is nuanced in that there may be several ways in which this can be achieved. Sufficient executive sponsorship together with employees and processes that support inter department integration and cooperation are fundamental to success.
A by-line that is often noted in this space is ‘Fail fast, fail often’. What does that mean, and is it something that senior executives want to go around saying? I try to explain this another way. You have a business question / problem or opportunity that you think can be answered by the data that your business creates or has access to. The quicker you can find out if the data will answer the question that better. I have built agile methodology’s to achieve this and I often say that those investigative initial projects have two successful outcomes. The first is that you find the insight / answer to your question that you were looking for. The second ‘successful’ outcome is that you don’t find the answer. Some might see that as a fail, but you have learned something. You need to change the question or get more data to answer the one you have. The quicker you can iterate through this process, the faster you can get the right answer to the right question.
Now, for some things that I think the IoT space in Ireland needs. The first is sufficient support for start-ups in the data analytics space. I was invited by the German-Irish chamber of commerce on an IoT study tour of companies, universities and research institutions in the Ruhr valley. On a visit to the CITEC University in Bielefeld, I heard how they give up to €50K worth of free support for start-up companies to take their research ideas to market. Ulrich Ruckert, their head of Cognitronics and Sensor Systems explained his reasoning. Companies pay taxes to the government, the government gives grants to the university to do it research so he feels an obligation on the universities to give something back to industry. What a great idea!
Another area that I think progress could be made is whether there is space for a better data science language? Language is used to convey ideas. Although I can readily see how using a principal component analysis on a wide data set, a logistic regression on a dependent variable of interest coupled with a decision tree model will help implement a predictive capability for managing that same dependent variable, does my business user understand what I am saying. A new language may not be needed, but we need more stories. This is the best way of conveying the value of data science techniques to potential business sponsors.