It’s always fun to watch business consultants and IT companies search for the next buzzword - throwing candidates to the wall, hoping that one of them will stick and help them own the next big thing. (IBM is particularly adept at the game, rotating buzzwords every two years.)
That being said, these people are smart, and we should probably pay attention when they start pointing in the same direction.
We believe this is happening right now. For example:
While the names may differ, it seems that fast data, continuous intelligence, the cognitive enterprise, industry x.0, VUCA architectures, the cognitive factory... share four main components.
The ongoing deployment of billions of connected sensors means that streaming data (or event streams, or time series) is invading business data sets.
"Modern enterprises live in a real-time world, where the moment between capturing data and having an opportunity to apply it is both shrinking and becoming more crucial to business success." - Forrester
The challenge is that most existing IT architectures and decision management systems are not ready for that.
Descriptive, and even predictive analyses are not enough. Managers are looking for personalized, actionable suggestions. That means taking into account existing business rules and physical models, and optimizing for actual costs and constraints.
We find that business rules and physical models are reasonably easy to collect. More difficult is to write down the relevant costs and constraints and dynamically integrate them with the business rules, physical models and predictive models.
Our advice would be to start small. Even a basic set of costs and constraints will deliver the quantum leap from predictive to prescriptive.
Machine learning is the only option for making sense in real time of today's massive amounts of data.
"Without big data analytics, companies are blind and deaf, wandering out onto the web like deer on a freeway." - Geoffrey Moore.
Nothing new here, except that...
- Machine learning is only starting to spread beyond its initial ethereal business applications (ex. online advertising, market finance), into the gritty industrial world. This calls for higher robustness and explainability standards, and means that pure open source is not a realistic option anymore.
- It takes a special blend of machine learning to optimize for real costs and constraints, with a combination of multiple types of models, based on streaming data.
Pragmatic IT architectures
We are well past the data lake craze. Even IT services providers now realize that enterprises are not going to ditch their billions of investments in legacy systems for the sake of tech purity.
The goal now is to evolve pragmatic IT architectures, mixing on-premise and cloud elements, gradually using best-of-breed solutions to bridge gaps and enable new services.
"By introducing modern architecture patterns and principles, legacy systems can be modularized and modernized which then opens opportunities to introduce digital technologies into legacy systems." - Accenture
The typical setup for Datapred's direct material procurement software is a good example of such pragmatic IT architectures:
- We install Datapred on a private commercial cloud.
- Our client installs a standard SQL database on premise, that collects Datapred's inputs from multiple legacy systems and outside data sources.
- Datapred's outputs are sent back to that database, and then on to the same legacy systems and to a commercial data visualisation software that may be on premise or in the could.
So watch out for technologies that facilitate the deployment of some or all of these components: streaming data infrastructure, specialized machine learning solutions, modern business rule management systems... bright days ahead!
The Datapred modeling engine is built for continuous intelligence applications - providing increased algorithmic performance and a 10 x acceleration of the garage-to-factory cycle. Contact us to discuss how it could help with your current continuous intelligence projects.