By Karl Fischer, MD of DVT Coastal, Head of AI, Data and Analytics
The rapid progression of data platforms and their capabilities has seen analytical models increasingly being used to model complex business scenarios for planning, operations, investment and innovation. Businesses continue to shift to data-driven decision making at all levels in the organisation as data streams, processing and subsequent insights become ubiquitous. Given the availability of these technological capabilities, the critical question is how to achieve success with these toolsets.
In the past, relatively scarce skills were required to perform statistical analysis. (How many people do you know that are comfortable with normal distributions and standard deviations?) Today’s data ecosystems and platforms though can easily facilitate connection to sources, wrangling the data and then structure, store and process with the elasticity of resources. Being on-demand in the cloud, these capabilities foster experimentation and ad hoc use that can deliver rapid results if you know the capabilities, dangers and have the people with the knowledge and experience to use them.
Tools to Help Managers Gain Insight from Data
Data is the starting point but just having it is a long way from gaining insight and advantage. Fortunately the major vendor platforms (Microsoft, Google and Amazon) provide relatively complete end-to-end capability that cover:
- Connection to edge devices (IoT) for data measure at source, in the field and in real time,
- Streaming of the data and managing the velocity,
- Filtering of the data to extract the key elements and ensure quality,
- Storing of the data for processing, and
- Then analysis and visualisation.
- Machine learning to build, train and deploy predictive analytics solutions to your data at scale.
- Cognitive services to enable organisations to develop AI solutions without the required machine learning expertise.
The challenge these days is less related to technology. Rather, it is more about asking (or at least establishing) the right and interesting questions to be asked. Thereafter, it’s a case of seeking the right data (valid, complete and sufficient) to assist in:
- Hypothesis formulation
- Hypothesis testing,
- Subsequent experimentation and/or modelling, and
- Finally, refinement or proof of hypotheses and lines of investigation.
The process should be familiar to anyone that’s taken a look at process improvement/waste elimination using methodologies such as Six Sigma or Lean. What practitioners are seeing is that such methodologies statistical analysis and presentation of results are now accessible in commercial tools and platforms. With these capabilities, it becomes possible for businesses to tackle questions such as:
- Where is cost optimization possible?
- How can customer churn be reduced by identifying customers that are at the risk before they are lost?
- How can you improve sales success ratio?
It all depends on you to establish the right questions, source the data, query, refine and then repeat. In fact, the process itself may lead you to questions (and answers) that were not yet even considered.
Key Dangers to your success with Data Analytics
- Ethics: Just because you can doesn’t mean you should. Privacy, data protection and your intent are issues that can and will cause major harm to your organisation should you not actively consider these elements and treat as core to every analytics initiative. Consider the downfall of Bell Pottinger as a case study of what not to do.
- Data integrity: The quality and completeness of your data is a fundamental to any analysis. Bad data as a basis for decision making is certain to produce poor outcomes. At the very least this will undermine confidence in your future initiatives and at the extreme result in decision making that is fundamentally flawed. Give specific consideration to data bias, insufficient training sets (immature models) and corrupted data sets (where did it come from and how was it prepared).
- Analytics skills: A title as a Data Scientist doesn’t make somebody an expert in data science. Too many practitioners have retitled themselves without the fundamental quantitative skillsets to match the actual role of data scientist. If your data scientist cannot explain the model outcomes your problem may be your scientist. If you do have a great, competent data scientist, your next challenge is holding on to them. Skills scarcity in this space is a major issue.
For organisations the opportunity to access phenomenal analytical capabilities (being the strength of the computing platforms) must be leveraged with understanding of the possibilities from such analytics and “learning” models. More than ever it is when the right combination of tools and human creativity are brought together that successful outcomes will be achieved. It is certainly an exciting journey and one the business world has begun in earnest.
Have some of your own thoughts? Feel free to send me an email on email@example.com.