Our Blog

where we write about the things we love

21

Aug

How to get started with Predictive Analytics

Recently predictive and especially prescriptive analytics started to receive a lot of press and attention. While predictive analytics has been around for decades, the latest technology advances have made the technical aspect of implementing predictive analytics much easier. 

Open Source tools raise and adoption

It is great to see the increasing number of university graduates who are competent in using R and Python. The functionality and flexibility of these tools have contributed to the growth of a worldwide community of data scientists who prefer an open source environment.

Increasing popularity of open source tools such as R, Python and Spark have prompted many vendors to recognise these tools and incorporate them into their product offerings. For example, SQL Server 2017 was announced to include in-database analytics and machine learning with Python in addition to R Services which are already available in SQL Server 2016.

Increasing maturity of machine learning services and cognitive services

Cognitive services capabilities continue to steadily improve in maturity and functionality as seen in Microsoft’s Cortana Intelligence Suite, making services such as Cortana Intelligence, Data Factory, Machine Learning more reliable, scalable and user friendly.

However, technology is only one side of the coin when it comes to predictive analytics. The main challenges for a successful implementation of predictive analytics remain the same as always:

Justification and measurement of business value

Some organisations still struggle to calculate ROI for analytics let alone predictive analytics. One will need to understand the business, its levers, and how to make better decisions to improve their future.

Resistance to cultural change

Predictive analytics requires adoption of entrepreneurial mind set – innovation, experimentation and acceptance of failure. It is much more comfortable to do things the way we have always done them – on an excel spreadsheet and relying on our “gut feel”.

Availability of skilled people

One of the biggest challenges in the predictive analytics journey is finding analytical skills. While I have mentioned university graduates earlier, the knowledge of technology itself is certainly not enough. The knowledge of R alone doesn’t mean that the person is ready to bring value; it requires much more to truly benefit from your data investments.

How can we ensure we are on the right track in getting the most value out of our data with predictions? I believe, among other things, the following are “must haves”:

Adopt the culture of experimentation

You will need to change your mind set and accept failure as a part of the process; with experiments, we need to fail fast. As Jeff Bezos commented on the failure of the Amazon Fire Phone, “If you think that’s a big failure, we’re working on much bigger failures right now”. The key to accepting mistakes and failures positively is to implement a framework for experiments or proof of concepts that will allow you to test your assumptions of the business value or abandon the experiment. When a possible failure is part of the plan and the process, and in fact, expected for a significant percentage of experiments, it becomes a norm and not associated with fear. Therefore, it promotes a culture of innovation when small number of successes pay for a significant number of failures.

Find skilled people

You will need people who understand statistics, data analysis techniques, data integration techniques, overall data architecture, as well as business drivers. Finding these people is one of the key factors to ensure your predictive project is successful. It is hard, I know, but not impossible.

Take one step at a time

Adopt a step-by-step method for realising the value from your predictive analytics. As a first step, enhance your organisational data with other data, i.e. geospatial data, sensor data or social data.

The next step is to socialise the findings and insights from your experiments with the business and identify the value of enhancing the insights further with your predictive analytics. For example, your business certainly may know that competitor prices and weather changes affects the revenue, therefore it is easy to justify the predictive analytics to deliver competitive up-to-the-minute pricing based on previous historical demand, current competitor prices and weather changes.

Additional value can be realised by going one step further and embedding your predictive analytics into your business process. Embedding predictions and prescriptions into your line of business application can help your business to get a better offerings to your customers and therefore improve revenue. For example, embedding the predicted best pricing into your ERP system will ensure consistent optimal sales volumes, high quality of customer services and revenue. 

Posted by: Anna Tarasoff, Data Insights Manager, Data Insights | 21 August 2017

Tags: Power BI, Data Intelligence, Digital Transformation, Less Busyness More Business, Data Analytics


Blog archive

Stay up to date with all insights from the Intergen blog