Predictive analytics: a retail game-changer
Friday April 26 2013
Big data expert Rakesh Harji examines the challenges and rewards of rolling out a predictive analytics solution in todays retail market
Conditions for retailers in Britain remain tough. Consumers have smaller budgets and a wider variety of places to shop, so it is getting harder and harder to build a loyal customer base.
Rakesh Harji, UK managing director of predictive analytics and Big Data specialist Blue Yonder
, said today’s environment means that, in order for retail business to remain successful, they have to move from using intelligence to look at past results and start using data to predict who wants what.
In fact, Harji claimed predictive analytics – if utilised correctly – is one technology that may provide help to struggling retailers. “Seen by many as a game changer it helps companies gain crucial insight into external factors including customer demand, market prices, or risk exposure,” he said. “Using predictive analytics enables companies to combine analysis of these factors with deeper understanding of historical data to forecast customer demand. These insights allow companies to become more proactive in their customer offering and adapt to select the next best action.”
Implementation interest growing
While implementations of predictive analytics remain relatively low, the analytics expert said there is a high level of interest and a number of companies who are looking to implement these systems. Blue Yonder-sponsored Forrester Research from January 2013
showed that one in three of large enterprises surveyed across North America and Europe had expressed an interest in rolling out predictive analytics systems.
But Harji advised: “Successful implementations need to overcome the same challenges associated with all business intelligence solutions. These include skill shortages and the quality of data that’s put into the system. Yet predictive analytics adds a new layer of challenges in the form of data output quality and the business fit. Traditional reporting is based on and delivers information about the past.
“Although this information may be tricky to interpret, you can be sure that so long as the original data inputs were correct the output will in turn be correct. However with predictive analytics the data output can be misleading, even if the original data was correct. The model applied to the data needs to fit the business question it was supposed to answer.”
The primary success factor for the use of predictive analytics solutions, according to Harji, is therefore the fit of the models used to the business use case. “Companies either need to have the appropriate in-house statistical/machine-learning skills or rely on pre-packaged solutions and external expertise to understand and to make sure the models used fit the business purpose,” he explained.
Best-practice deployment tips
Once the model and business use are established, he said companies then need to consider a range of other factors to ensure the successful roll of a predictive analytics solution.
Align business and IT with executive sponsorship. “Predictive analytics projects should be driven by business to insure tight integration with and usage by the relevant business process owners,” Harji commented. “Ensuring a project receives appropriate support from within the business is dependent on support from the top, predictive analytics projects should always be on a board room’s agenda. For successful implementation projects, business and IT need to team up closely under an appropriate executive sponsorship that can align business and IT.”
Build your own predictive analytics roadmap. “Define your plan for predictive analytics from a clear business scope to the selection of the appropriate data sources, getting the right models in place, and measuring the results in relevant business KPIs [key performance indicators],” he added.
Good results need iterations. Here, he said: “There are many possible use case scenarios for predictive analytics in every company. Start with an achievable scenario and sharpen the business use case through a series of iterations and improvements.”
Integrate with business processes
Feed the beast with accurate and timely data. “The more data you have, the smarter and more accurate the predictive forecast — assuming that you have strong information governance in place that ensures high-quality input data,” he stated.
Integrate the results into the business processes. “Don’t run predictive analytics as a standalone initiative — for example, within a small group of data scientists,” he continued. “For maximum impact and success, feed the results of your analysis directly into your business processes for users to make better decisions or for the automation and optimisation of the business process.”
Get the right expertise on board from day one. “More than most other analytics technologies, predictive analytics needs the right skills to ensure not only high quality data input but also the appropriate models that fit the business scenario for accurate data output forecasts,” said Harji. “It is important to choose the right solution at the outset Companies need to ensure that the appropriate skills are on board for successful predictive analytics projects right from the very beginning.”
Consider using outside help. “The challenges of skill shortages, the business fit of applied statistical models, the level of integration into IT and business processes, along with change management are all good reasons to involve third-party support in predictive analytics projects,” he added. “Third parties can complement existing, internal know-how with external best practice expertise — for example, identify which data sources to use for which business scenario.”
Ensuring maximum competitiveness
While it is clear that rolling out a successful predictive analytics solution requires careful planning and investment of both time and resource, Harji restated that the benefits will be worth it. “Even if the economy recovers faster than predicted, retail conditions will remain competitive. Businesses need to understand not only the past decisions of their customers, but also need to change their product and strategy to better adapt to customer behaviour in the future,” he said.
“Predictive analytics is ideally suited for this purpose allowing organisations to improve their business processes and enhance their decision-making. Those who invest in the right solution now will better manage the present and increase the probability of future success,” concluded Harji.
Tagged as: Blue Yonder | big data | data | predictive | analytics | forecast | business | intelligence | BI | process | skills | decision-making