Why predictive analytics success comes down to trust
- 17 January, 2014 15:57
Organisations looking to make the most of predictive analytics must recognise that it takes time to build trust, and that success requires the involvement of the entire company towards a data-driven culture.
The advice from TWDI research director for advanced analytics, Fern Halper, stemmed from the research firm’s new study into the benefits and use of predictive analytics. Its Predictive Analytics for Business Advantage report delves into how predictive analytics is being used to date, the types of technology investments that need to be made, and drivers now and to 2016.
The research made clear predictive analytics is gaining ground as the natural succession to business intelligence (BI). Its new-found accessibility is thanks to cheaper and faster compute power, better insight into the value of proactive customer insights, economic considerations, big data, and ease of use.
“Predictive analytics has finally found its groove and there is a lot to say about the value it provides,” Halper said during a webinar to discuss the results. The list of drivers for predictive analytics adoption is led by understanding market trends and behaviours, followed by understanding customers and predicting their behaviour.
Business process was another reason cited by respondents, and TDWI cited several organisations using predictive analytics to drive better business performance, strategic decisions, and operational efficiency.
Marketing and sales led the charge in terms of adoption by user group, with top use cases being direct marketing, cross-sell and up-sell, and retention analysis. Seventy-two per cent of those surveyed said they plan to use predictive analytics for retention analysis over the next three years.
While most of the work done today is on structured data, TWDI also saw unstructured forms such as text data rising in importance in the next few years.
Why predictive analytics matters
How predictive analytics is helping AMEX tackle customer attrition
Big data, budgets and building predictive analytics
Yet despite the hype, there remains a significant knowledge and skills gap around predictive analytics’ role, timeframes, skills required and where to start.
“Part of building out a predictive analytics business is cultivating relationships and building trust, because sometimes professions are suspicious of new forms of analysis,” Halper commented. “It’s important to involve business and other parts of the organisation. Collaboration between business users and IT, as well as other business groups, is key.”
To do this, Halper advised starting with proof of concept projects with a business sponsor that demonstrate value for the business, and get the ball rolling.
“Success breeds success. A lot of organisations start with BI, get their organisation analytically aware, they start to see results they can use, and then that drives value,” she said.
Organisations also need to work in steps. “Business sometimes gets frustrated it’s not part of the change, so it’s important everyone gets involved, but that you also take it one step at a time.”
One of the biggest mistakes organisations make as they move into predictive analytics is not coming up with a business problem worth solving, even during proof of concept stage. “Even with big data, you can go in and explore, but you still need to at least be thinking about the area you are interested in,” Halper said. “This is about value.”
Another part of the process is determining the cost benefits of the models you want to build and deploy, and the skills available to you to execute the plans, Halper said.
“Enterprises are organising around predictive analytics in different ways,” she stated in the report. “Some are building out dedicated analysis teams. Others are building cross-functional centres of excellence and may have teams within the centres that serve different business areas. Information and best practices are shared across the entire team.
“It is a rare company that can assemble a group of rock-star statisticians to build and deploy predictive models. Even where that is possible, predictive analytics is not simply about building a model. Remember, different people across your organisation will have to get involved, especially if you plan to operationalise the model.”
As with most types of data-driven analysis, Halper also recommended adopting a good model management approach that can scale, as well as investigating different types of data, including unstructured, to utilise. Data integration, governance and a solid BI infrastructure are also important.
TWDI’s research incorporated 373 respondents, 34 per cent of which are actively using predictive analytics in their business today. Fifty-two per cent are investigating the technology, including 20 per cent engaged in predictive activity. Two-thirds of those surveyed were business sponsors or users.
Follow CMO on Twitter: @CMOAustralia, take part in the CMO Australia conversation on LinkedIn: CMO Australia, or join us on Facebook: https://www.facebook.com/CMOAustralia
