Predictive analytics has become a critical part of the marketing toolkit in recent years, but in truth its influence has been rising for years, says a new report from marketing software company Kenshoo.
Studies by Accenture into the work of analytics professionals between 2009 and 2012 revealed almost a tripling in the number of organisations that primarily applied a predictive modeling approach to analytics.
Since then, the data science star has only risen further, so it is little wonder that the market for predictive analytics software is set to grow to $6.5 billion by 2019 according to Transparency Market Research.
Marketers and their thirst for customer insights are a big driver of this trend. As Kenshoo’s recent whitepaper, “The Kenshoo Guide to Agile Marketing” (download here) noted, the growth in the amount of data collected by marketers has largely outpaced the ability of companies to analyse that data.
“Traditional analytics aren’t very well-suited for uncovering the full value in today’s large data sets, so technology is needed to perform some of the most valuable analytics for digital marketers.”
But here’s the rub. “It’s not just about analysing historical data and current performance,” say the authors of the report. Perhaps more importantly it is about how those insights can be activated and applied to make future predictions.”
Underscoring this is a technology approach called machine learning.
According to the report, “Machine learning represents technology that allows marketers to access the vast amounts of data that we have been accumulating (and which are only continuing to grow exponentially) to help uncover and understand the patterns that emerge and act on those predictions.”
The authors say the machine learning approach allows for processing of billions of data points which when taken as a whole, paint a valuable picture. “Machines are able to use all the ingested data points to test the accuracy of predictions and train the system to improve forecasts. As the system becomes skilled at recognising valuable patterns, these models can be easily distributed and disseminated across a predictive marketing platform at scale.”
Of course organisations operate at a different levels of maturity where data driven marketing is concerned and it’s useful to understand where your own company sits on the curve.
The authors offer four questions marketers can ask themselves;
- How can we account for all the different inputs?
- How do I know if I’m interpreting the recommendations properly?
- What do I do once I have the insights and predictions?
- How can a machine know my business better than I can?
“To address these questions, it helps to start from the bottom up. First and foremost, you have to put your faith in the system. Recognise that, indeed, a machine can’t know your business as well as you and that’s why you get to pick which machine you use and program it with your goals and objectives,” say the authors.
“It’s part art, part science; part people, part platform. From there, pick a machine that can not only deliver insights, but can help you act on them in an automated manner.
“For example, seek a platform that can provide a forecast and recommended budget pacing and also implement it through a push of a button that sets advertising bids on search engines and social networks. It’s also important to ensure your platform has easy-to-digest dashboards that clearly show the opportunities.”
And, finally, they suggest, leverage a technology that can ingest custom inputs from internal systems and third-party partners such as inventory feeds, promotional calendar, or CRM databases.
For its part, research company Gartner offers a simple model with five levels which run from adhoc at the start of the data marketing journey through to the final state which they call the “decisive level” where organisations use real time segmentation and budget for experimentation.
Most importantly though, for these marketers analytics differentiates the business.