Never make predictions, especially about the future.
We don’t exactly have a great track record of forecasting the future, so this seems sage advice.
From IBM president Thomas Watson declaring in the early 1940’s that there would be a world market for “about 5 computers” (admittedly, there may only have been room on the planet for 5 of the early IBM machines ), through to the Y2K  hullabaloo, people can’t resist wading in with grandiose, often wildly inaccurate, predictions.
The rewards for knowing the future in advance are too great to resist giving it a go, but we rely rather heavily on human intuition to form our projections. As such, the rewards all too frequently remain unclaimed.
This is a rapidly evolving industry, however, and advances in artificial intelligence (AI) could soon see us base our future projections on reliable statistical models rather than our familiar-but-flawed intuition.
Within this three-part series, we will explore the potential role of artificial intelligence in developing accurate, accessible predictive analytics that lead to improved business performance.
This article will begin with an analysis of where the predictive analytics industry stands today, along with some tips to help businesses make the most of the available technology and data.
What do we mean by ‘predictive analytics’?
Predictive analytics is a form of data mining that employs machine learning and statistical modeling to predict future states of affairs, based on historical data.
There are examples of predictive analytics in action all around us already. If your bank notifies you of potentially suspicious activity on your credit card, it is highly likely that a statistical model has been used to predict your future behavior based on your past transactions. Serious deviations from this pattern are flagged as suspicious.
As a simple proxy for understanding the level of interest in the field, we can see from Google Trends that search volume for the topic ‘predictive analytics’ has risen significantly over the last 5 years:
We can look at this line and predict that it will continue to grow. But that is really just based on the recent historical trend, or the fact that we have heard a lot of buzz about the topic in the industry. It would take a lot more investigation for us to assert with any real certainty where the line will go next.
It makes sense that so many businesses are intrigued by the topic, too. It is predicted  that over $76 billion will be spent annually on ‘Big Data ’ technology by 2020. The best way to get a return on this investment would be to use all that data to anticipate future demand trends.
This is a difficult task for people to master, as we have seen. We need a bit of help if we are to start making correct predictions.
As a result, in Gartner’s ‘Analytic Ascendancy Model’, predictive analytics is viewed as an evolutionary leap from both descriptive analytics and diagnostic analytics.
That said, the desire for accurate predictive analytics is not new, and nor are attempts to use analytics to model future consumer behaviors. Many analytics professionals engage with this field everyday to calculate figures such as the lifetime value (LTV) of their typical customer, for example. The availability of vast and varied data sets has helped improve the accuracy of these calculations considerably.
What is relatively new is the application of artificial intelligence to plug gaps in our skill set and extend what is possible with predictive analytics.
This combination has led to more sophisticated statistical models that spot patterns in past consumer behaviors and use these to map out likely future actions.
But why is artificial intelligence, in particular, so effective in achieving what we have found nigh-on impossible on our own?
Creatures of habit: How is predictive analytics applied in the real world?
Predictive analytics is helped greatly by just how predictable people are.
As unique and free-willed as we would like to believe we are, AI can quite accurately predict what we will continue to do based on our past actions and the actions of similar people.
A study  by scientists at the Media Lab of the Massachusetts Institute of Technology in 2007 discovered that “a good 90 per cent of what most people do in any day follows routines so complete that their behavior can be predicted with just a few mathematical equations.”
A lot of marketing campaigns have operated on this assumption, but we can now apply this principle with greater accuracy and accountability.
Where AI comes into its own in this field is in its ability to identify broader patterns that humans simply would not see. We select areas for investigation based on what we believe to be safe assumptions, but AI can identify other variables that, when altered, have an impact on each other.
For example, based on my location, age, past purchases, and gender, how likely am I to buy milk if I have just added bread to my basket? An online supermarket can use this sort of information to automatically recommend products to me, based on my predicted propensity to buy these items.
Moreover, a financial services provider can use thousands of data points created by my online interactions and those of similar people to decide which credit card to offer me, and when. A fashion retailer can use my digital profile to decide which shoes to recommend as my next purchase, based on the jeans I have just bought.
This helps businesses to improve their conversion rate, but the implications are much wider than that. Predictive analytics allows companies to set pricing strategies based on consumer expectations and competitor benchmarks. It allows retailers to anticipate demand, and therefore ensure they have the right level of stock for each product.
Predictive analytics can even suggest ideas for new product lines by spotting changes in customer preferences. This marks the transformation of analytics from a retrospective tool for data specialists, to an essential predictive function that shapes business strategy, improves customer relations, and creates operational efficiencies.
The evidence of this revolution is already all around us. Every time we type a search query into Google, Facebook or Amazon, for example, we are feeding data into the machine. The machine thrives on data, growing ever more intelligent as it receives these feedback signals.
This phenomenon brings with it a host of benefits for marketers. Google has been using this technology for some time already, through its Smart Goals product in Analytics, and its Session Quality Score  feature, launched late last year. These are examples of predictive analytics in action, powered by machine learning technology.
There is an argument that prediction is the bedrock of intelligence, so this is quite the feat for AI.
This is only the beginning, however. Much of the current work in predictive analytics centers on making suggestions or recommendations, but there is scope for AI-based projections to form the fulcrum of marketing strategies.
Recent developments provide a lot of cause for optimism (or trepidation, some might say) in this respect. Google’s DeepMind team has just created an AI  that is capable of planning for the future and considering different outcomes before acting.
This is relevant within the scope of predictive analytics, as imagination is a fundamental aspect of creating projections. This ability will only entrench the role of AI as an essential component of a successful predictive analytics campaign.
How can businesses integrate predictive analytics?
To capitalize on the potential of artificial intelligence and predictive analytics, there are four elements that organizations need to put into place.
1. The right questions
The best predictive analytics projects begin with a sound hypothesis to test. Although we should provide machine learning algorithms with the room to make their own objective associations between data points, we need to set out with a business challenge that we are looking to overcome. This helps to provide some shape to the endeavor.
2. The right data
Advances in data science over the last decade mean that we can derive insights from large volumes of unstructured data with greater accuracy, but we do still need complete data sets to arrive at convincing conclusions.
Therefore, the next stage after defining the questions you want to answer with predictive analytics is figuring out what data is available to you and whether it will be sufficient to answer your questions convincingly.
3. The right technology
As implied with the projected value of $76 billion by 2020, big data technology is a booming industry. Data is created at such a rate that we require ever-improving technological capabilities just to capture, store, and make sense of it.
Many of the leading analytics software packages have already launched predictive analytics tools, but they do vary in their methodologies. To decide which solution is best for your business, it is more important than ever to have a team in place that has experience of each and can identify the most suitable option.
4. The right people
This brings us back to step one, essentially. Without the right people, it’s very difficult to pose the right questions. It is also difficult to know what data might be required to answer them, or to get the best out of the latest technology. For all of the talk of AI replacing people, it has only really sharpened the focus on getting the right people to make the most of the new opportunities it creates.
The applications of this technology are already widespread, but we are still just scratching the surface. In the next article in this series, we will take a look at five businesses that are using predictive analytics to drive improved business performance today.
Marketers undeniably face a wide array of challenges, but understanding and making use of customer data is key to each and every one of these. In Part 3 of our three-part series on Customer Data Platforms, we’ll look at exactly how a Customer Data Platform can address common problems faced by CMOs.
A Customer Data Platform (CDP) can help companies unify their data and acquire that single customer view that is so vital to marketing success. But how does a CDP differ from other data management solutions like a Data Management Platform (DMP), and what essential functionality does it need to nail?
Despite the Digital Marketing Association noting that 77% of brands believe real-time personalization of their marketing is crucial, a whopping 60% still struggle to do it. In this article, I look at why this is, and how a Customer Data Platform can help.
- ^ early IBM machines (upload.wikimedia.org)
- ^ Y2K (en.wikipedia.org)
- ^ predicted (www.snstelecom.com)
- ^ Big Data (www.clickz.com)
- ^ study (www.newscientist.com)
- ^ regression analysis (en.wikipedia.org)
- ^ report (resources.everstring.com)
- ^ Session Quality Score (searchenginewatch.com)
- ^ created an AI (deepmind.com)
- ^ … read more (www.clickz.com)