What if we could use our customers' complete buying history, including all transaction data, to predict the next most likely purchase for each customer? This information would allow us to conduct individual marketing communications and provide tailor-made offers at the right time to each and every customer without even asking one single question.
Here we go
Big Data is making all this possible with its ability to detect correlations and patterns where human beings only see data chaos. Still important, however, is making sure that data is up-to-date, relevant and comes from guaranteed trusted sources, which sounds easier than it is. Consistency checks and data cleaning is unfortunately a task located somewhere in the no-man's land between marketing, controlling and IT. The fact that data is often still generated and kept in different systems - the famous "data silos" - accessible only by the corresponding department, make things worse. These systems must merge and get properly integrated, which requires controllers that have expertise in both IT and process optimization and are first and foremost future-oriented minds. This is best ensured with powerful analytical systems, be it descriptive, predictive or prescriptive analytics.
Make my day
Descriptive analysis is the classic reporting feature that has always been familiar to the controller and describes the status quo, the historical development of the company, based e.g. on target deviations. The resulting "diagnostic analysis" tries to track down causes from these results, such as e.g. “why are the costs above budget or why are sales going down?”
In future, the time gained through automation will be used for predictive analysis, and there is a plethora of software tools available that can detect patterns from the vast amounts of data or draw correlation between seemingly unrelated data. Such software is used not only to make forecast by means of algorithms, but to calculate the probability of occurrence of certain events and to determine the distribution of risk.
Today, computers can predict relatively accurately who will cancel a newspaper or magazine subscription or when someone will change the insurer. The supermarket chain Target allegedly analyzes the surfing behavior of female customers to find out if they are pregnant, so that they might be able to supply things like diapers or baby food. Life insurance companies supposedly found out that people who retire early usually die earlier, and airlines know that vegans miss their flights less often. In human resources, predictive analysis is often used to determine whether employees consider quitting their job. So, the art of predicting the future isn’t really an art, but the result of careful evaluation and proper interpretation of huge amounts of data.
Without asking any question
A typical online experience you are familiar with is when you look at certain content and the site recommends more content that other people have viewed. In online commerce, these recommended products can even be tracked on other sites, thanks to AdWords. In other areas, these recommendations are highlighted even more subtle and sometimes unnoticed by simply placing these items further up the page. What happens here is that a website uses a so-called "recommendation engine," a system that constantly learns from purchases and page views of previous visitors. This knowledge will be used in the next step to propose the next product that you are likely to be interested in, provided you follow the patterns of the other customers. The proposal is then your "best next offer" and the technique, used to figure this out based on your transactions, is often called "collaborative filtering".
However, to be able to make forecasts of future developments in the company, reliable data on probability of occurrence and risk distribution are required. Only then, reliable scenario calculations can get compiled and decisions made on the given, solid database. Unfortunately, financial planning (profit and loss account, balance sheet, and cash flow) and operational planning (sales, production, personnel) are in most companies often kept separate, which can lead to different expectations and significant loss of time.
Clearly, it is faster and much more efficient to link the business model directly with predictive analytics as it results in better predictions and clarifies the impact on business operations. Once that is done, the next logical step is self-evident, namely the automation of business decisions, which brings us to "prescriptive analytics" that can provide answers for price optimization, but also the fine-tuning of the marketing mix or fund management.
Wherever time-critical decisions have to be made, analytics come in handy, since they manage to act in fractions of a second. Analytics systems can provide timely recommendations for action, which are then discussed in management - and ultimately decided by people - saving precious time that’s literally worth a fortune of money.
By Daniela La Marca