By 2020 Big Data will consist of 40 zettaoctets of information (1021 octets), which is times 33 what it was in 2010, while research on the subject has increased twentyfold since 2009.
However, data has been collected since long, at a time when no one ever spoke about Big Data. But collecting data without clear purpose or strategy, as it had been done for a long time, is clearly not profitable. Fortunately, companies that collect data without knowing how to put it to good use are a rare minority. Big data has increasingly become a part of predictive marketing.
Predictive analysis consists in defining models through algorithms that use collected data. These models intend to anticipate and predict trends in consumer behavior. This allows identifying customers who are likely to become less involved with the brand and those whose needs evolve.
Prediction grants access to marketing’s Holy Grail: sending the right message to the right audience at the right time. An old marketing rule that is hard to apply without access to the right information our without analyzing it properly.
The ability to offer tailored advice at the precise moment a need is being born has significant impact on client satisfaction, and thus on their engagement towards the brand and the company’s financial results.
But prediction is only possible if collected data is good quality and varied enough. In order to have a good predictive model, the following steps must be followed:
Defining the problem(s) that the predictive model should solve.
In sum, predictive analysis allows better decision-making, risk avoidance (anti-churn for instance) and differentiation thanks to an optimized client experience.
Cathy O’Neil does not define herself as a Data Scientist but rather as a “Data Skeptic”. Mrs. O’Neil, author of “Weapons of Math Destruction” and holder of a Ph.D. in mathematics from Harvard University, warns in her book about the potential side effects of predictive models. According to her, there are many sources of errors, including the following:
Use cases come in large numbers and many forms, depending on the industry. The first limit is creativity, but possibilities offered by predictive analysis depend on the volume, quality and relevance of the collected information.
Second party data can play a key part in predictive analysis. A car maker may have a difficult time knowing that a consumer’s family is growing, except if a partner specializing in goods for newborns provides him with that kind of information. It is then in the car maker’s best interest to propose to its potential customer a personalized offer, adapted to their new situation, by serving personalized advertisements or modifying the website’s homepage according to the user’s profile, and highlighting a vehicle they might considering purchasing.
Predictive analysis also lets mobile operators identify users who constantly make more phone calls than those allowed by their mobile phone plan, allowing the companies to better target users with new adapted plans.
Predicting entails serving content adapted to each and every consumer’s particular situation, not only making efforts to sell a product or service among users for whom a clear need has surfaced. In this environment, well-performed predictive analysis represents a strong competitive advantage for companies that engage in such practice. It is becoming a key element in terms of decision-making.
According to a Forrester study, predictive models have become more precise and effective in recent years – even within firms that lack skills related to that field – thanks to new tools that are easier to use.
Several companies may believe that data analysis requires hiring a Data Scientist, but such a belief is not all that accurate and they should focus on training the human resources they currently have. According to Forrester, the larger the number of employees working with predictive analysis, the better. Including those lacking specific statistical skills.