Predictive Analysis - Enabling us to See the Future
Posted in Operations & IT Articles, Total Reads: 2344
, Published on 31 October 2014
Web analytics can be used to track basic website metrics. But how do we go beyond that and forge long-term and solid relations with our customers? We use Predictive Analytics.
If there has been one thing that has radically transformed the very nature and basis of marketing in digital space in the recent times, it is Predictive Analytics. Predictive analytics is a business intelligence technology that harnesses a sophisticated mathematical model to be used as the basis of making predictions. The model sources its input from the multitude of data collected on each individual.
This model helps to provide insight to the marketers on the characteristics and the behavior of the buyers. It differs from business intelligence on the basis of the richness of the data and the rigor of the mathematical modeling. It can be used to optimize marketing strategies and website trends to better leverage the consumer behavior. It calculates a predictive score for each of the user and accordingly deduces a path of action for the particular user. Predictive analysis consists of six stages as follows:
Figure 1: Stages of Predictive Analytics
Predictive analysis is a dazzling concept yet with an intimidating and daunting perspective. To foresee and predict what your customer might buy may seem unimaginable or a complex feat with calculations from statistics, probability and mathematics. But this number crunching helps in higher returns as it rather focuses on separating the segment of customers who might behave differently compared to the rest.
Gaining perspectives from customers,understanding their needs and turning them into products is the basic keynote of any marketing strategy. Thorough analysis of behavioral data helps companies make informed decisions on the marketing strategies they need to adopt for their product. Predictive Analytics not only aids in logical decision making but also enables the companies to adopt the best practices using data driven consumer insights. Companies leveraging it have access to a rich and focussed database that puts them in an advantageous position.
Figure 3: Access to marketing is crucial to success
Relative advantages of using Predictive Analytics for companies
The company that uses Predictive Analytics mostly has advantages in two marketing metrics as compared to the companies that do not use it. These are as follows:
1) Incremental sales lift from a marketing campaign: The consumer experience almost a two-fold boost in the sales lift of a marketing campaign propelled by Predictive Analytics.
2) Click-through rate: The click-through rate from a marketing campaign for a predictive analytics user is significantly higher than a non-user.
Figure 4: Positive Impacts of Predictive Analytics
The preponents of Predictive Analytics are empowered to carry out the following:
1) To test the effectiveness of a marketing campaign.
2) Discovering new insights from the human resources dedicated to data mining.
3) To estimate the value of a customer in terms of lifetime customer value as well as profitability.
Usage of Predictive Analytics in STP of a product
Firms investing in Predictive Analytics can harness the data collected on the behavior of the customer to segment the audience and correspondingly target offers and personalised messages for them.It can be used to target high-end customers and offer special services for them.Going beyond superficial trends,it helps identify correlation and patterns. It enables them to keep abreast with the channel proliferation and the changing consumer behavior.
Figure 5: Behavioral data used to segment,target and position a product
Predictive analysis and Big Data:
Predictive analysis along with Big Data helps in understanding the customer's insights and predict customer's preferences. This combinaion helps in shift of view from historical view to a more futuristic perspective approach.
Companies or stores can analyze the data from loyalty points and predict the prefernces of the customer and recommend their future purchases. Websites can also offer a customized interface experience based on the previous searches made by the customer.
Predictive analytics can be applied to any type of data irrespective of whether it is in the past,in the present or in the future. Predictive Analytics enables the firms to achieve their objectives faster by turning the dormant data into opportunities.
Figure 5: Harnessing Big Data for Predictive Analytics
Predictive Analytics in Fraud Detection
With the help of predictive analysis we can plug in loop holes by accurately finding fraudulent transactions irrespective of them being online or offline. They also help in identifying false or fraudulent insurance claims. It also helps in speedy disposition of genuine claims. Predictive analysis uses a combination of text analytics and sentiment analytics to scan a claim report which generally extends to a multiple pages. Predictive analysis along with big data helps in sifting and sorting through the unstructured data to detect possible fraudlent insurance claims. Usually claim adjusters write lengthy reports when they are investigating claims. These long reports contain small key words which are the hidden clues to find out genuiness of claims. The computer system which has the predictive analytics software works on business rules and spots these words as evidence. Another most important thing to observe is that fraudulent claimers usually tend to story over a period of time. Such errors are easily picked up by the system.
Figure 6: Steps in Insurance fraud detection using Predictive Analytics
According to Infinity insurance company which uses this predictive analytics, the claims fraud system which filters out fraudulent claims has an increased success rate of 88% from 50%. It also reduced the timed required for the investigation of such fraudulent claims up to 95%.
For using this predictive analytics in insurance claims we have to follow a ten step approach. We perform a SWOT and then build a dedicated management team. Once these both are in place we decide whether there is need for buying or building the required skill set. Then we integrate all the data and frame relevant business rules. With these rules in place we set thresholds for anomaly detection. Then we use predictive modeling and SNA and build an integrated management system leveraging on them. Such system can run efficiently and give an increased rate of return on investment. The data has to be periodically updated from additional sources to be ever efficient.
Figure 7: Fraud detection using Predictive Analytics
Predictive Analytics in Amazon’s ‘Anticipatory Shipping’
Amazon has recently obtained a patent for its pathbreaking concept of “Anticipatory Shipping”.It is the system of delievring products to the customers before they even place the order.It plans to ship the products which its customers might buy preemptively. The company decides the goods based on the customers previous searches, wish lists and on the duration of time that a customer spends over a particular article online. If this concept is executed in an efficient manner,then it will propel Predictive Analytics to an alltogether new level and will enable Amazon to greatly expand its wide base of loyal consumers.
The underlying rationale of this system is that a worry-free customer is always loyal,contented and would not switch on to the competitors. By ensuring that they can deliver products before the customers run out of them ,adds to the credibility and trust in the brand of Amazon . Most of the shopping is tiresome and mundane and when Amazon takes over that part, it absolves the customer of his responsiblities and hence promotes goodwill for the brand. The reason for the mind boggling popularity of Amazon has been that it remembers what we bought and when had we bought it;and recommends us products to buy based on our shopping history.
Amazon’s system involves the use of two computers: One is used to identify the general shipping location and a second is the one that waits for the delivery address to be finalized.
Figure 8: Anticipatory Shipping by Amazon using Predictive Analytics
When a customer finally places the order, the item would already be halfway to its destination. This will reduce the delivery time to less than a day. It can be reduced even further if the location of the customer is closer to the Amazon distribution center. The benefits of predicting customers’ orders could help to boost sales and potentially reduce shipping, inventory and supply chain costs.This will further strengthen its position as the number one retailer in the world.
Figure 9: Targeting customers using Predictive Analytics
Predictive analytics is not limited to digital space alone.It is taking over the public sector too.A CSIRO big data project in Australia has leveraged Predictive Analytics to predict the emergency admissions in the Gold Coast Hospital on any day of the year with 93% accuracy. The concept of Predictive Analytics was even used in Barack Obama 2012 Presidential Campaign to influence voters (https://www.youtube.com/watch?v=5vct0-e9xxc)
Therefore, when it comes to Predictive Analytics, one prediction is absolutely clear: there is a bright future in using Big Data to predict what is coming next.
This article has been authored by Mohd Zeeshan and Maanasa Mallela from IIFT Kolkata
1) Aberdeen Group-Predictive Analytics for Sales and Marketing