As technology continues to empower our ability to conduct analytics with “Big Data”, the “Internet of Things” has arisen as an area where devices themselves capture and transmit data albeit machine-level type data. Let’s discuss some of these devices.
Telematics is revolutionizing the auto insurance industry as devices are placed in cars which actually record the behavior of the driver. For example, how fast am I going, or how quickly do I brake, etc. are all types of behavior that can be captured by the sensors in these devices. In Canada, Desjardins Insurance , a Quebec-based insurance company, has been the most active proponent of this type of technology within the Canadian marketplace. The use of discounted premiums has been one method to promote the use of these devices by its customer base. It is still early days in the use of this technology within the auto insurance industry. But think about the benefits. The ability to collect actual driver behavior is simply going to provide insurance with better information on how to price a given policy based on predicted claim behavior.
Weather-related data also represents a new phenomenon in its use within predictive analytics. Data related to temperature, wind speed, sun levels, and precipitation is collected continuously thereby allowing analysts to construct pattern or trend-related variables. For example, how does change in sun and precipitation impact sales of sporting clothes? How does temperature impact mall sales versus retail one-unit stores? Most recently, one organization has been using weather-related data to better predict call volume within its customer service area.
The challenge for data scientists, though, is again the volume of data which is being collected continuously. At what time intervals should the data be collected? Of course, this will depend on the business challenge or business problem we are trying to solve. If variability of sales by week is important to the organization, then how sales varies within a week is important. For forecasting purposes, weather data can be extremely important since it can be argued that weather does have an effect on consumer behavior. The frequency of its collection, though, is based on how far ahead we are basing our forecast on. For example, forecasting a week out would imply that we gather forecasted data that is seven days old. If we are looking at time-series analysis where the record of interest is a prior event or outcome rather than a consumer, we would expect that weather data would have some impact on a future event or outcome. But in cross-sectional type analysis, we build predictive analytics solutions that are at the customer level rather than at an event/outcome level. Under this scenario, practitioners might ask whether or not weather trends impact customers differently. The initial assumption used by practitioners is that the impact of weather exhibits the same behavior sensitivity across all customers. But then again as practitioners, we would look at the data to determine indeed if this type of weather sensitivity exhibits different patterns amongst different groups of customers.
Mobile technology have increased the amount of data exponentially as cellular devices collect data even without the user exhibiting any overt behavior other than the movement of the cell phone based on the location of the individual. For example, if a mobile user’s WIFI is turned on, the strength of the signal can be assessed based on the distance from a given router and the customer’s location within the establishment. What exactly does this mean? The signal itself creates data and information relating to the distance between the router and the phone. It also creates data pertaining to when the signal first appeared on the router(entry) and when it disappears(exit). Think of the valuable information here. If you are a restaurant or retailer, you now have data pertaining to visit behavior utilizing mobile technology. Recency of last visit, frequency of last visit, and duration of visit can all be captured due to mobile WIFI data. Besides the traditional recency and frequency metrics, advanced analytics can also be conducted to develop predictive models. Retail outlets and specific locations can be grouped into high, medium, and low value groups based on the mobile visit behavior within that location. Mobile technology has simply extended the ability to apply data science and advanced analytics to data that was simply not accessible in the past.
Beacon technology provides even more access to data as the customer moves through a given store. For example, the technology identifies the products that are in close proximity as the customer moves through the store. Previous customer analytics may have indicated that a customer has a certain predisposition towards cookies. As the customer moves through that section, specific offers related to cookies could be promoted through the customer’s phone.
In the home, this digital explosion and more important access to the data is now in the process of being extended to such devices such as refrigerators, oven, stoves,etc. Similar to what we observed with mobile technology, information relating to use as well as time of day provide valuable information regarding the use of these devices to the analytics practitioner. Think of the analytics possibilities as the practitioner begins to integrate all this information. For example, the use of household devices within the household can be integrated with the property insurance policyholder information such as claims, house demographics, coverage, etc. Establishing a new rating structure in theory would in all likelihood include variables related to the use of these devices that are most likely to impact the likelihood of a claim and its resultant amount(loss cost). In fact, specific insurance products and offers could be developed for these products based on the usage of these devices. In the financial services sector, perhaps greater use of these digital devices might be used to target those households that are more receptive to lending offers which are related to these devices.
TV viewing is already being used by telcos to better understand behavior as well as building tools that predict a given customer’s telco behavior. If we can integrate data related to TV Viewing and the use of a refrigerator, patterns might emerge that demonstrate increased refrigerator usage with increased TV usage on Sunday afternoons in the fall. One obvious and likely hypothesis is that increased snacks and food are being consumed while watching sports, particularly football on a Sunday afternoon. If I am a grocer, I would definitely want to direct my advertising towards snack-related messages that are now targeted to these now-identified households.
The above just represent a few examples that can leverage this type of data. The Internet of Things , being one of many developments in our Big Data world, is simply another source of data for the data scientist. Yet, the traditional approach used to be to “extract as much data as possible” whereby data reduction techniques would then be applied in filtering out the data to more meaningful data. Although data reduction techniques are still applied as part of the data science exercise, more attention is now devoted in extracting the right data. There is simply too much data in the digital world where the overwhelming portion of this data is really irrelevant for a given business problem. It is the business problem that now defines a process in how we extract the “right” data. But what is the “right” data. This is where the “domain” knowledge of the data scientist is used to understand the real business problem. Along with their deep mathematical and technical knowledge, the ability to quickly acquire domain knowledge is becoming an increasingly more important skillset for the data scientist. Those data scientists that have the deep technical and mathematical skills complemented by an appetite to acquire domain knowledge will be in tremendous demand. Technical skills can be taught as universities and colleges will support the demand for these specific skills. But the ultimate challenge in leveraging information from the Internet of Things(IofT) era will be to identify the “hybrid” individuals that have both the technical acumen and the ability to effectively apply these skills within a given business setting.
Richard Boire, B.Sc. (McGill), MBA (Concordia), is the founding partner at the Boire Filler Group, a nationally recognized expert in the database and data analytical industry and is among the top experts in this field in Canada, with unique expertise and background experience.
Mr. Boire’s mathematical and technical expertise is complimented by experience working at and with clients who work in the B2C and B2B environments. He previously worked at and with Clients such as: Reader’s Digest, American Express, Loyalty Group, and Petro-Canada among many to establish his top notch credentials.
After 12 years of progressive data mining and analytical experience, Mr. Boire established his own consulting company – Boire Direct Marketing in 1994. He writes numerous articles for industry publications, is a well-sought after speaker on data mining, and works closely with the Canadian Marketing Association on a number of areas including Education and the Database and Technology councils. He is currently the Chair of Predictive Analytics World Toronto.