Meet Annie. She’s running a small financial consulting firm and dedicate all her energy to customer service and business process optimization. She barely had time to integrate a CRM and follow up all clients through it. Over a year her consulting division has grown from 4 to 16 people. One day, looking through a schedule of meetings she found that there are more appointments for the last week of each month and too many cancellations for the first week of the month. Turned out that the majority of appointments are canceled right after potential clients get a paycheck.
Annie advised to reschedule all appointments with potential clients to the end of the month, when they feel more insecure with their current financial situation and thus are more willing to get advice from a consultant. Guess what? Annie’s firm has doubled the number of clients in three months.
It may sound counter intuitive but here’s how Big data and simple data processing can be applied even in SMB. All it takes is to acknowledge your data as a ‘corporate asset’. And what’s more, this asset is available even to the companies that are using simple customer relationship management system.
How Many Data is Already Big Data?
If “Data is New Oil” than it needs a refinery. Crude oil is always cheaper than gasoline and it should be refined to power car engine. This metaphor perfectly applies to Big Data and Machine learning engines.
Storing raw data is a bad idea. Only big enterprises can afford to store it in bulk. The Excel rows with missing values, inconsistent formatting, black spots, time gaps are those factors that will take extra time and money for pre-processing.
Keep in mind that Big Data is about quality, processing and discoverability potential. For instance, if your company has two subscription plans (A and B) and only 2 out of 1000 customers are using Plan B, then there is no point to look for dependencies between those columns. This data is not representative. In some cases, even the date field may be useless. In addition, there is no direct correspondence between the amount of data and performance of an analytical model. At one point the efficiency and accuracy of the model won’t increase just because you’ll get another SSD full of data.
It’s like kids when they go to school. When they learning math they learn some arithmetic but they don’t learn every possible combination of numbers to add together. They learn a few and then they learn a pattern of that. If you give a kid a few dozen of arithmetic problems and practice they will learn the patterns, but if you give them more they won’t get any better at the arithmetic.
Machine learning is just like your kids at school: it will learn what it can, but it comes to the point when giving them more data won’t make any better because it learned everything that can be learned.
Who Will Benefit from Gathering and Processing of Data
- Executives will become swift in making informed decisions.
- Business Analysts will get more insights to work with and find new business opportunities.
- Data Scientists save time on data acquisition and focus on actual studies.
- Your Customers will get products and services tailored to their needs.
Data Processing in Different Industries
Macy’s adjusts prices in near real-time for about 73 million product positions it has to sell. All prices are adjusted based on ongoing demand and inventory. The process is handled by an intelligent analytical algorithm developed using data Macy’s had previously gathered. The data includes out-of-stock rates, sell-through rates, price promotions, etc.
Nordstrom has made even deeper dive into Big Data. The company has collected user’s movement data by processing Wi-Fi signals in one of its area stores. The acquired data was later processed to reveal actual shopping trends and personalize ads.
Even if you’re gathering data and keeping track of visitors’ taste preferences, there is a problem to spot certain groups of customers. For instance, customers dining with children. Even when a visitor joins your loyalty program there is a small chance you’ll know that he/she has children.
Paytronics has used Hadoop data to identify these group of customers by their activity footprint. Those visitors are dining in groups early and ordering children’s entrees, Shirley Temples, or milk. From here your restaurant manager can target customers with specific discounts, free desserts and make dedicated dishes for children.
Fintech often use data to analyze applications to define a Default Probability rate. For instance, the word “business” is associated with higher probability of default. Thus, when someone is borrowing to fund a small business than it is more likely that this loan will go bad. On the other hand when someone is filing for consolidation of a credit card, then it is more likely to be a good loan.
The same method can be used to process employer description field in order to get Default Probability Score. For instance, giving a loan to a nonemployed person is not much worse than to loan someone from a bank. But if a person works in university or city then he/she has lower Default Probability.
Having a good analytical data underneath may be extremely helpful for your bank customer relationship team. Especially if you just rejected a loan application and your client needs to get an intelligent answer. The kind of answer that will be based on cold facts, rather than equivocal phrases.
If you’re doing portfolio management you may be want to match investment opportunities and the customers by their risk appetites, which can be determined by researching previous decision and investment portfolio customers already have.
As the world is getting more conscious about energy sources it becomes more clever and diverse in using energy resources. But it poses no problems, especially in the renewable energy industry.
For instance, solar panels can degrade over time. The data and machine learning algorithms built on top of it can identify which solar cells need to get maintenance work in the first place. In such particular case, those that are in the center of a solar plant, because they are not exposed to the wind flow and heat up more.
There is so much buzz with hype words like Artificial Intelligence, Deep Learning, Big Data, etc. Don’t fuss about this hype. What matters the most is that you can identify right business problems and data to solve it. That is what you need in order to implement Data Science and get new business models that improve to your bottom line.
Start from what data you have now, identify your business problems and then look for methods to solve those problems with data processing techniques. Data Science only works when business people ask the right questions, have right tools to get answers and take actions with the answers.
Got the Right Questions?