Liveblogging CSI Pune Lecture: Applications of Business Intelligence
I am liveblogging CSI Pune‘s lecture on Applications of Business Intelligence by Narender C.V. of SAS R&D India. These are quick and dirty notes of the lecture – not intended to be a well organized article, but hopefully it gives you enough of a flavor for the area to get you interested and excited enough to check it out on google and wikipedia.
The amount of data is doubling every 11 months. And we have easier and easier access to all this data from all over the world. The problem is making sense of all this data. The amount of time at our disposal remains the same. So we have to use sophisticated software and algorithms to figure out how to use this data to improve business and efficiency. That is Business Intelligence (BI).
This talk is the second in a series of talks on BI. PuneTech covered the first talk which gave an overview of BI and data warehousing. This lecture focuses on who uses BI and why. A major portion of this talk will be a bunch of examples of use of BI in real companies. So on to the examples:
Example 1: Getting a better grip on Reality (i.e. Seeing problems earlier)
First case study will focus on using BI to simply get a good picture of the situation as it exists. Seeing Reality. Last year, US based companies paid $28 billion in servicing warranties or recalls. This is money you don’t really want to spend. Biggest problem in this is identifying these problems as early as possible. Seeing reality early. Typically, an issue first appears. A little while later, the issue becomes visible to the company, and it is prioritized. Later it is “defined” and decisions taken by the decision makers. Finally the issue is resolved, and money paid out. A study by SAS shows that the “detect” part of this cycle takes about 90 days, the prioritize part takes 20 days, and the define part takes 75 days. That’s a total of 185 days to fix the problem.
A business intelligence system helps to reduce each phase of that sequence because of better data gathering and statistical analysis. This results in 27 days detection, 5 days, prioritization and 46 days to prioritize, for a total of 78 days. This is a huge improvement, and each day saved results in money saved.
How is this done? First simple reports: defects per thousand, per product. Dashboard with easy to see defect reports. Then a library of reports that various people in the company can use easily to see and analyze defects and warranty claims. Then a statistical analysis engine to detect “emerging issues”. Use algorithms that can detect, from early trends, issues that are likely to become “big” later on. Text mining and analysis to read unstructured reports of service technicians and being able to determine, simply by looking at the keywords, which product or part or defect was the cause of that particular incident. And there are other analytics, like forecasting and trend analysis that are used. Bottomline? Shanghai GM was able to reduce detection and definition time by 70%, resulting in reduction of costs by 34%. Which is pretty cool for simply running a bunch of mathematical algorithms.
Example 2: Manage and Align Resources to Strategy
Everybody agrees that it is important for a company to have a strategy. And that everyone should understand and execute according to that strategy. Obvious?
This is a reality based on a survey: Only 5% of the workforce of a large company understand the company strategy. Only 25% of the managers were incentivized based on the strategy. 60% of organizations do not link budgets to the strategy. 86% of executive teams spend less than one hour per month discussing strategy.
How can BI help in this case?
It is possible to define objectives for each person/team in the company. Then it is possible to define how this objective can/should be measured. Then BI software can be used to capture and analyze this data, and figure out how everybody is contributing to the end objectives of the business.
Example 3: Retail Optimization
The problem to be solved. Need to stock the exact quantity that people are going to buy. Stock too much and you lose money on unsold items. Order too little and you get out-of-stock situations and lose potential profits. Need to be able to forecast demand. Optimize which sizes/assortments to stock. All of you must have an experience of going to a shop, liking an item, and not having that available in your size. Sale lost. Profit lost. Can this loss be reduced?
Use BI for this. In case study, a department store sent the same mix of different sizes to all stores. SAS did clustering of stores, to create 7 different sub-groups that have different size mixes for each sub-group of stores.
Example 4: Personalized, real-time marketing
Take the example of marketing. Consider a traditional marketing mail sent from a company. Customers hate that and the success rate is a pathetic 3% or so. That’s just stupid, but exists when there is no alternative. Better is event based marketing. When you do something, it triggers a marketing push from the company. This is often convenient for the customer, and has a 20% success rate. But the best is customer initiated interaction which has a 40% success rate.
Note that as you go down that list, it gets more difficult to quickly, in real time, determine what marketing message exactly to push to the customer. If you call a pizza delivery place and they point out that that they have a buy-one-get-one-free offer, it might or might not be interesting for you. Better would be an offer focused specifically on your needs. Use BI to analyze individual customers and forecast their needs and then tailor the offer for you. An offer you cannot refuse.
Another example. Customer puts digital camera in online shopping cart. The online shopping software contacts the BI system for offers to push to customer. It looks at customer history. Figures out that customer is non-tech savvy customer who buys high-end products. Also, customer’s demographic information is consulted, and finally some accessories are suggested. Since this is very specific recommendation, this can result in a high chance of being accepted. This significantly increases profit on this transaction.
Example 5: Understanding Customers
Mobile company, simplistic view: Customer is leaving. Offer them a lower value plan. The might or might not leave. BI gives you better tools. Cost is not the only thing to play with. Understand why people are leaving, and also understand the effect of them leaving on your business. (Sometimes it might be best to let them leave.) And based on this, determine the best course of action – what / how much to offer them.
First, use predictive analysis to get an estimate of how much profit you are going to make from a customer over the course of next N years based on the data you have gathered about them so far. Use this figure, the “customer value”, to drive decisions on how much effort to expend on trying to get this customer to stay. Forget the low value customers, and focus on the high value ones!
Another possibility. If you have marketing money to spend on giving offers to some customers. Let us say there are 3 different kinds of offers. Use BI analysis to figure out which offers to send to which customers, based on customer value, and also chances of customer accepting that offer. This optimizes the use of the “offer” dollars.