Predicting system failures to improve customer satisfaction

Wednesday, May 14, 2014

Companies like Amazon, Google, and Netflix have done an amazing job of providing a great customer experience. For example, when you use Google’s search engine, it quickly figures out if you are just researching a topic or planning to a buy a product/service. Accordingly, it tailors content to show you relevant ads or chooses not to display any. Similarly, when you buy a product on Amazon, it displays other products that may be of interest to you. Netflix offers suggestions of movies you may enjoy based on your viewing behavior. How do they do this? It’s not magic, but sometimes you wonder how they could be so accurate. The short answer is that these features are enabled by some cutting edge analytics.

You may wonder what Amazon, Google or Netflix’s analytical capabilities have to do with what we do at GLASSBEAM – i.e. helping companies become proactive and predictive in avoiding costly system failures. The answer is simple – Product manufacturers can apply the same principles of analytics to not only predict system failures but also prevent them by taking proactive actions.

Let’s first discuss a typical system failure scenario. When a system fails in the field, the customer is the first to notice and usually he/she contacts the product manufacturer. The manufacturer will then need to decide whether to ship a replacement part/system or send a field technician to fix the issue. This method of reactively managing system failures is not only costly, but also negatively impacts customer satisfaction. The good news is that all this can be avoided.

The two critical things required to prevent this scenario are data and the ability to analyze data. Most systems shipped today are instrumented to generate a lot of product operational data i.e. machine data. However, building applications to analyze this data is not an easy task since it requires sophisticated parsing, cleansing, extraction and loading processes to prepare the data for deeper analytics. In addition to machine data, one also needs to import data from other application systems like CRM, Bug Database, Knowledge Base etc., so that the resulting data model for analysis is rich across multiple dimensions. Many people underestimate this "staging" process where data has to be "prepared" for basic reporting and deeper analytics. Glassbeam’s analytics solution excels in this part of the process. Our core IP of SPL, our parsing language and SCALAR, our IoT platform enable conversion of incoming files or streams of machine data into actionable information. Compared to HOMEGROWN SOLUTIONS THAT BALLOON INTO MULTI-YEAR PROJECTS that take up valuable engineering resources and expensive infrastructure, our SaaS solutions can be deployed at a fraction of the cost and time.

In my next blog, I will address the second part of the solution i.e. the ability to analyze the data. Stay tuned for more….