Analytics for the rest of us
This blog was originally published here on September 17th, 2013.
How do you get started using analytics?
Stories about analytics and big data are everywhere. The coverage usually talks about clever ways to use non-traditional data sources to gain insight and edge out competitors. What is absent from typical coverage is how and why firms decide to start down this path in the first place. There are some basic steps to implement “analytics for the rest of us”.
1. What’s the problem you are trying to solve with analytics?
Often organizations start this process because something needs to be fixed…once in while an organization starts down this path with the sole focus of exploiting data. (That tends to be organizations with a lot of data and a staff of quants.) Most of us should stick to the problem we are trying to fix. Typical problems include fraud detection, marketing to the rising power of the individual consumer, managing customer churn, and operational improvement (to name just a few).
2. What data will you use to solve the problem?
Companies benefit by starting with their own data (e.g., sales databases, product databases, business systems…). You know that data; you trust that data. Once you have pulled your data together you can examine underlying trends and determine if you need additional third party data.
3. What modeling approach will you use to anticipate future changes?
Organizations succeed when they look for the least complex model to answer the problem. It does not have to be simple, but you do not want your management team to place their faith in a “black box”. This is an iterative process and must reflect the context of the business problem. There are 100s of models and algorithms available for statistical modeling and machine learning. Selecting your algorithms and modeling architecture is a tradeoff depending on the type of input data and the complexity of the issue.
4. How will you make analytics fit your organization (not the other way around)?
Top-down approaches (even knowledge driven ones) are hard to implement in a distributed work environment. Two techniques have worked well for us.
- Hold focus groups with staff and ask them what can make them more efficient. When you implement your analytics solution, tie it back their feedback.
- Ease staff into your approach gradually. Listen to their feedback and build on their experience to refine the analytics solution.
Remember, analytics is suppose to solve a problem – if nobody wants to use it, it will sit on a shelf. Success is when you can make decisions with clear information v. gut and guesswork.
If you want to start exploring predictive analytics, sign up for a FREE LIFT account now.
Not ready, learn more in our resource section.