Analytics is basically using data for different applications. Applications are often called use cases: the most difficult part with getting started with analytics, from a strategic perspective, is to identify relevant use cases that relate to a specific business need or to a future business opportunity.
In the analytical world, a use case is a demarcated area where data can be used to rebuild an analysis model, i.e. populated with new data as it is collected, for example, by sensors, from customer data, etc. As new data is collected, the model is continuously refined. If the refinement is done automatically, it is called machine learning.
Examples of use cases for different companies
Company A - Personalized online shopping
The company wants to create a personalized online shopping experience to increase conversion. A use case for analytics can then be:
How can we predict what every customer visiting the website is most likely to be interested in?
Company B - Optimize customer service staffing
The company wants to optimize customer service staffing to save costs and increase customer satisfaction. A use case for analytics can then be:
How can we predict how much calls come into the customer service due to different external factors?
Company C - Optimize inventory
The company wants to optimize its inventory to save costs and reduce the loss of sales opportunities. A use case for analytics can then be:
How can we predict how much demand there will be for our different products at a given time?
Company D - Reduce maintenance costs
The company wants to reduce its maintenance costs by getting early warnings of potential equipment failure. A use case for analytics can then be:
How can we predict which components are likely to develop a fault based on known behavior patterns shown before breaking apart?
Real-time data allows for real-time insights
Previously, historical data has been used to produce forecasts that form the basis for traditional planning. Now it is possible to work with real-time data that runs against different predictive models, which in turn can be linked to a dashboard or against different triggers.
This increases the accuracy of the models drastically. A trigger may be that if a signal exceeds a certain value, an alert should be sent to a technician who can act directly, instead of making maintenance rounds at fixed time intervals.
Another trigger may be to highlight a particular offer to a specific customer, based on his/her earlier searches on a website instead of giving all customers the same generic information.
A third trigger may be information sent to a staff manager about how many employee hours are needed for the next period of customer service, based on real-time demand instead of following regular forecasts based on historical data.
Normally, there are two distinct phases in analytics:
1) Prototype phase (or proof-of-concept)
The goal of this phase is to create predictive model(s) based on historical data to validate if it is possible to solve the identified problem(s)
There is almost no barrier to starting the prototype phase. However, there are some common obstacles that limit the ability of a company to go to a production phase, such as lack of sufficiently high-resolution data or lack of an easy way to retrieve and restore data between data warehouse and analytics tools.
2) Production phase
This phase is when the predictive model is put into a production environment, which usually involves activating triggers, but may also consist of continuously presenting insights into a dashboard.
But before taking the step towards production, one needs to have overcome the prototype phase and truly achieved a good “buy-in” from the rest of the organization.
In many companies we work with management teams who want to see a business case at this stage before they are willing to make a bigger investment.
The prototypes luckily provide a good indication of the business potential of a full-scale implementation and any system changes that will be required to reach it.
Do you want to get better and faster insights by finding the right use case for your business or need help to drive through the initial prototype phase? If so, please contact us by clicking the button
You probably know someone who also should read this? Don't forget to share!