As a CEO who focuses on developing software purpose-built for consumer goods companies, increasing data leverage is a topic that I frequently discuss with both users and potential users. Given our application focus areas which include trade promotion, retail execution, and supply chain execution, we manage a wealth of insightful data within our customer’s environments. The key is the elevate that data into information which is actionable as part of an intuitive workflow within the application UI.
To deliver on data leverage for our clients, we must understand our client’s views on data external to their companies and how this syndicated data either plays or might play a role in their decision-making today. As you can imagine, this discussion typically expands across multiple data types —from government records to GS1 data, from e-commerce to social media and everything in between. One could refer to this expanded data set as ‘big data.’ However, for the most part there is more noise than signal in this data.
In finding the signal amongst the noise, the question becomes which data should we use? How do I analyze the data I choose? Where is the payback for all this data? All these questions boil down to one important question: How do I make my data actionable?
Alex Ring, VP Product Management of TPM Retail & FS, and I presented on this topic at our recent AFS 2016 User Conference. As a first step we must determine how expensive the problem is that we are trying to solve. Now expense in this sense can have many dimension, but the application of technology should be metered by the expected return on investment. You can start by answering:
- What is the opportunity payback of fixing the problem?
- How expensive is the solution to your problem?
- Where is your smart data (confirmed by an ROI analysis)?
Once you have answered these questions, you need to find the data. The challenge with noisy data is it requires you to sift through and find the nuggets that will deliver value. You will want to look for data that is interpretable, relevant and novel. Focus on data that is tied to processes and is easily accessible to update when you need it. Basically use your business process workflows as the filter on that value of your data.
Next, you need to apply intelligence. There are three types of analytics you can apply to your data:
- Descriptive analytics: Descriptive statistics that summarize the data.
- Predictive analytics: A statistical model that uses existing data to predict data that we don’t have.
- Prescriptive analytics: A deeper analysis that based on the analytical results can offer advice on how to proceed based on possible actions. This level can also lead to scoring as well as gamification where that makes sense as part of the future model.
Prescriptive analytics by definition are actionable enabling us to solve problems before they occur, and can also provide the closed loop feedback we’ve always wanted in terms of whether the prescribed action actually generated the expected results. We’ll revisit analytics in a future column. For more information visit slideshare.