1. Create a strong culture for the availability and use of data
A strong analytics culture is foundational for getting the most out of big data and analytics. How can you create this culture? Let’s take an example. Think of yourself as the Vice President of Sales for a company. You have called several of your sales managers together to discuss the coming quarter and expected shortfall in revenue. You ask for input on how to optimize the available resources to drive better results. The sales managers start sharing ideas including reassigning resources and new accounts to go after. You respond, “What data do you have to support that? Is this gut-feel supported by facts?” If you are at IBM you would ask the sales manager, “Have you followed the recommendations in the Coverage Optimization for Profitability analytics model?” This example illustrates that encouraging employees to use data to support their assertions can get the whole team thinking about data-driven decision making. Changing culture is never easy – expecting decisions to be fact-based is a good start.
2. Build a team with the right skills
Solving a business challenge using big data and analytics requires a collaborative and multi-disciplined team. A person with expertise in the business is essential. Skill in the business processes is particularly valuable. Also key is an IT person with data expertise in the business. Finally, you need an experienced data scientist or analytics practitioner to understand and prepare the data and to develop and evaluate the analytics model.
3. Estimate the ROI for your project
The Return on Investment (ROI) of an analytics project can be estimated through a disciplined methodology that identifies key value drivers for the business area and assesses costs (for example, estimates of time, talent, software costs, training costs, Subject Matter Expert (SME) time). Value driver trees are an effective way to determine the benefits the project can drive. Identify the key value drivers for the process and quantify the impact that an improvement in that process could drive. Existing research and SME input can guide the value tree development and quantification of both hard and soft benefits. Once you understand those value levers you can estimate the expected return. It is important not to underestimate the training costs of getting people to use the analytics within the business process. Change is hard and some end users will require more knowledge of the tool than others.
4. Start with the data you have
Waiting for perfect data can take time, causing you to miss an opportunity for action. Use analytics techniques that can fill in gaps in imperfect data so that business value can be realized. Think about the jigsaw puzzle analogy that IBM Fellow Jeff Jonas uses to illustrate this point. Each piece of the puzzle represents a transaction or data. As you put more of the puzzle pieces together it becomes easier to complete the puzzle; in fact the last piece of is the easiest to put in. Even without the last few pieces of the puzzle you are likely to have enough context to infer what the missing pieces would be like. Analytics can be used to fill in those gaps like the last few pieces of the puzzle so that business value can be derived sooner.
5. Deliver results iteratively
Rather than planning to develop a large analytics solution in several years, with one delivery at the end, put a stake in the ground to drive progress and get results. Iterative development of analytics solutions has several advantages. First, it allows you to obtain feedback from your stakeholders and target user group early. This feedback may cause you to have to make adjustments, which are much easier to make early than later. As soon as you have a working prototype, you can use it to create buy-in from your target users. Second, developing iteratively decreases the time to value.
6. Engage target users early
Deploying a new analytics solution widely is key to increasing the amount of value realized by an enterprise. Every target user who fails to use the analytics solution reduces the value realized. One of the best ways to help your target users buy into your new solution is to expose them to an early prototype both to allow them to see what the prototype can do and to get their feedback. Fortunately, those following tip 5 will have an early prototype. By way of example, a team developing a new quality detection solution needed a way show why yet another quality detection solution was needed. They asked target users to provide historical data containing their gnarliest problems so that this data could be fed into the prototype. The prototype team was not told what the problem was. The prototype identified some of these historical problems 6 weeks earlier than the tradition quality detection solution did. These results caused the target users to look forward to the new solution.
7. Use proven analytics solutions even if you do not understand the underlying analytics technology
Analytics is a broad field encompassing a range of techniques for extracting insight from data. Many of the techniques have been verified both through rigorous mathematical methods and through extensive evaluation and have been made available for use as robust software packages or services. Understanding when and how to use the methods is essential; understanding exactly how the methods work is not. Just as most users of digital technology, ranging from cameras to music players to cell phone, don’t understand the details of how sound or light is converted into bits to be stored or transmitted, and then converted back to sound or pixels to be enjoyed, it is not necessary for users of analytic methods to understand how the analytic algorithms process numerical data to find correlations or patterns, or explore a set of possible solutions to pick the best one. It is, however, important to understand when the use of a method is appropriate and how to interpret the output of the method. As in any technical field, in analytics there will be a small number of people who create new methods and a larger group of people who create new applications of methods (both old and new), and a very large group of people who use the applications to create enterprise or personal value.
8. Take action from insight to realize value
Analytics is often described as the process of extracting insight from data. While insight is far better than hindsight or no-sight, insight alone does not create value for an enterprise. It is insight, coupled with actions inspired or influenced by that insight, which can create value. Of particular value is insight that allows you to understand the likely result of specific actions. Then you can select from among possible actions the one most likely to produce the desired result. Further, using advanced analytic methods such as mathematical optimization, you can use this type of insight to understand and evaluate complex multi-part decisions, such as the allocation of scarce resources among activities. Finally, it is important to consider the action(s) taken, the expected result, and the actual result as additional data. This data should be further analyzed to gain additional insight that can be used to influence future decisions. This final step makes analytics an adaptive and learning process, which can continually improve outcomes.
9. Measure to gauge success
The primary reason for running a big data and analytics project is to achieve a business outcome. In order to know if a business outcome was achieved, the outcome must be measured. As an example, a proactive retention project used predictive analytics to identify high-value employees with a propensity to leave the company. Human Resources (HR) knew the attrition rate of high-value employees in the previous year. After predicting which high-value employees were likely to leave, HR increased salaries or paid retention bonuses. Attrition of high-value employees for the current year decreased. Measurements allow you to evaluate progress and decide if the outcome is acceptable or if additional measures should be taken to improve the outcome.
10. Share big data and analytics assets across business units to drive additional value
In addition to being used to support individual decision processes, analytics can be used to facilitate collaboration and learning within an enterprise. The simple act of agreeing on the source and meaning of data can bring together process teams from different organizations within an enterprise. Discussion of how the different data elements are used, whether for measurement and reporting or for decision making, can lead to greater understanding of the overall drivers of enterprise performance. Further, identifying the various decisions that are made based on this data in the end-to-end execution of business can lead to both better alignment of unit measurements and the identification of opportunities for collaboration. In the transformation of IBM’s supply chain, analytics revealed opportunities to share both physical and information resources between units to better serve customers. It also enabled collaborative decision making between IBM and its distributors.