Keys to Analytics Success
In a flat world, where traditional sources of competitive advantage are rapidly eroding, analytics holds out the promise of being one of the last points of differentiation. However, as with any strategy, competing on Analytics requires certain ingredients and processes as pre-requisites. Otherwise, there is a good chance that you will spend tons of $$$ and effort with nothing to show for.
From our experience of helping our clients, here are some recommendations to profit from predictive analytocs.
1. Think Big, Start Small. People tend to think of predictive analytics as just another initiative on the tactical side of the business. Viewing it this way leads to underestimating its impact and the data and process change management effort required to profit from it. To maximize the value from analytics initiatives, organizations need to focus on their distinctive capabilities, ensure that solutions are business driven, and start small
Focus on your distinctive capabilities. Given the breadth of analytical possibilities, you should prioritize your analytics initiatives on enhancing your chosen distinctive capabilities - resources that provide the basis for your competitive advantage. If your distinctive capability lies in your ability to attract and keep the most profitable customers, you should focus on analytics initiatives that help you attract and retain the right customers and maximize the value of your customer relationships
Business-driven initiatives. Profiting from insights generated by predictive analytics may require your organization to adapt your business processes to take advantages of these insights. Hence, these initiatives succeed only when they are aligned to organization's goals and have the top management's buy-in. Business needs, rather than other considerations, should drive analytics initiatives, not the other way
Do not boil the ocean. At the risk of stressing the obvious, we advise you to start small, go for low hanging fruits, and balance faster time to market with solution complexity (data availability, process changes needed, etc.)
2. Define the Business Case. Identify a list of business pain points that can be addressed through analytics
a) Define the objectives. Define the problem in clear terms, using key performance indicators. Be SMART about your objectives – specific, measurable, achievable and realistic, and time-bound. For example, if your issue is the high attrition rate for a key segment of your subscribers, define the current and targeted retention rate, quantify the gap, and the time by which the gap is proposed to be closed
b) Define your solution approach. Clearly describe how the proposed analytics solution is expected to address the gap. Outline the data requirements and the process changes that might be needed to ensure that insights from analytics are used for making decisions in the process
c) Define the costs and benefits. Create multiple scenarios of success for the analytics initiatives – conservative, moderate and aggressive. For each of these scenarios, identify the expected lift from the solution, the data needed, the process changes needed, the complexity of the analytics models, and attach a cost to each of these elements, and compare it against the expected dollar benefits. If possible, perform a Net Present Value (NPV) analysis to identify if the project is worth devoting time and effort
d) Prioritize. Based on the cost/benefit considerations and the time to market, prioritize your initiatives. Assign resources and responsibilities to coordinate solution definition, development and implementation
3. Ensure you have the right resources. Ensure that you have access to the right resources - data, tools, and people
Data. Data is the foundation of predictive analytics. Ideally, more the data, the better. However, what is more important is the quality of the data. Key dimensions that you typically need include (in decreasing order of importance) customer transaction history, customer demographics, customer interaction and attitudinal data
Tools. To maximize productivity, invest on tools that are widely used (so it is easier to get people and build a team), that have a large variety of algorithms, and which can scale to manage your data. When it comes to tools there are a variety of options – a later post will address some of these options
People. When it comes to people, make sure that you invest heavily in smart analysts who have the business acumen, and who can intelligently tie together all the terabytes of customer behavior, demographics and interaction data into bite-sized chunks of insights. Otherwise, you run the risk of drowning in data while starving for insights. As a rule of thumb, I recommend that you follow the 10/90 rule suggested by Avinash Kaushik (applied to Web Analytics, it’s probably applicable to other domains as well)
4. View Analytics as a Discipline. Predictive analytics is no silver bullet, and profiting from it requires a disciplined approach. We recommend a disciplined approach with a clear plan of action, and follow-through to ensure execution excellence
Define a plan and a methodology. Create a detailed plan, with milestones, resources and responsibilities. The plan should ideally have four phases: data preparation, data exploration, model development and validation, and solution implementation. Define a methodology for project execution based on industry standards, such as CRISP-DM. and break down the project into doable tasks that can be assigned to people
Follow-through. Ensure that you execute well. Look for deviations from desired tolerances, and rigorously work to close them. Conduct team meetings to question and follow-through on the tasks, to ensure that schedule and deliverables are under control. Validate the insights that are generated from these models with the business managers based on their experience and knowledge
5. Expect incremental progress. According to experts in the industry, organizations that go for the gold rush with predictive analytics are bound to become frustrated from the weight of massive, unrealistic expectations. The reality is that predictive analytics provides benefits in little percentages, rather than unexpected, multi-million dollar discoveries.
Be prepared to persist with delivering incremental improvements to business processes or customer insights that generate accumulating value over the medium term.
These recommendations capture some of our thinking on how best to use predictive analytics. Do you agree or disagree? Would love to hear from you.