Organizations are increasing investments in data management & analytics. Business managers are keen to move beyond traditional BI to achieve analytics led business transformation and benefits. Here are the 10 steps that will help business and analytics professionals get maximum leverage out of their analytics initiatives.
1. Define enterprise view of analytics
Incomplete analytics vision leads to – unmanaged and expensive analytics growth, waste of resources and doesn’t give maximum possible benefits.
Steps to defining enterprise view of analytics –
- Identify analytics requirements across 3 types of decision making – strategic, tactical & operational. This is done across business functions.
- Set up the vision of how the pieces fit together.
- Prioritize analytics initiatives
- Allocate people, technical infrastructure & funds
2. Innovation funds are ideal to get analytics started
In the initial phase, allocate funds from innovation budget rather than waiting for well-defined business case. Use this for (1) experiments and (2) establishing the value. Use small teams, open source technologies & cloud subscription. Once you establish business case and get business funding, continue to allocate a small innovation budget for new initiatives.
3. Adopt a managed self-service analytics approach
Business users needs speed and more control in their decision making process. IT managed centralized BI applications cannot fulfill this. Over a period of time analytical silos emerge to satisfy number/data savvy managers. You need enterprise wide self-service analytics platforms that gives more users more analytics capability. Key characteristics of such platforms should be –
- Easy to deploy and customize for multiple users
- Business user should be able to configure & re-configure solutions easily
- Should have advanced analytics capabilities and not just descriptive & diagnostic capabilities to generate accurate & trusted results
- Should be scalable for big data
- Ideally accessible on web
4. Use industry specific solutions
Each industry has its own set of business problems and corresponding analytics solutions. Generic analytics solution needs more time & resources to adapt to industry specific business processes & data. Look for self-service analytics platforms with industry specific solutions to maximize leverage from existing resources.
5. Define success metrics
Success for analytics initiatives needs to be measured & tracked for improvements in 3 aspects – data & information management processes, business decision making processes & shareholder value.
- Data & information management processes can be measured using metrics like – % of sources in analytical datamart, % of sources considered for single views, count of managers using a report for decision making, ratio of report visits vs ad-hoc requests raised, timeliness of reports, data audit reports, data load performance reports, etc.
- Business decision making processes needs to be measured separately for strategic (for e.g. – value of options), tactical (for e.g.- forecast accuracy, retention rate) & operational (for e.g.- sales conversion rate, % first time right, resource wasted due to rework) decision making.
- Stakeholder value can be measured using metrics like – customer spend, brand share gain, cost of raw materials, marketing spends, etc.
6. Leverage cloud & open source
Existing RDBMS/ analytics capability might be insufficient. Cloud and open source can help you get started with your analytics initiatives by providing flexibility and scale without upfront investments & vendor commitments. Such an ecosystem can be leveraged during the initial (or experimentation) phase to build a few analytics solutions, demonstrate their value and create a community of analytics users within the organization.
7. Build a comprehensive analytics data mart
While enterprise data marts are created with an objective to get ‘data in’ and support data management, analytical data marts are created with an objective to get ‘insights out’ and support analytics initiative.
Analytical data marts should contain data from 5 sources –
- Transaction data that exists in enterprise data warehouse
- Industry specific data bought from aggregators like Nielsen, D&B, IRI, etc.
- Open data freely available from government sites
- Social media data from blog, tweets, yammer, Facebook & LinkedIn.
- Data present in docs or emails and not part of data warehouse, but important for analysis and insight generation
The power of analytical data marts lies in its single views. Single views should be created for each analysis level like – customer, product, promotion, campaign, store, etc. Multiple single views capturing different time spans can also be built.
Robust data validation processes should be built to ensure authenticity of data.
8. Data governance
As more and more data and analytics become available to more and more members of the organizations, the risk of privacy violation & data misuse also increases. Assign team with right tools that allows them to regularly monitor 3 aspects – data access, data quality & data perishability.
Data hoarding in the hope of future analytics need also needs to be managed.
9. Move towards advanced analytics
Look for self-service environments that moves beyond data discovery capabilities towards more advanced predictive & prescriptive analytics. While several organizations have struggled to demonstrate the value of business analytics, there are a significant number of organization where advanced analytics has helped them attain business transformation & game-changing success.
Organizations need to move from Descriptive analytics (What has happened?) to Diagnostic analytics (Why something has happened?) to Predictive analytics (What will happen in future?) to Prescriptive analytics (What are my options and what should I do?). As organization matures in analytics journey, human role in decision making will reduce and thus allowing them to act with agility and responsiveness is required for quickly changing business needs. In fact, barring a few strategic decisions, all tactical and operational decision making should be done at a rate faster than human capability.
10. Embrace agile development
Basis enterprise vision and priority pick up analytics initiatives to proceed incrementally with agility.
Traditional BI or IT solutions follow a structured approach with clearly defined business problems. Analytics initiatives can stifle in such environments where answers are explored for ambiguous or non-existent questions.
Agile development (where requirements and solutions evolve through collaborative efforts of IT, business managers & data scientists) is best suited for analytics initiatives especially in experimentation phase.
Steps to managing an analytics initiative –
- Understand – Analytics done in silos with just data and statistics is never actionable. Make the analytics team interact with the business, spend time in the business processes, go to the stores, etc. so that they realize the business problem. Analytics project usually fail during implementation primarily because analytics team never had an understanding of these challenges when they set out to build the solution.
- Explore – Internalize the data. Explore unknown patterns in the data. Link different sources & aspects of data to realize the complete story.
- Build – Use advanced techniques to build the solution. Try multiple approaches or solution frameworks. Overlay business rules on top of statistical outputs to make the results actionable. To realize actionable insights always try to answer the question – “So what?”
- Refine – Use business feedback to make the results and insights more actionable. Realize implementation challenges if any.
- Implement – To implement an analytics solution additional layers needs to be added. In some cases, actionability is achieved through a set of dashboards, in other cases it can be a complex process for e.g. campaign lists getting generated basis micro-segments or predictive score or recommendation engine, they are then getting dispatched and tracked for improved customer behaviour.
- Re-configure & Implement – Markets change very fast. Customer behaviour is also dynamic. Analytics solutions should be built in a way that re-configuration is quick and implementable. For e.g. changes to an existing customer segmentation structure should be easy, recommendation engine for a new category of business should be quick.
The key aspect of any analytics solution is its iterative nature of build. Usual BI or IT solutions or even traditional analytics solutions are typically build for long usage without changes. Whereas analytics solutions need re-configurations every time there is a change in market or customer behaviour. To retain the actionability of analytics insights, it is essential that we use self-service platforms that allows us to perform re-configurations in quick time and implement them with ease.