What happened to the what if analysis?
When I first started working in the Business Intelligence industry one of the key concepts was ‘What If Analysis’. One of the drivers of using an OLAP/Decision Support or Business Intelligence system was to be able to do predictive analytics. What If Analysis is undoubtedly a big value add for business, however, my recent experiences with clients have made me think that not every one is in a position to do this yet.
At Rittman Mead we tend to look a Business Intelligence through a series of levels of maturity, these typically go from being able to produce operational reports from a single system with no control and governance, to building an analytic system where data is integrated, dimensions are conformed, metrics are clearly defined, control and governance is applied to the data and power users can do sophisticated analysis of the data and make use of features such as drilling down, drilling across, aggregation and filtering. Reports built in this system could then be arranged on dashboards to provide user end user access to the information.
Users get value by being able to compare data from different business processes, being able to aggregate data and being able to look at data over time. There can be a number of steps between the basic operational reporting and this level of analytic reporting, but this level basically describes the core features of OBIEE’s Answers and Dashboards products.
We find most organisation are at a more basic level of maturity. In order to get traction from Business Intelligence projects, we have to engage at that organisation’s existing level of reporting maturity, there is no point proposing some wonderful high brow analytics system if the organisation is still doing most of its reporting using spreadsheets, I don’t believe you will get traction. You have to grow the maturity of the organisation with the development of the Business Intelligence landscape.
So this leads me to my original question of where the what if analysis has gone, how should organisations move to this? From a product point of view there are some obvious candidates: Oracle Data Mining (ODM), Essbase, Oracle OLAP and Crystal Ball. The questions are: what is the use case for implementing these? Where can an organisation get business value using this type of analysis? How could you construct a ROI case for these?
A meeting with a colleague identified the following scenario which I quite liked and thought made a clear and strong case. The starting point is at a level of maturity where an organisation had a good analytic system. The use case is the organisation wants to change their existing business model to either target another customer segment, or add another value proposition or product.
The most basic way of doing this would be to create a simple spreadsheet that read a bit like a Profit & Loss statement and had a figure for expected revenue and some figures for costs, these estimates would be based on the experience of the spreadsheet author and would at most contain worst case, expected and best case values. They could potentially turn out to be accurate, however they are not really going to carry any weight, especially if significant investment is going to be required to implement this change. What is required is a more qualified approach to modelling this change.
Using the products mentioned above one approach to solving this problem could be to start with ODM. The organisation could use some of the algorithms in ODM to start to predict measures like revenue, based on customers likelihood to churn and sensitivity to price. These calculated measures would be based on historic data the organisation had and would give a much more reliable view or model of the revenue for the new or revised business model.
However so far we have only got one version of the model and it may not feature all the measures we need, for example it may be possible to calculate some revenue figures, however cost may need to be manually input.The next step would be to use Essbase. If the model comprising of the associated dimensions and facts was then loaded to Essbase, along with the ODM calculated data the organisation could then publish this to several users and allow them to input data for measures that couldn’t be calculated and to create different scenarios for the model, based on different interpretations of how successful the model is, for example worst case, best case and most likely. In additional profit would be able to be determined for these scenarios as all the relevant measures would be present – either calculated by the analytics tool, ODM or user input.
Coming back to the question of investment, whether it was either internal or external, the organisation would now have a much more compelling case to persuade a potential investor. However I think it is possible to argue that there is still a missing element from this which is risk. The investor would be well within their rights to ask what risk was associated with their investment. This is where the final piece of the jigsaw comes in, Crystal Ball.
Crystal Ball allows users to run a series of (Monte Carlo) simulations based on high and low water mark parameters that are run against the model, the end result is that a level of confidence can be ascertained about the likelihood of an event. Crystal Ball works as an Excel plug-in, and as such can also work in conjunction with Essbase. The Essbase data can be loaded into Excel through its plug-in, then have the statistical analysis from Crystal Ball performed on it, then potentially read back into Essbase as another scenario.
The end result would be a business model both in real terms and loaded into Essbase that an organisation could go to an investor with understanding of revenue, profit and level of confidence (risk) and be in a much stronger position to secure the finance they need. As a further benefit the organisation may be able to negotiate a significantly lower interest rate on any investment as they can demonstrate low risk.
I think this story sets out a clear picture of how these technologies can be used together to support a genuine business scenario and it is certainly something I will be investigating internally with some business initiatives we have coming up over the remainder of this year.