Companies have always been interested in getting new insights using the data. But in past the access to data was restricted and so were the capabilities of utilising the data. Structured data was the prime source of information and BI was the prime analytic mechanism. Companies didn't focus on getting data but focused more on efficient data warehousing mechanisms and effective business objects to create reports. But in last few years’ access to data have become very easy for a lot of companies and have left them with headache of how to best utilise the information. Not a bad headache to have I must say.

Let's take a closer look at this problem. Traditional Analytic systems were level 2 processes. Data was gather and was organised, relationships with the data were created and then it was perfectly stored in different databases and files (level 1). Data from different sources was selected and was copied to a single database or file (Level 2) which was used by BI tools to create beautiful reports for intelligent people. But now level 2 is deemed to be too late and slow. We are talking about level 0. Using data as soon as it is available. Even before it can be structured, truncated, normalise etc.

But when it comes to analytic the question is whether companies should ride the trend or should they wait until the distinction between hype and reality clears out? But it is certain that companies cannot give a blind eye to this trend. First thing that most of the companies would consider is Cost. Cost benefit analysis is required before taking on any new venture. There should be "Horses for Courses" and there is no standard solution for developing analytical capabilities. Companies have to make difficult decisions and position themselves at what they want to achieve if they want to transform or even build their analytical capabilities.

Some of the important factors companies need to consider are

1         Scope of Analytical capabilities

Defining scope of the system is as important as the building the system itself. Failure to outline the scope can greatly affect the results. Even from Cost-Benefit perspective it is wise to build analytic system where benefits can be realised. It is debatable but if scope to benefit analysis is done the graph may turn out to be a bell curve. There are two aspects of defining the scope.

1.1        Segmentation by size

• Process
• Department
• Company
• Organisation

This can directly influence the cost and benefits. If company wants to up the sales then no point building system for whole company.

1.2        Segmentation by target

• Individual
• Aggregate

If data is analysed for individual entities then the quality of data becomes vital. Effectiveness of processes like analysis and implementation totally relies on quality of data. Best example of this can be selective marketing or suggesting based on consumer behaviour. But when it comes to group, large amount of data will mostly exhibit the group characteristics so quality of data is important but not as important as it is in case of individual consumers. But for a group, analysis and implementation becomes tougher and to be effective these systems have to be robust. For e.g. government agency wanting to improve the public transport time table.

If customers are to be targeted individually then it is important that company has excellent screening systems. Implementation and prediction can turn out to be easier if correct data is collected. For e.g. selective marketing. But when it comes to group data collection and screening is not as important as for an individual customers but implementing or predicting a solution can be a headache.

2         Time frame decision for Analytic

The complexity of the purpose of analytic may determine the complexity of the systems. It is imperative to establish the time-frames for which company wants insights. The complexity of the system and capabilities of resources required increase significantly for complex purposes.

• Past: - Analyse historic data to understand what happened in the past. Most obvious example can be performance evaluation of individuals and businesses.
• Present: - It comprises of during historic analysis as well as trying to understanding what is happening currently. Trend analysis can be one of the examples.
• Future: - Predictive and prescriptive analytic systems

3         Risk acceptance

Another decision point for companies can be the nature of risk they are willing to take and the strength of insights companies are looking for. Company’s willingness to work with following information can determine the extent of analytic capabilities required:

• Facts: - If the basis of decision making for any company are facts then it is not hard to determine what kind of analytic system they want. The focus for such company should be gathering and structuring data properly as the analysis process and coming up with results would be easier in this case. Not only the systems but analytic expertise requirement may also be not as demanding as it will be in other cases.
• Strong Hypothesis: - Companies who want to gain competitive edge through insights but are reluctant to expose themselves to high risk and uncertainty should look at developing analytic system which is based on strong hypothesis/correlations. For e.g., if someone buys a birthday cake there is a high chance that he/she might buy birthday candles as well. But these insights might not always be able to provide competitive edge as competitors might also be deduce these information though they might be very useful for internal decision making process.
• Loose Hypothesis: - "Small insights can have big impact on businesses". Companies who thrive to be unique and have high risk absorption capabilities (usually due to size and diversification) try get as many insights from data as they can. But such systems require extra validation and simulation systems on top of their analytic system to evaluate the strength of insights and validate the risk profile. The infamous ice-cream sales and shark attack example is a loose hypothesis.

4         Mechanism

Objective of performing analytic plays a key role in defining what needs to be done. There can be different flavours to this as well.

• Specific question: - Answering the burning question decides what the analytic process should be and how it should be built. For e.g. company looking to understand why sale of their brand if low even when demand for the product is high or what impacts consumer buying behaviour etc. The end objective is clear and it is easier to differentiate what is useful and what's not. The system should be fine-tuned to focus on a particular aspect based on the question.
• Exploration: - Companies can also build their system in such a way that they are focusing on getting something out of the Blue. There is no specific question but area of analysis is defined. For e.g. Analysing point of sale data can through interesting insights about various products, consumer segments, locations, stores etc.
• Rewriting history: - Another objective can be doing what if analysis to take future decision. Rewriting history to understand the consequences if alternative decisions were taken in past. Such analytic process can be very useful if the company is looking to be more efficient in future. The intermediate events and their sequences are important and form the basis for future decision making.

Before investing in Data Analytic companies to decide on what should be the scope of analytic systems, decide on the time-frames over which they want to do analytic, companies also will have to understand the risk and accuracy of data they are willing to play with and finally they should establish the mechanism through which they want to carry out Analytic.