I was speaking to a COO the other day about the drive to deliver BI projects in short timescales. "That’s fine," he said. "But I’ve never known IT to deliver a project in six weeks."
This underlined to me one of the key challenges that is now facing information projects - the ever shortening timelines in which the business is expecting to see value from a BI investment. If we look back to five years ago, there was the general agreement that insight was something worth waiting for, and the business was prepared to accept at least nine months before the BI system was up and running. Nowadays, with the multiple channels available and the transient nature of the information itself, the requirement is measured in weeks, if not days.
In order to address this challenge there needs to be recognition that the business requirement needs to be split into two parts:
- Insight - where the business has a hypothesis as to where value could be realised and where it needs information to support a quantitative assessment.
- Industrialisation - whilst the more agile approach is able to realise value early there are still instances where a waterfall approach is required e.g. in the case of standard reports or where there are downstream applications. In these instances there is a need to follow a more structured development and testing approach especially where regression and change management to the business is required.
It is the former of these that is driving such aggressive timelines. However whilst this information is needed quickly, in the majority of cases it is only needed on a one off basis, which as a result reduces, if not removes the requirement to industrialise it within a data warehouse. Instead, the key requirement is the ability to visualise this data quickly and to be able to extract the insight efficiently.
Minimising the need to industrialise the solution means that the lead times for delivering insight are significantly reduced and the critical path becomes the quality of the data. Given the single time use of this information, this requirement can be addressed quickly, where it is needed, using data profiling and/or quality frameworks on an ad hoc basis.
The challenge that businesses face, especially those that are not so mature in their analytic capability, is that they often have a limited understanding as to how the data can be fully utilised or which metrics are required to track process changes. Therefore, the advantage of this rapid visualisation and working in an iterative manner is that they can confirm whether their initial hypothesis is robust. If it is, they can then define the optimal set of metrics that can track the benefit realisation. It is at this point the following can occur:
- Linkage to value - there is a clear understanding of the value that is expected to be realised to underline the success of the delivery within the business.
- Industrialise within the data warehouse - as there is a clear understanding of the data, metrics and reports required and the degree of data cleansing needed, the effort / cost of incorporating into the data warehouse is a lot clearer and can be more efficiently implemented.
With respect to the timelines, process change typically needs three to four months to be implemented and therefore this can run alongside the industrialisation process. This allows the business to identify, change and then monitor process changes within a significantly shorter timeframe. There is also a side benefit with this value-led approach in the adoption of the solution. Where the business can clearly see value being delivered, there will be a strong pull to maximise the use of the system. Hence why my mantra is, "first visualise the value and then industrialise it."
by Will Gatehouse, Process & Information Management, Accenture