April 14, 2008
Anarchy and centralization
How do you decide what kind of organization you want when you’re talking BI - what belongs in the Business Intelligence Competency Center (BICC), and what doesn’t?What are the drivers behind establishing the BICC, and what level of centralization vs. anarchy should you choose?
One of the more important features is to support the amount of anarchy there may be in the business lines. The Business Intelligence system should not prevent or hamper the efforts in the individual business lines to leverage the knowledge that supports the decisions made on a daily basis.
The BI system should act as a repository for knowledge, for facts about the company, and if the business lines need to expand on those fact to make them fit to their reality, they should be able to - but how much anarchy should you allow?
In my oppinion theres anarchy in two dimension - in the processes/innovation and in the data (heavily inspired by Ross, Weill and Robertson - EA as strategy).
What you need to find out is - What does my company look like? The reason this is important is that it will define the optimum roadmap and endresult for the construction of a BICC (Business Intelligence Competency Center), what kind of roles you should have in it, and where you should anchor it.
Separation
Within the separation space there is no room set for a BICC, and each subset of the company will need to build up their own BICC. There will be a marginally small gain to creating a centralized BICC holding the governance and doing the development, since everybody have the opportunity and ability to break with the governance, and will often be motivated to do so. The BICC will at most consist of experience sharing groups.
The reasons for having a seperated strategy within the company are obvious - it will put the focus on agile exit and entry of new companies, and will be due to an organicly growing company with a fast turnover of subsidiaries.
Imitation
Within the imitation quadrant there’s multiple ways of creating your BICC, there will however be certain inhibitors on the roles and responsibilities that you can take into the BICC. Within the Imitation quadrant there will be certain key processes that needs monitoring across the company, and here you will face the problem with aligning the data. This gives you the challenge of building a BICC that is very knowledgeable on the sources within those key processes, while at the same time allowing for a distributed, decentralized, BICC, either functionally split according to source systems, or functionally split according to business units.
The key role of the BICC will be to provide knowledge on the data in the source systems, and provide the ability to align those source data across functions or across business units.
Total alignmentIn the total alignment quadrant the setup is easy - the systems should all be integrated, and the meassures should be the same across the entire company. All you need to do is define the key KPI’s, and setup the system to report those with all the relevant processes.
The roles in the BICC will be easily defined, and the BICC will be able to take care of a major chunk of the daily work in the different knowledge driven departments (like market analytics and finance).
Collocation
In the collocation sphere there is differences in processes across the companies, but the core processes are all contained within the same datasphere, or the interdependencies are described within that sphere - perhaps in a Master Data Management setup.
In here you have the possibility to create the entire BICC setup, but you will have to take the differences in processes into account, and constantly push for aligning the businesses - if you create the BICC right, then you will gradually be able to move the business towards total alignment.
Posted by Peter Møllebjerg Andersen at 2:45 AM | Comments (0)
April 10, 2008
Orthogonality
Lately I’ve been thinking a lot about orthogonality in KPI’s.
Perhaps I should specify - when I talk about orthogonality, I’m pretty much talking in vectorspace, meaning that if the KPI’s are perfectly orthogonal they do not contain any form of information about eachother. Or in other words - independent KPI’s.
The trouble with interdependent KPI’s is that it will give an overestimation of the challenges in the company - imagine a situation where you have 4 KPI’s, one on customer satisfaction, one on revenue, one on orders quality and one on employee satisfaction. All are red.
Which one would you focus on?
Well - most companies would implement an revenue improvement - by going more aggresively in the market, after that they would ensure that their share of voice would grow, and become morepositive by branding themselves better, while getting the employee satisfaction up by having more socialization internally in the company - e.g. giving an office party. Finally, off course, they would get the orders quality up by putting out a memo stating that employees need to be more thorough in their job when putting orders into the system.
What should be realised is that the 4 KPI’s might be interdependent.
What actually happens is that:
1.The quality of orders data is low
2.Due to low quality data - orders do not get delivered on time - if at all
3.Customers get dissapointed, and move their business elsewhere
4.People feel that they cannot perform as well as they would like to, while listening to a lot of customers complain that they work at a bad company - their satisfaction falls
5.Revenue falls (naturally) stemming from the fact that the customers do not reorder
The right action to take would have been to impose a new governance on orders quality, or change systems such that the quality would go up.
Off course you have the same problem with weighted importance within all KPI’s, and you could argue that this is where the good management comes into play - finding out what KPI’s to focus on, but in many cases of missing orthogonality you will not be able to realise that you actually have a interdependence problem, and will try to isolate the problem and solve on a point basis.
Another form of interdependency is when we meassure the same process in different ways - such that the process is overrepresented in the KPI map. This kind of overrepresentation will bias the attention towards areas that it will be less efficient to focus on, and will ensure that you direct your focus towards a less than optimal area.
So how do you avoid creating these kinds of problems for your company? I will write it here when I have the definite answer.
Posted by Peter Møllebjerg Andersen at 5:30 AM | Comments (0)
