BI implementations are like icebergs


Once again, this article is inspired by a question on one of the many groups. As these are only viewable to members, I’ll confine myself to a general link to the site. The subject was what might be the best BI tool for a particular project. I was enormously encouraged by the number of people who said that this was putting the cart before the horse.

My experience is that while having a snazzy BI tool (Cognos PowerPlay being the one I have most often used) in place can win you plaudits, this is only because it is sitting on top of a warehouse that embodies the business information that the organisation needs. The four keys to a successful BI implementation are: –

  1. Forming a deep understanding of the key business questions that need to be answered.
  2. Piecing together the various elements of corporate data that relate to these questions (assuming that they exist) and establishing how they are linked together.
  3. Working out how to transform the data to meet the questions.
  4. Managing the behavioural changes required to ensure that use of BI becomes pervasive.

The above steps (which are typically iterative) form the foundation of a BI initiative that actually adds value. If you have followed them assiduously, then whatever front-end tool you choose, your users will like what they see*. You can argue about the precise figures, but 80-90% of your overall development project will relate to the three steps above and only 10-20% to layering a BI tool on top of your data.

When I presented at the Informatica World 2005 in Washington, DC I tweaked Thomas Alva Edison’s famous aphorism to suggest that: –

Business intelligence is 10% presentation and 90% integration.

Taking my admitted cheesiness to one side, I stand by this observation.

* Before I am accused of being agnostic about different BI tools, I should stress that they are not all the same, some are better than others and it is worth putting in the effort to pick the right one for your organisation. Instead my point is that if you do not have the correct data and business foundations, it is irrelevant how good your BI tool is. Equally, if you do have these underpinnings in place, then most BI tools will at least be adequate.


18 thoughts on “BI implementations are like icebergs

  1. These were my initial comments on the thread:

    Your BI tool for “slice and dice” analysis is really only the icing on the cake. What counts is having a really well designed data warehouse that transforms the available data to meet pressing business requirements. Once you have a knowledge of what numbers are needed to run the organisation, what data is required to support these and how to transform the latter into the former, then your choice of front-end tool should just be around ease-of-use and robustness.

    If you are a member, you can view the discussion here.

  2. Hello, Peter!

    I jumped to your side through Linked In connection and really like your very common sence and no-nonsence approach which is clear and based on reality. Not much of this nowadays unfortunately.

    You a very right about this 90 and 10 division. One other addtion I would make – for BI it is absolutely necessary that someone at the top ( or even better ) many people at the top are truly interested in the resulting data and its quality. Interested to the extent that they will be looking at it, knowing ( at least in general) what it means and what goes into it, and using it for the business management.

    As I am now somewhere between CIOs and “real life” people I can clearly see what garbage and what mess can be happening on middle layer. Not possible to have a good data on top with those things down.
    And until someone wants to make sure that all layers are in order and not afraid to find the truth – no good BI is possible.
    Of course we can use BI just for report of invoices by customer and materials – but then should we actually call it buisness intelligence? Nothing new there – existed in thouthands years.

    So – will fully depend on the company and management desire to go into this difficult path and emerge with clarity.

  3. Galina,

    Thank you for your comments. You are of course right about the need for a commitment to BI at the top of an organisation. If that does not exist, then a key role is of BI advocate, selling the idea of BI and explaining the value that it can unlock. I have performed this role and have found that it is important to find allies and to build coalitions to help promote your ideas.

    On the issue of layers complexity and good data, it is interesting that I was recently thinking about this area and offered a slight twist on orthodoxy in this recent article.

    Thank you for your input.

  4. nice article – though I would add another dimension to the iceberg analogy: data quality. BI is useless without data and misleading without cleansed and realiable data.

This site uses Akismet to reduce spam. Learn how your comment data is processed.