27 April 2009

The value of an embedded Business Intelligence toolset

Why OEM Embed?

It's the old build versus buy debate.

My new Toyota came equipped with Bridgestone tyres and a battery from some manufacturer I did not know of. Without these components, my car would not be complete. However, Toyota did not have to open and fund a tyre manufacturing plant or a battery factory, they sourced these from specialist vendors, thus they got a final product to me that really was "best of breed".

In the same way, your software application can source the reporting & BI components from a specialist vendor who understands this space very well.

The typical dilemma faced by many software organizations is as follows:
We have just spent 15 - 40 person-years (or more) getting our application ready for the market, where data flows in and around the system exactly as designed (or redesigned), but now we need an analytic module and reporting module to round off the offering so that our customers can analyze their information in flexible, visually pleasing and easy ways.

Do we build it or source it in via OEM embedding it?

Questions that should be asked around this topic are usually:
  • How long will it take before we have a go-to-market ready offering?
  • How flexible is the organization we will be dealing with?
  • If we OEM embed a solution, can we re-badge the solution as our own? If so, what are the limitations?
  • What are the once-off integration costs and what are the ongoing costs?
  • How often do we receive upgrades?
  • What happens if we have specific enhancement requests?
  • How do we protect our investment?
  • How compatible is the solution with our core offering?
  • How will it be supported? What is the ramification of bugs in need of fixing and what priority will our customers receive for 2nd level support requests?
  • What is the skill ramp-up required to provide 1st level support to our customers?
  • What skills will be needed that we don't currently have?

At SeeMoreData, we are aware of these questions and have a solution that addresses these with the proof that our solution runs as an embedded OEM component at some of the world's largest companies. This in turn improves the overall value that we deliver to our OEM partners, as we have harnessed the knowledge, flexibility and other lessons learned over 18 person-years, into a few weeks of integration effort. This has produced some phenomenal outcomes for our OEM customer at a small fraction of the cost of developing an inhouse solution.

If you would like to know more, please visit us at seemoredata.com to get more information.

21 April 2009

What exactly is meant by 'Business Intelligence'

Ask 5 people what BI (Business Intelligence) means and you will probably get at least 6 different answers.

Today's post is short (and hopefully sweet) and describes the various forms that BI takes on.

In short, BI is either:
  • Strategic
  • Tactical
  • Operational

There is a new emerging category, which bridges all 3 types: Predictive

Strategic BI
This is where the Inmon-style massive data warehouse is involved and tries to answer the strategic business questions of trends over longer periods of time, where the business is headed and what the business should focus on in terms of overall business, where the money is coming from and going to, how the business interacts with its customers and the market at large.

This type of BI is slow as it can involve many years' worth of collected data, deliberate and the audience is generally executive management of the organisation.

Tactical BI

Smaller in size, generally restricted to a few (or often just 1) data sources and answers more immediate questions like
  • What is my best selling product per region for this quarter / month / week?
  • Who are my most efficient employees this week / day / month?
  • What is the ratio of open-closed service requests for the day / week / month / quarter
  • Who has more worked assigned than they can reasonably cope with?
  • What are my HR stats for the month/ week (OH&S, absenteeism, payroll processing, complaints, compliance, etc.)?
Many of these types of analytical views will form the content of various dashboards to be used from executives down to team leaders.

Operational BI
The target audience here is generally the regular employees, on whom the organization relies to actually get things done (as opposed to management who control, motivate, monitor & analyze), thus the amount and scope of decision making is limited.
Typical uses:
  • Daily closing figures report
  • Customer history / analysis reports
  • Employee analysis reports
  • The data will be restricted to a lower latency so that it deals closer to real-time events.
Usually, the data will be restricted to the user's area of responsibility (like team-based visibility in Siebel) or the data that the user is responsible for.

Some of these reports do not need to go against a data warehouse or data mart, thus they can be run against real-time datasources. Often these are predefined in their content and scope so you would (should) not find many power users doing analysis against real-time OLTP data sources. The datasources for these types of queries need to be protected with governors from runaway queries otherwise your normal online business operations WILL be affected. Been there as the DBA looking for the culprit (turned out to be the CFO using a point-and-click query generator against live production datasources, ouch!).

Using BI-based alerts (when a certain business event occurs, do something) falls somewhere between operational and tactical BI and starts to deliver on the promise of BI from way, way back, which is to deliver actionable, relevant, accurate information. An example of this is: When the power grid consumption nears peak, send a warning to the peak consumers of a pending blackout, via SMS text message, or email or some SOA-enabled process.

Predictive BI
The market is now ready (and pushing hard) to have the corporate information serve predictive needs. I am not going to get into the nuances of predictive models here but typical examples of this might be found on web sites such as travelocity.com (beautiful example of predictive analysis in my opinion) whereby the system is combining real-time transactional events and trying to predict consumer behaviour based on past trends and delivering the results in real-time and sub-second turnaround.

Last words
There has been some near-religious wars fought online as to what is Business Intelligence and Business Analytics, whereby reports and dashboards were categorized as BI and any sort of trending, exception alerts and prediction was categorized as Business Analytics. This in my opinion seems like a marketing effort to produce something with their time, regardless of relevance. A rose by any other name is still a rose.

Hopefully this has served to clear up in everyday language, what we BI people are on about.

Have an awesome week!

07 April 2009

The importance of good communication with End Users

I've said it before and I will say it again: "The success of BI starts and ends with the end users".

Last week I had the privilege of running a brief education stint with the senior management of one of my favourite customers, to educate them on the efficiencies they would now have in getting relevant information with our BI implementation, which they could invoke directly from inside Siebel with the click of a single button.

After a lot of lively, meaningful discussion, we realized we probably would not achieve world peace this week, but we also found out that although the BI implementation was developed exactly to specification and worked flawlessly with all its drilldown, slice & dice functionality, it did not provide value because the wrong things were being measured, and in the wrong way. Another reason why I like this customer so much is that instead of a witch-hunt for who interpreted their requirements into a specification, we took the opportunity we had with all the affected business leaders in the same room and nutted out exactly what was required. This is a level of maturity that is refreshing to find, especially in the midst of this global financial turmoil.

In BI, often end users don't know what they want until you show them something, this is a very common reality and is rooted in the spiral nature of BI projects, as opposed to the waterfall "design everything up-front" methodology of typical commercial OLTP type projects. This statement did have a placating effect on my audience last week and allowed us to return to the issue and opportunity at hand.

We flipcharted and whiteboarded (are those even verbs?) all the thoughts and considerations to be taken into effect and came up with a way forward, in a way we could reuse many of the existing components, but now we had a lot more support, since this was the first time we had all the top levels of management in a single room for a 3 hour period, and we finally got inside their minds and found out the answer to WPMO (what pisses me off?) and WKMA (what keeps me awake?) questions. Now we could design a solution that would provide value to them.

This is also proof that it is better to do the right things than to do the things right.

If BI projects are to succeed in delivering value, to enhance the lives of the audience they are targeted towards, we HAVE to "spend a day in the lives of our target users".

Put another way, if my car's user interface (dashboard) showed me the compression in each of the cylinders, or the amps being drawn from / to the battery from the alternator and other electronics, I probably would not pay the slightest attention to these measures as I drive towards a traffic cop parked on the side of the road. At that time, what matters most to me is that I know the speed limit of where I am driving and I know the speed of my vehicle in relation to it (as in, stay under the limit). The intelligence I get from my car needs to be relevant to my conditions at the time. Effective business intelligence needs to begin with an approach as simplistic as this.

Have you had situations where you have gone to deliver the final 'ta-daa! presentation' only to find that what you have delivered means about as much to your audience as a symphony being delivered in sign language? Share your experiences via feedback & comments.