Subscribe to DSC Newsletter

Where Should We Place Statistics Within The Corporation?

I am curious about where people think statistics/advanced analytics should be placed in the organization. I have heard a great deal of talk about this lately.

Views: 353

Tags: Corporate, Organization

Comment

You need to be a member of AnalyticBridge to add comments!

Join AnalyticBridge

Comment by Paul Wilson on July 7, 2010 at 2:06pm
"Have a loose, informal organization of all the analytics folks that are sprinkled througout the business lines."

Anthony, I'd have to disagree with this statement, unless I didn't understand correctly what you're saying. Analytics needs support from the most senior levels of the management in order to be succesfully inplemented. "Lose" and "informal" structures are not likely to be receiving the support necessary as they will be on executive's back burner. However, if one of them is tasked to be the owner, leader and the champion of anaytical competency in the organization, analytics will be on his/her radar.

"This way the analytics professional is placed where the "context" is most important: the data source"

I think you're missing a point here. I think in the ideal case scenario, when the Analytics Center of Excellence is formed, there needs to be an initiative to reconcile all the IT systems that store the data from different business lines into one data mart, so all the information is available in one, single datasource. Then all information is available to anyone at any given time and noone has to be "close to the data source".

I agree with your statement in regards to consumers being treated like customers.
However, I don't anywhere see a statement where analysts should be reporting.

Without a clear reporting structure there will be no accountability and without accountability no one will care to push analytics forward.
Comment by Anthony Arrott on July 7, 2010 at 8:06am
The discussion here makes it seem like locating the analytics function in the organization is a choice between the lesser of two evils.

1. Report to the line-of-business that uses the statistics/analysis.
2. Report to a central analytics organization.

The problem with reporting to the line-of-business relates to the "telling truth to power" dilemma.

The problem with reporting to a central analytics organization relates to the "don't let the strategic plan interfere with strategic planning" dilemma.

(If someone wants to know what I am implying here, I can expand.)

I suggest there is a third way.

In my experience the analytics function works best when the "consumers" of the analytics are treated like "customers" not "bosses" by the analytics function. (This has an enormously beneficial effect on the unbiased integrity of the analytics results.)

But placing the analytics function in a central organization is typically horrendous at effectively allocating resources to disparate projects. Local department heads are much more efficient at allocating resources than central planners. (That's how the West won the Cold War).

So, here is my twist:

Have a loose, informal organization of all the analytics folks that are sprinkled througout the business lines. It acts like a support group to the isolated analytics professional; measures and rewards collaboration when one professional helps another; and provides quality control and professional standards enforcement for analytics activities throughout the business. Then place the analytics professionals in the line of business that is the source of data not the consumer of the analytics. They are then functionally part of the liaison of the data-source organization to the results-consuming organization. Examples of this include:

a. Manufacturing data-source for Finance results-consumer.
b. R&D and Tech Ops data-source for Marketing results-consumer.
c. Sales data-source for Executive results-consumer.

This way the analytics professional is placed where the "context" is most important: the data source - not the results consumer. Let the manager of the "data source" organization absorb the pressure to produce meaningful measurements of what is going on in their line-of-business. Since they do not have an advocacy agenda for the results (unlike the "results consumer" manager), this "data-source" manager is then primarily concerned with allocating limited resources to produce quality measurements in response to pressure from the various "results consumers" pestering for measurments.
Comment by Abd Errahmane Mouhab on July 3, 2010 at 10:49pm
There is a erroneous belief in business that analytics/Statistics is context related (just look jobs advertisements for analytics), yes a context experienced analyst will take less time in doing his analysis, while a non context experienced analyst has to do extra work in investigating the context before doing his analysis. which means extra time, although it is not bad due to the fact the analyst will be more careful in his investigation and thus analysis, while the context experienced might fall in the trap of "as usual analysis" and miss very important features of the data and thus his analysis will be of less value.

In addition to the above erroneous belief, managers tend to think having an analyst in their line will make decision process faster, and I also believe that "Power play" has a say, that is no outsider will affect their final decision. that is they can significantly affect the final recommendations of the analyst and thus the decision without any interference from higher management/business support.
Comment by Randy Bartlett on July 2, 2010 at 3:05pm
@ Paul Wilson
I read the paper and found it to be interesting. I especially liked the Analytical Decision Process Spider Diagram. I will look for the book.

The paper claims that the best analytics companies have a centralized group. We agree. As previously written, why do you think so many corporations lack a centralized statistics group?

I agree with you that there are advantages to having the statistics group report outside of the line-of-business. Why do you think so many line-of-business statistics groups report into the line-of-business head?
Comment by Randy Bartlett on July 1, 2010 at 9:29pm
@Abd Errahmane Mouha

I think this is a great structure. It is flexible enough for most businesses. The resources of the central group can be stronger technically and the line-of-business groups can focus on the local issues.

Yet, few corporations have this structure. Most have analytics only within the line-of-business. Why do you think this is?

Perhaps, you can validate some of my thoughts on why:
1. Analytics is grown organically. Someone in the line-of-business has an epiphany and they request resources. Done.
2. As previously written, (A) senior management wants to assign functional skills within each line of business rather than apportion them from a central group.
3. As previously written, (B) senior management thinks of analytics as context dependent.

What are your thoughts?
Comment by Paul Wilson on June 24, 2010 at 12:20pm
Randy,

Here are my opinions in regards to your questions:

"Here are two interesting questions: 1. Do businesses benefit from an ‘enterprise wide’ analytics group?"

Yes, they do. I'm a part of such a centralized analytics group and as someone who's reported directly to the line of the business before I can tell you there are many benefits.
First, the business is able to leverage a single view of the customer with built in insights drawn from many perspectives as opposed to many different groups drawing conclusions based on their own background only, ultimately risking confusion at the top of the decision making pyramid.
Second, blending analysts from a variety of backgrounds allows for better exchange of information and know-how, leading to better skill sets and employee engagement.

2. Should the statisticians supporting a line of business report into that line of business?"

I don't think they should. At least not initially. As the Analytics Center of Excellence evolves, more of it's members should be embedded on the lines of business, but at the same time they should be alligned to the central group in that new hybrid of centralized-decentralized approach.
That will ensure that everyone speaks the same language, consistency in definitions and making sure the efforts are not duplicated.

Here is a good paper on the analytical centers of excellence:

http://www.umsl.edu/~sauterv/DSS/pdf/BI/wp_6426.pdf

As well here is a book title with more examples from the real world:

Business Untelligence Competency Centers, A Team Approach to Maximizing Competitive Advantage by Miller, Brautigam and Gerlach
Comment by Abd Errahmane Mouhab on June 24, 2010 at 2:44am
Statisticians/Analysts should be at a corporate level that is part of "Decision making and business support" that takes care of strategic decisions and management changes, however business lines need staff to be trained in using statistics and performing simple analysis needed in the day to day operations and in addition will support the group of statisticians, given that they have more hands-on knowledge of the business line.
Comment by Randy Bartlett on June 23, 2010 at 8:55pm
I agree with both of you. If statistics is placed in a particular line of business, then it is not very ‘enterprise wide’ (shared, centralized group). Also, there are many benefits to be derived from a more independent analytics group.

Here are two interesting questions: 1. Do businesses benefit from an ‘enterprise wide’ analytics group? and 2. Should the statisticians supporting a line of business report into that line of business?

Many businesses have 1) separate analytics groups within each line of business (or groups thereof) that 2) report into that line of business. E.g., marketing analytics reporting into the marketing head, clinical statistics reporting into the clinical head, QC statisticians reporting into the manufacturing czar, and so forth. These businesses lack a central clearinghouse for solving statistical problems. Also, they lack independence from local political interests.

I think that leaders perceive statistics as a functional skill that is bound to the context. A) They want to distribute functional skills within each line of business rather than apportion them from a central group. B) They perceive statistical analyzes within clinical, marketing, and manufacturing as incompatible.

What do you think? Is there more to this?
Comment by Tejamoy Ghosh on June 21, 2010 at 4:19am
I agree with Paul.

Analytics can be a powerful device for the high-level decision makers when used optimally incoprorating all aspects of a business. For instance, if marketing wants to show good numbers by acquiring any customer it can without any heed to the risk (s)he brings in (for BFSI business) or the net LTV, we may have negative impact on business - whereas if risk and marketing works in tandem such situations can be avoided. I have one experience of such thing impacting business negatively - there the marketing department was responsible for/incetivised on the number of repsonses it generates through campaigns and not the ultimate conversions. As a result their campaigns were targeted only to maximise response without any knowledge of conversion propensity. And this was not understood by the senior management untill a convesion model was eventually employed in conjuction with a response model for targeting customers and roi immediately improved significantly. This could have been avoided if analytics department was not under marketing and was a direct strategic device under a C-level person looking after the overall business.
Comment by Paul Wilson on June 20, 2010 at 7:31pm
Depends on may factors, but if analytics is to be a shared, centralized group (i.e. not report to silos/individual line of business like, say CRM), there should be a separate stream with a VP, Analytics type of position available for optimum visibility ultimately reporting to COO or similar C-level executive. I think it'd be a mistake to have such a group to be stuck under IT, Finance or Business as I believe it should be positioned on the cusp of these 3.

On Data Science Central

© 2019   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service