Sign up | Login with →

Comments by Meta S. Brown Subscribe

On Netflix Asked the Wrong Analytics Question

Bill, there's another fundamental problem with the way the question was asked.

Scoring was based on the root meansquared error. That's a measure suited to estimation of ratio or interval measures. Movie ratings are ordinal measures - they have order, but should not be treated as numeric values suitable for arithmetic use.

Treating ordinal measures as intervals is a common practice. I'm not a purist, so if it works, it works. In most cases, though, people who do this do not actually have evidence that it produces useful results for them. Analytical methods designed for use with ordinal data exist, and Netflix should not have defined success criteria that depend upon assumptions that were not appropriate for the data.

It's just one more case of the math and the programming obscuring meaningful goals for solving a business problem.



December 10, 2015    View Comment    

On A Complete Guide to Overcoming Executives’ Concerns about Hadoop


Do you often encounter CEOs getting involved in these technology decisions? Why wouldn't they leave that to others, such as a a CIO or appropriate VP.

April 18, 2014    View Comment    

On Breakthrough: How to Avert Analytics’ Most Treacherous Pitfall

Sadly enough, I recently witnessed a conulting firm using a slogan along the lines of "we torture the data until it confesses." This was actually on their signage at a trade show.

March 7, 2014    View Comment    

On Technology Training Needs a Hands-On Approach

At the Big Data conference in Chicago a couple of months ago,one IT manager explained that he had a team of programmers who learned Pig in 2 weeks. They were mature mainframe programmers. Think of all the talent like that being wasted in the US today, because we lack proactive involvement and investment from employers.

I spoke with that manager about his approach to training. His methods were simple and straighforward; they could easily be applied at other businesses, and the costs were modest. It requires little more than  the will to do so.

Chicago, where I live, has programs for job retraining, but without commitment from employers, I don't see the results for job seekers. For example, the city is training many people for PMP certification. I could identify a dozen unemployed or underemployed PMPs from my own contacts alone, and all of them have experience and good work ethics.



March 6, 2014    View Comment    

On Technology Training Needs a Hands-On Approach


There is so much to be said for training!

Recently I heard that the City of Detroit wants to import immigrants from other countries to fill tech jobs. Asked why they aren't looking at people from the US, they claim the schools are not producing what employers want. Heaven forbid employers would invest in training employees to fill their needs.

Once upon a time, a friend of mine got a job with a pharma company. She had majored in psychology, but the pharma company trained her to be a systems analyst. She was good at it, and she earned a good salary for her work. I've had the good fortune to meet a couple of the programmers from the Mercury, the first US manned space flight program, and they, too, learned computing on the job (there was no such thing as a computer science major then!) That kind of story used to be common, but not so today.


March 5, 2014    View Comment    

On The Dirty (Not so Secret) Secret of IT Budgets


It's interesting that you give the example of banks here. Some businesses have a lot of flexibility to experiment with cloud-based data storage and applications, but I would not have thought of banks as one of them. Surely there are special considerations in this industry regarding data security and privacy. Can you tell us more about that? How can banks explore new tech offerings while respecting compliance concerns?


Meta Brown

September 30, 2013    View Comment    

On Do You Really Need a "Sexy" Data Scientist?

You may dare, but you'd prove nothing. Human bias toward attractive people doesn't equate to causation of success for an analyst. You could say the same for gender or race. That doesn't make gender or race a cause of success in itself - bias, cultural behaviors are major determinants of things like hiring and promotions.

Indeed, hiring, promotions and so forth tell us little about successful data [science, analysis, or whatever you may call it]. The object of the job is to produce actionable analysis that yields good returns for the employer. Measure that.

May 13, 2013    View Comment    

On Do You Really Need a "Sexy" Data Scientist?

Well, you know what statistics is - a form of measurement used with noisy data. It's an element of thoughtful research, but not the end of the process. A kiss is just a kiss, a sigh is just a sigh, and a correlation is just a correlation.


May 13, 2013    View Comment    

On Do You Really Need a "Sexy" Data Scientist?

If I dredged up some data indicating that men are more successful in their careers than women, something indicating that they earn more, get hired sooner, get promoted quicker, and so forth, would that be a good case for hiring only men? How about if the data suggested that whites are more successful than blacks? Would it then make sense to recruit only white men? I'm certain you understand the flaws in that line of reasoning, and that you see the similarities to the attractiveness requirement.

May 13, 2013    View Comment    

On CEOs: Hold Your Team Accountable for Data Analysis

Is there actually evidence that JC Penney does not use data in its decision making?

April 28, 2013    View Comment    

On Big Data Success Stories: Take Them with a Grain of Salt

It's true, many analytics programs and projects fail. I've written and spoken on this topic often.

That's not to say that analytics, Big Data analytics or any other kind, can't be successful. The problem is almost always that the investment is made fast and furious without a decent plan for success. Often there are not even any success criteria established. How these investments get approved is a mystery to me.

Why would anyone pick on someone for talking honestly about the fact that failure is possible, and common? This is something we all need to acknowledge in order to help us plan and avoid pitfalls. Perhaps the word "thousands" was a bit of hyperbole, but is that a big deal? As for the criticism that the author works for Teradata - he's up front about that, it's in the profile right on the page. What exactly is suspicious about this message coming from someone who works for a vendor? Are people going to hand over their bank accounts because he warns them that projects often fail?

If we don't get the hype out of the analytics biz, then many businesses are going to waste money and conclude that analytics is just another meaningless business fad. Give Paul credit for having the guts to frankly admit that failure is possible.

March 11, 2013    View Comment    

On So, You Want To Be A Data Scientist?

Lillian, it's very interesting to hear from an environmental engineer, one specialty I don't think we've seen on Smart Data Collective before. I'm in total agreement withyou on the value of knowing something about the field you're investigating.

Here's my own take on the same topic:

So You Want to be a Data Analyst

January 13, 2013    View Comment