Aventine Partners

Bubble Charts, Good or Bad?

I like bubble charts.  It seems many others don't.  This is a quick review of bubble charts as a tool, with thoughts about why I like them, why others don't, and how best to use them.

At Aventine, our focus is marketing analytics, but we have a strong commitment to the related discipline of Information (or Data) Visualization, InfoVis for short.  By using good InfoVis techniques, we are able to create better graphical presentations of the analyses we do for clients.

I've watched with some interest recurring discussions in InfoVis circles about the use of bubble charts, including one recently on a site we like, Junk Charts, panning a graphic in the Financial Times using bubbles.  By and large, the InfoVis experts hate bubble charts, feeling people can't do a good job distinguishing differences of bubble size. 

The interesting, parallel development is that motion charts, which are just a type of animated bubble chart, are getting increasing attention thanks to Google and a remarkable presentation at a TED Conference in 2006.  These motion charts are getting traction despite the criticisms.

In Online Ad World, "Value" Doesn't Drive Price, Supply Does

Despite the increasing ability to deliver highly targeted marketing messages, the online advertising market will never, ever command premium rates.  

We love that our clients are able to put analytics to work with campaigns in the online world that take advantage of increasingly sophisticated targeting capabilities and highly segmented audiences.  But it's astonishing to hear the never ending lament from the online publishing side that rates are low and ad buyers just don't place a high-enough value on their product.  The inescapable implication is that the marketers and media buyers just don't get it; that they're somehow not bright enough to see what targeting can do.

I'm sympathetic to being passionate about the value one brings to the table, but I think the ones who don't "get it" are those that believe buyers who are getting a great product at a low price are dim witted.

An Ad Age report this morning, Online CPM Prices Take Tumble, continues to attribute ever falling online CPMs to confusion about the superior value of the product.  Here's the analysis, in brief:

Problem:

• CPM rates are in the basement, below $1 for remnant inventory.
• Sell through rates are so low, around 30% for some online publishers, that remnant inventory pricing is driving the market.

Diagnosis:

• Marketers who have become accostomed to using the web for high-volume, low-yield campaigns never developed the sophistication to take advantage of the better targeting that can generate higher yields.
• Too few publishers have the discipline to "stand up for themselves", sell to the value, and hold the line on pricing.

Solution:

• Better educate buyers about the superior value of highly segmented audiences and ever-better-targeted impressions.
• Be more disciplined about pricing, and refuse to give away inventory below its true value.

This analysis might hold water if not for the most critical characteristic of this industry: nearly unlimited supply at a marginal cost of effectively zero.

Asset Portfolio Risk vs. Customer Portfolio Risk

Nearly every financial institution is undertaking a comprehensive reevaluation of methods and processes for measuring and managing asset portfolio risk.

But what about customer risk?

This came to mind recently when reading an interview with Wells Fargo's CEO, John Stumpf, in Investors Business Daily.  IBD chose Stumpf as their CEO of the year because he had the foresight to keep WFC away from the most toxic forms of mortgage assets, sacrificing considerable profits in the short term but keeping the bank out of trouble in the long term.  WFC may be down 33% from Jan 1, 2008, but that's a far cry better than the 60%+ drop in value for the banking sector as a whole.

I noticed in the interview a story we've been hearing from Wells for years -- an emphasis on cross selling and signs of higher-than-average retention experience in their customer base.  

Measuring and Modeling Culture and Cultural Change

How much of our marketing has been calibrated to a culture and set of values that may be on the verge of radical change?

There is no shortage of opinion seeing our society as hyper-consumerist, bearing the seeds of its own self-destruction.  Some industry commentators see the cultural turmoil afoot ushering in big changes in marketing.  A handy example is the cover article of the most recent issue of Media magazine, “Age of Dissent: Has America’s 50-Year Shopping Spree Hit the Wall?”, by Steve Smith (free registration is needed to access the article).  Smith does a good job of moving beyond mere tactical adjustments to address the near-term economic malaise and sees permanent changes to how marketing is done because of a pervasive change in culture.

Quite apart from questions of whether these changes are good or bad and who and what are to blame for our economic predicament, the fact is significant cultural changes are far more likely now than a year ago.  Furthermore, the probability of dramatic and longer-lasting change is higher, and the direction and profile of the changes may be easier to discern. 

So what does this mean for marketers and marketing analysts?

The Power of "R": The Open Source Alternative for Statistical Analysis

A core philosophy at Aventine is that marketing analytics initiatives can be done pragmatically without breaking the bank.  

One of the developments that made it sensible for us to detach from larger consulting firms is the open source movement.  As the open source options mature and become more robust in areas like databases (MySQL, PostgreSQL) business intelligence (Pentaho, JasperSoft), and even customer relationship management (SugarCRM), starting a client down the path to being analytically savvy no longer needs to be tied to big, expensive technology deployments.  Companies don't need to wait until they have a few million dollars in the budget to start down the analytics path.

But the core of what we do is statistical analysis, and this field for years has been dominated by SAS, a powerful, but expensive, statistics application.  Having SAS as part of an analytics capabilities solution has never been entirely consistent with our low-cost, pragmatic approach.

Fortunately, an open source alternative called "R" is out there and is beginning to take hold.  As a sign of it's coming of age, the New York Times profiled this statistical analysis package in a recent article.  

We've been experimenting with R for about six months and find it to be a good alternative, but with a bit of a learning curve, especially having used some of the commercial alternatives for so many years.  That said, we plan to explore its flexibility and ability to develop add-on packages to perform make possible some data visualization approaches that are difficult with SAS and the alternatives.  We will share what we discover.

Meanwhile, for companies that have not yet committed to a solution, R is well worth investigating.

Here at Aventine, we are remaining non-committal for now.  In addition to SAS and R, we use SPSS, a long-standing commercial alternative that has functionality very competitive with SAS and a much more attractive licensing model (i.e., perpetual vs. annual for SAS).  I've also recently discovered a powerful Excel Add-In called XLSTAT that might be a good cost-effective and low learning-curve solution for some companies.  The downside for us at Aventine is that we are a Mac shop, and the XLSTAT add-in doesn't work for the most recent version of Excel for Mac.

What's in a Name?

When commenting on Warren Buffett's classical allusion in yesterday's post, I realized that nowhere on this site have we described the origin of our firm's name, Aventine.

We started our firm as HCE Analytics (the "H" and "C" for Hughes and Cahill), but as we began to grow became convinced we needed to jettison this insipid name.  

We both have a liberal arts background, retain our love for the classical tradition, and bring a philosophical outlook to our work.  So, combing the ancient world for inspiration, we stumbled upon Aventine, one of the seven hills of Rome.  

It turns out that the Aventine Hill features in the story of the founding of Rome, as told Ennius, in which Romulus and Remus were vying for rights to name their new city (Roma or Remora).  They decided to settle the matter through augury: whichever could spot the flight of auspicious birds first would gain the honor.  Romulus was smart enough to climb the Aventine Hill, from which he was able to first espy the high-flying birds.  The name Aventine likely comes from the Latin for birds, aves.

Hence, today we throw three coins in a fountain in Rome, not Remusville.

Hopefully you see the analogy.  Using data and analytics is like being smart enough to climb the Aventine.  Work with us, and you, too, can be king!

You may have noted the normal distribution curve in our logo.  It's also a hill.

Learning from the Quants

"Beware of geeks bearing formulas," quipped Warren Buffett about the risks of relying on arcane financial models to make investment decisions.

Not only do we enjoy the classical allusion in Buffett's observation, but this illustrates one of the important philosophies behind Aventine's approach to marketing analytics and modeling.  

In short, a model isn't a black box.  It's a mathematical description of patterns of behavior that should be developed and used in conjunction with qualitative descriptions of cause and effect (or at least relationships between events or conditions).  In other words, I should be able to explain in layman's terms what is going on, and it should make sense.

We're not unique in this regard, but we find more and more modeling done by people who don't understand the business but know how to throw data into a stats package and get an answer.  They then pass it off with great claims of "accuracy" to marketers who now have a new way to target but lack real insight into what is really going on.

(And let's not even start with the opposite absurdity championed by many software vendors who want to "empower" marketers with no statistical training to build predictive models using their handy-dandy do-it-yourself predictive modeling module.  Encouraging customers to make marketing investment decisions based on the result of a modeling "wizard" is malpractice, and should be criminal.)

Coming from a corporate and investment banking background, I've tried to keep tabs on the world of quantitative finance and look for ties to what we do in marketing.  Analytics and modeling in the financial world have in many ways developed far beyond the marketing analytics realm, having benefited from massive amounts of money that still today slosh around in that system.  

Paul Wilmott runs a great site on this topic and has been a critic of modelers run amok in his industry.  In the aftermath of the financial meltdown, he recently posted the Financial Modelers' Manifesto.  This extended appeal to be humble in the face of complex reality and responsibly practice within your limitation is also relevant to us in the marketing analytics world.

Here is an excerpt that could be easily translated to any marketing model situation:

Models are at bottom tools for approximate thinking; they serve to transform your intuition about the future into a price for a security today. It’s easier to think intuitively about future housing prices, default rates and default correlations than it is about CDO prices. CDO models turn your guess about future housing prices, mortgage default rates and a simplistic default correlation into the model’s output: a current CDO price.

Our experience in the financial arena has taught us to be very humble in applying mathematics to markets, and to be extremely wary of ambitious theories, which are in the end trying to model human behavior. We like simplicity, but we like to remember that it is our models that are simple, not the world.

In short, if there is no intuition or insight, no hypothesis about behavior driving your model, beware.