• In a complex and busy world, where a marketer’s time is at a premium, will brainstorming, learning, and piecing the world together, become passé?

That would be very, very bad if that were the case. Knowledge is king.

Excellent, provocative post as usual, Paul. From my perspective, for marketers to be truly effective, they need to find the balance here. If marketers only work on the premise of Mr. Anderson's idea that quick correlation sans context is adequate, I think they're in for a rude awakening. Besides bona fide trends, quirky fads pop up. Is that what marketers want to react to? Every passing fad? Every great idea by competitors? Can it work sometimes? Sure, but short cuts won't work much of the time. Marketers need to balance this approach with ongoing research and the compilation of true consumer models to understand what is really going on, ie., true trends. Granted, it's much more work to do that. But if marketers don't do the real grunt work, they can't possibly understand their customers' drivers. Hiding behind endless study sans taking any action is a cop out, too. Armed with real knowledge, inside brainstorming is a valuable tool. Ideas can be put into action, but those ideas are then grounded in solid information.

No, Paul, understanding and constant learning are never passe.

Paul,

It is about balance. However, I don't believe that waiting for statistical analysis serves Marketing well. We marketers need to understand and respond to consumer behaviors. By the time causation is determined, behaviors change. I want the statistical analysis as a way to better understand cause and effect for future planning. Meanwhile, I need to act and be fleixble as well as able to change quickly.

Thanks Paul for this post. I read Chris Anderson's article and it concerned me a bit. I really appreciate your context.

There is an art and a science to marketing and to depend too much on the numbers distances marketers from the fact that customers are human beings. Some would argue that humans have free will (that can be a whole other discussion) but if you agree with that fact, you need to understand the why to be able to influence behaviors.

Neil, thank you for commenting. While I'll agree that knowledge is king, it's also what you do with that knowledge/insight that counts!

Great post, Paul.

This gets to a jugular question: how can we truly know why people do what they do, when very often people themselves don't know?

Often, when we claim we "know", what we really mean is "we've made up an explanation we feel good about that we can sell to management. Hooray logic!"

If we have the courage to let go of our illusion of control a little bit, the approach of "not-knowing" is potentially very liberating.

What new things could we see if we shut off the babble in our brains and focus our attention on what is *actually* happening?

If that sounds crazy, consider this. Whose market cap would you rather have? Your company's, or Google's?

Claire and Lewis, thank you both for commenting. Your points about the need for balance are well taken.

Lewis, as you mentioned, the world is quite complex and fluid. Things change quite rapidly. We don't always have the luxury of waiting for a detailed causal analysis. That said, correlation can lead us down the rabbit hole--sometimes the wrong one. It's always going to be a balance of how much analysis is enough before we need to make a decision...

Claire, you rightly pointed out that research into customer drivers is hard work. It's been my experience that too often the temptation is to take the easy way out, where we spot the correlation and say, 'eureka', when in fact as any good marketer knows correlation is definitely not causation. I believe that modeling and piecing together customer profiles, while hard work, is still the best approach.

Marjorie, thank you for commenting. Much like you, when I first read Mr. Anderson's article, I was under the impression he's presenting us with a false dichotomy. To me, it's not an either/or situation. In many enterprises (both government and commercial), petabytes of transactional data present us with an opportunity to collect, analyze, and explore customer behaviors. Sure it takes time, money and expertise. But as you mentioned, understanding "the why" is very important!

Tom, thank you for commenting. Your comment, "how can we truly know why people do what they do, when very often people themselves don't know?" is why fMRI (neuromarketing) research is gaining in popularity. People often say one thing in survey, but their behaviors may prove otherwise.

We are often quick to want to define causation aren't we? It's tough to sit back, observe and observe some more without trying to discern causation right away. Your point about how much we can truly learn by just patiently observing is well taken.

The more research we do, and the more research our clients do, the more we're apt to discover that our first assumptions were wrong. Or, if they aren't wrong, they're too simplistic in a world filled with complex human beings, who happen to be our customers. That's why real research work has to be conducted, Paul. Most of the time, great results are the result of a lot of hard work. Who was it who said that Genius is 1% inspiration and 99% perspiration?

Ted, you are so right about how our first assumptions are not necessarily the ones that should be acted upon first. I think that's why I'm fundamentally at issue with Mr. Anderson's take on how we no longer need modeling to help us understand—that correlation is enough.

The world is complex, and so are human beings. We need more than “report” functionality –i.e. what happened—to make marketing decisions. Building models helps us understand why things happen. We can then use those understandings to make better decisions on pricing, targeting, budgeting etc…

Hi Paul,

If all marketers do is reply on correlation, they are making assumptions about the future based upon the past. Without understanding why those decisions were made, blind prediction based upon history will lead to many wrong assumptions about consumer behavior. Both correlation and causation are necessary quantitative tools. In addition, for indepth understanding some qualitative tools should be used as well.

Those who understand their consumers the best are better positioned for success in the future.

Hi Paul,

If all marketers do is reply on correlation, they are making assumptions about the future based upon the past. Without understanding why those decisions were made, blind prediction based upon history will lead to many wrong assumptions about consumer behavior. Both correlation and causation are necessary quantitative tools. In addition, for indepth understanding some qualitative tools should be used as well.

Those who understand their consumers the best are better positioned for success in the future.

Camille, thanks for commenting. I agree with you in that quantitative tools are necessary, but only give a partial picture. I'm encouraged by the field of text analytics which bridges both worlds by taking documents, survey responses, emails, call center notes etc and then attempts to enumerate key phrases or selected words so as to put a quantitative spin on qualitative feedback.

Hello, I really like your post and I'd like to give my 2 cents.

Correlation without "understanding" actually produce the following results: I used Google translate to translate Chris article from English to French and although it's overall pretty decent. It translated
" Data without a model is just noise."
into
"Données sans un modèle de bruit est juste"
which would actually mean in French, data without a model for noise is correct.

It's just a 4 words sentence and actually this should give hope to people like me that don't like the 100% behaviorist world that's behind Chris article !

I'm actually reading Chris claim the other way round: online advertisers and mass media advertising have framed and modeled consumers so that they behave like clicks and machines and therefore can be analyzed with simple correlation models.

So it's just measuring what it built.

However, with the rise of social media, fragmentation of channels, re-localisation of communities, this may be a feature of the past.

It may explain why web2.0 companies face such difficulties to monetize their social success into financial success with business models & advertising scheme inherited from the "world according to Google".

I'm a Google fan and Peter Norvig is clearly a great thinker. However, I think there was a particular moment where the computational power was growing very fast while the volume and diversity of content on the web was more in "controlled growth".

In the recent years we've seen a major shift with almost everyone becoming a publisher and a exponential growth in production of diverse content.
People that use to produce "queries" now write full articles, with humor and finesse, using languages and presentation capabilities to their full extend.

I would bet that understanding is poised for a comeback !

Dominic, thank you for taking the time to comment. Some observations from your responses...

1) Machine translation has a long way to go, doesn't it?
2) Behavioral targeting will help complete the picture of our customer profiles, but I think you're alluding to the challenge/difficulty in just using clicks as proof of determining what our customers really want/need
3) Web 2.0 companies, in my opinion, are having a tough time monetizing for various reasons, key among them is privacy/security concerns. This topic would make terrific fodder for another post...

Again, thanks for adding to the discussion.

• As a marketer, do you want to understand and be able to explain the causes of why customers/companies/competitors do what they do, or is observing and acting on the behaviors good enough?

Modeling is good enough for big companies, not for the little guy. When you're a giant, you can spend less money per person and get higher ROI on modeling. I only say that because if you're Dick's Sporting Goods and sell addons (upsells) to 5% of your customers you'll make much more money than if you're the local sports shop that upsells 10% of its customers. I think it just comes down to customer volume.

• Building accurate models requires tons of data collection, the proper analytical technologies and applications and of course the know-how. It’s not easy work. Do you agree with Mr. Anderson that for most decision making opportunities in the Petabyte age, correlation is enough?

I think for most large companies correlation is enough. But that's not the case for everyone. Let's take my favorite example: Amazon.

I buy lots of things at Amazon. They are notorious data munchers. They recommend products to me all the time and they're usually way off base. Just because I bought cereal doesn't mean I'm interested in milk. Maybe I eat my cereal dry. And therefore, a better suggestion would be ziploc bags for me to tote the cereal in.

Either way, Amazon still gets plenty of money from me--just not as much as they would if they had better algorithms.

• In a complex and busy world, where a marketer’s time is at a premium, will brainstorming, learning, and piecing the world together, become passé?

For the majority of companies: yes. For innovators and growth companies: no.

For myself, knowing why is paramount. Knowing why means being able to coevolve with the customer. If you know why, you can anticipate the next move and possibly even beat the customer to that line of thinking.

Computers are only as smart as the people who program them. Computers are really good at calculations and finding patterns in exorbitantly large data sets. When it comes to critical thinking and reasoning, they're a bit lacking.

A computer can search through 5,000 retailers and give you the best price on a Widget model 300 but can't tell you why Widget 300 is better than Fidget 300 or Gidget 300.

What it all comes down to is are you the kind of marketer or company that would rather have lots of customers and lose lots of money to bad marketing or have a very small number of customers but be able to know them better than they know themselves and sell them everything they need and nothing they don't?

There are advantages to both, I just prefer to dominate an industry rather than get the crumbs that fall from the big boy's table.

Michael, thank you for taking the time to add valuable commentary to this article.

I think many of your points are spot on, however I think it's important to define small vs big companies in regards to the value of modeling and then predicting future courses of customer proclivities/actions. If we define small companies as "mom and pop", then I'd say modeling isn't for them. However, there are some companies with <300GB of customer data that are doing customer modeling, and sophisticated BI tools aren't necessary--sometimes Excel (which is very robust) will do.

Whether modeling is right for your business is highly dependent on whether you have enough data, the right IT infrastructure, data collection processes, and the know how to analyze and interpret that data.

There's also a willingness factor - quite frankly some companies are perfectly happy with transactional relationships.

For a terrific real world example of how a major US bank uses models to make better marketing decisions, look up this article:

http://www.marketsharepartners.com/documents/HBR_May_2008.pdf

Paul, there's a fascinating article in the current issue of "The Economist" that relates to this discussion.

See http://preview.tinyurl.com/5qpn8j

I have the feeling that this will ultimately turn out to as useful as phrenology. But, I could be wrong.

Tom, thanks for pointing me to the article. I'm about two weeks behind on my Economist reading.

As you rightly pointed out, even the debate between the Princeton economists (who don't care about the "why") and other behavioral economists (who are trying to figure out why we make decisions the way we do) rages on.

Thanks for contributing to the discussion!

I think yoiu're being rather kind of Anderson's article, which seemed like a great example of a false dilemma leading to idiocy.

Marketers aren't scientists. If your goal is to maximize the efficiency of your marketing programs, correlations are enough; you don't need to know why something works to decide to put more of your budget into it, refine your approach, etc.

Scientists are not marketers. Figuring out the why leads to greater knowledge, better understanding of physical processes, and unexpected breakthroughs.

The right balance between analysis and the search of root causes, and taking action based on correlations? Well, that depends on who you are and what you are trying to accomplish, doesn't it?

John, thanks for taking the time add your comments. This post is much richer because of the insights from DailyFix marketers like you.

The discovery of knowledge--in science, marketing (or when both worlds sometimes cross over) will never be passé!