Chief Marketing Technologist

You are what you don't automate

"You are what you don't automate," one of ion's engineers commented in a meeting the other day.

It was in the context of a series of time consuming, manual steps that had to be done for a particular task. He attributed the adage to super-programmer Jeff Atwood, although I've not been able to find the reference.

It struck me as a brilliant way to frame the challenge of marketing automation.

See, in software engineering, most developers feel that time spent doing anything other than designing and coding great software is, well, kind of a drag.

So they try to automate as many of the boring processes of assembling, testing, deploying, operating, and debugging their software as possible. When they find themselves doing a rote task, the machinery in their heads starts whirling: is there any way to write a script so I don't have to do this by hand again?

To engineers, "you are what you don't automate" is a chiding remark: don't have you something better to do with your time than type that same sequence of commands over and over again?

Automation is about eliminating mindless busywork to make more room for creative and meaningful work. More time for distinctly human contributions.

In marketing automation, however, I often hear people talk about automating some of the most human aspects of their jobs: segmenting customers, judging opportunities, designing truly delightful customer experiences. Not surprisingly, such visions for automating that which humans do best often run into trouble.

Marketers should automate their own mindless tasks — things that take up their precious time that don't require thought. Many marketers I know still get bogged down in a lot of busywork of this kind, so there's clearly plenty of opportunity for automation to help.

But don't be quick to blindly surrender what your insight and instinct serve best.

"You are what you don't automate" can also be a badge of honor. What are you? The quintessential, best parts. Don't automate that.

More insight, less rhetoric in politics and marketing

If you haven't already seen Jess Bachman's brilliant "1-page book," Death & Taxes, you've got to check this out (click to enlarge):

It's a superb example of what Edward R. Tufte calls a "supergraphic." (For more on the art and science behind such supergraphics, grab a copy of Tufte's latest book, Beautiful Evidence.)

Seth Godin, whose new publishing venture The Domino Project published Death & Taxes, wrote this in a post on the power of visualization: "Data is not useful until it becomes information, and that's because data is hard for human beings to digest."

The U.S. budget, in its raw form, is a classic example of undigestible data. Jess Bachman has transformed it into something that anyone can digest as useful information.

"It is not possible to spend less than 10 minutes looking at this, and more probably, you'll be engaged for much longer. And it's definitely not possible to walk away from it unchanged. That's a lot to ask for a single sheet of paper, but that's the power of visualizing data and turning it into information."

My takeaway, in the context of U.S. politics, is that it's a relief to have some actual concrete, digestible information to inspire meaningful discussion and insight, instead of the endless unsubstantiated rhetoric from most politicians and the media that love them. But don't worry, this isn't becoming a political rant blog.

Rather, for marketing technologists, this is the bar that should be set for marketing analytics.

While many people are talking about the power of "big data" — a hot topic in marketing, as according to IBM's latest CMO study, over 71% of CMOs surveyed now feel painfully underprepared to handle the explosion of data under their roof — there probably isn't enough emphasis on the visualization of data.

Not just visualization for the sake of eliciting "ooo's" and "ahh's." But visualization that genuinely reveals meaningful information, generates useful insights.

Because if there's a runner-up for a profession that could benefit from more insight, less rhetoric, marketing would certainly qualify. The transformation from (merely) impassioned rhetoric to (also) information-driven insight, is a big part of the core of the new marketing.

In related news, Visual.ly — a new web start-up focused on letting people create amazing infographics — has just raised $2 million in seed funding.

Computational science/computational marketing

This morning I was reading a transcript of a talk the late Jim Gray gave a few years back on eScience: A Transformed Scientific Method, and I was struck by the parallels in our current transformation of marketing.

This terrific slide from Jim's talk sums up the evolution of the paradigms of science:

Although the purpose of science has remained constant — just as the purpose of marketing is indifferent to the digital revolution that has engulfed it — the day-to-day work in pursuit of that goal has changed radically. Jim remarked:

People now do not actually look through telescopes. Instead, they are "looking" through large-scale, complex instruments which relay data to data centers, and only then do they look at the information on their computers.

The world of science has changed, and there is no question about this. The new model is for the data to be captured by instruments or generated by simulations before being processed by software and for the resulting information or knowledge to be stored in computers. Scientists only get to look at their data fairly late in this pipeline.

Sound eerily familiar?

Later in his talk, he notes that experimental budgets in major science projects are becoming 1/4 to 1/2 software. "Even in the 'small data' sciences, you see people collecting information and then having to put a lot more energy into the analysis of the information than they have done in getting the information in the first place."

The challenges this presents, in Jim's list of "generic problems" (slide below), are nearly equivalent in science and marketing:

We are seeing the evolution of two branches of every discipline, as shown in the next slide [Figure 2]. If you look at ecology, there is now both computational ecology, which is to do with simulating ecologies, and eco-informatics, which is to do with collecting and analyzing ecological information. Similarly, there is bioinformatics, which collects and analyzes information from many different experiments, and there is computational biology, which simulates how biological systems work.

Marketing informatics and computational marketing, anyone?

Jim's suggestion? A renewed focus on tools to make X-informatics and computational-X more efficient and effective. Or, in our world, the marketing technology landscape.

Is marketing becoming a math state? Is that good?

There's a fascinating debate raging in the DigitalNext section of Ad Age that captures some of the current "culture wars" on Madison Avenue between marketers ("mad men") and technologists ("math men").

A few weeks ago, Kendall Allen wrote a piece titled: The Dangers Of Online Advertising's "Math State" that begins with this shot across the bow:

Ever notice that in online advertising no one talks about the message? It's all about the mechanics, how an ad is bought, served, delivered because that's where the cool money is. Even though the supporting technology and platforms have never been more abundant, capable, efficient and smart, we remain a math state, which ultimately will backfire and dumb down the agency ranks.

The rest of her article rails against the increasing dominance of the "math state" mindframe, especially in the agency community, and the harm that is doing to brands and the DNA of agency talent:

With our hyper-focus on efficiency, mastery of the mechanics, tuning of the [tool sets] — what can we really expect to do long-term for the brand? Even though they all have certain performance goals and sales targets, was the brand marketer's ultimate goal always such large-scale ROI? What about competitive marketshare; brand visibility; consumer advocacy; and so on? The way we operate now suggests that mass efficiency and uber ROI are the only imperatives. This cannot be an authentic long-standing objective that clients really have in their heart of hearts. We absolutely bred that beast with our math and our fancy tools.

Joe Zawadzki responded the following week with a rebuttal: Why the "Math Men" Will Set Online Advertising Free.

He pushes back hard — not that technology should rule the roost, but that rather it's an enabler for the very goals Kendall is championing. He writes, "In the past three years, all of those scary concepts of RTB, DSP, ad exchanges and data exchanges have been about radically fixing digital advertising — using technology and math [to] do what computers and models do best — freeing people up to be able to think."

A section of his article that I found particularly compelling:

There's no unringing a bell: math and technology are part of [the] future of marketing.

But it's not art or science, people or machine. It's mad men AND math men.

Marketers can rise up above the drudgery of execution. They can put systems in place — yes, the demand-side platform — that provide the technical and analytic infrastructure to succeed in this next generation of marketing. Math and technology have already been bundled and made easy to use, so why resist: let the machines do what they are good at, and let them help the marketer be smart again.

Then marketers can focus on stunning creative concepts, or bold re-imaginings of the brand promise. The machine won't come up with the message, but it will optimize the medium like it's no one's business.

Kendall (and a whole mob of other people) then join in the comments.

Kendall replies, "I do not eschew the math. But, I am deeply wary of its dominance in the business and brand-building equation. That is dominance implied by a total math state that we run the risk of establishing if we don't keep mindful of the balance."

Good points on all sides. What do you think?

(Thanks for alerting me to this, Audrey.)

Analytics: it's not about the math

Earlier this month, I attended eMetrics and Conversion Conference in San Francisco, part of a "data-driven business week" federation of related events. One of the highlights was a keynote by Tom Davenport — author of the bestseller Competing on Analytics — titled "The New Quantitative Era."

Tom started with the observation that analytics teams are blossoming like flowers in Springtime throughout organizations. Web analytics, marketing analytics, HR analytics, supply chain analytics, predictive analytics, even actuarial analytics (the dismal science with a dashboard?). Collectively, these different beachheads combine into enterprise analytics.

He called it the "analytical equivalent of the European Union." I'll call it the United States of Analytics. It's a loose affiliation of independent groups in the enterprise who are harnessing similar techniques to their different domains. They're starting to connect to each other. And together, they share a common agenda: making better decisions in business based on data.

It's not about the math

But this transformation shouldn't be framed as data over human dynamics. To the contrary, the two should be entwined into a new culture of analytical leadership.

To illustrate this point, Tom quoted Karl Kempf — who has the title of Chief Mathematician at Intel and is an Intel Fellow — as saying: "If you want to be good about analytical decision-making, it's not about the math."

Yes, according to the Chief Mathematician, it's not about the math, "It's about the relationships."

"The math person must understand the intuition and speak the language of the business person." This is an asymmetrical relationship where the math person needs to have enormous respect for the business person, and the business person should have at least a little respect for the math person. The right models can help businesses compete, but models in the abstract have little value unto themselves.

A big part of capturing the value in analytics is being able to translate data into stories. "A lot of people can tell stories, a lot of people can work with data, but there aren't too many at the intersection," said Tom. "Without a story, you're not going to influence a decision very much."

New analytics: engineering + liberal arts

In Tom's opinion, there are 4 skills that analytics professionals need to step up to the next level:

1. The ability to tell a story with data.
2. The ability to help frame decisions with data.
3. The courage to stand firm when necessary.
4. Don't just identify problems in data — propose solutions!

Ideally, these skills should be applied in "agile analytics" — an approach, like agile development and agile marketing, that favors user-driven deliverables with frequent outputs and check-ins.

Tom sees this as essential to developing "a new analytical culture" that emphasizes:

• facts, evidence and analysis as central to the decision-making process
• a pervasive "test and learn" approach when there aren't yet facts available
• a free pass for push-back — okay to question people, "where's your data?"
• never resting on your analytical laurels — no model is a sacred cow

In surveying hundreds of corporate decision-makers, Tom found that the most common ways to "make better decisions" alternated between left-brain and right-brain techniques. Analytics was the #1 answer. But "changes in culture and leadership" was #2. Better data was #3. But "changes to process" and "better education" where #4 and #5 respectively.

"If your job is to make better decisions, you have to alternate between these two worlds."

"A new model for analytics is emerging," Tom said in closing. "There's no better time to be in analytics. But it will require new management approaches and new ecosystems. This is not business as usual — this is a historic opportunity to transform your industry and function."

If it's in an infographic, it must be true

Making the rounds on Twitter is a hilarious "infographic" by smarter.org and byJess.net, Why Apples Are Better Than Oranges, satirizing half a dozen different ways in which data, statistics, and visualization can be — and often are — thoroughly abused.

As the soft art of marketing has been juxtaposed with the mathematical precision of analytics, fascinating hybrids of evidence and imagination have proliferated — often wrapped in very pretty illustrations.

Want to keep yourself and your company out of the quagmire of publishing — or worse, making decisions on — bogus infojunk? Darrell Huff's classic How to Lie With Statistics (1954!) is a great primer for avoiding and debunking the common errors of logic and visualization, intentional and unintentional, packaged in this way:

• correlation does not imply causation
• avoiding selection bias with truly random sampling
• visual tricks of scaling that confuse and distort

Or you could read pretty much anything written by Edward Tufte.

Or you could simply grok the different levels of what's being satirized here (click to go to the original site):

Infographic by Smarter.org

Is this the future of marketing education?

While reading MapMyFollowers is a Great Twitter Data Visualization on ReadWriteWeb this morning, I saw this ad for a new graduate program offered by Columbia University:

I know, the fact that someone paid attention to an ad is almost newsworthy unto itself these days.

But what a fascinating program! I've remarked before that journalism is one of the few professions that's suffering more disruption than marketing (for many of the same reasons). Here's an attempt by one of the world's leading journalism schools — now partnered with a top computer science department, if I may stroke my alma mater — to tackle this challenge head on with deep integration between journalism and hard-core software engineering.

If you visit the school's page on Dual-Degree: Journalism & Computer Science, their program brochure starts with this call to arms:

Columbia University's innovative dual-degree program, accepting its first students in 2011, will offer a Master of Science in computer science and journalism. Students will receive highly specialized training in the digital environment, enabling them to develop technical and editorial skills in all aspects of computer-supported news gathering and digital media production. The goal of the program is for its graduates to refine and create new news gathering and digital media technologies that will redefine journalism as we know it.

If you substitute "journalism" with "marketing," and replace "news" with "customer," this would sound like a killer program for fostering a new generation of marketing technology leaders. Instead of two semesters in the journalism school and three semesters in the engineering school, simply swap those two semesters in the journalism school with two semesters in the business school on a marketing concentration track.

Wired wrote a terrific article on this program earlier this year, Will Columbia-Trained, Code-Savvy Jouranlists Bridge the Media/Tech Divide? The heart of their vision is not merely aligning IT and journalism, but fusing the two disciplines together:

The Columbia program ... seeks to attack the barrier between journalists and the increasingly important IT professionals whose web and digital savvy are crucial to any form of news gathering, reporting and delivery. The problem: users really don't know what to ask developers for (or how), and developers have no real idea what their software will need to do in the hands of the users.

"The IT Department [at a news organization] comes up with software programs that the journalists don't use; the journalists ask for software that is computationally unrealistic," said Julia Hirschberg, professor of computer science at the Columbia's Fu Foundation School of Engineering and Applied Science. "We aim to produce a new generation of journalists who will understand both fields."

Again, substitute "journalists" with "marketers." The list of technologies that Wired reports these new graduates are hoped to contribute include:

• Automated journalism modules
• Data visualization
• Deep data mining and analysis
• Device-driven journalism
• Digital trust
• Identifying under-reported events
• Next-generation narrative

Who will create the marketing/computer science version of this? A master's degree in computational marketing?

Ultra-large-scale marketing operations

A couple of major trends in software development — in particular, open source collaboration and the design of social network/user-generated content platforms — may provide useful insight for the future of marketing management.

After all, the increasing number of marketing channels and the increasing granularity of initiatives in them combine to form ultra-large-scale marketing environments that share similar properties to ultra-large-scale (ULS) software systems.

It's not coincidental that many of the leading web sites whose value is derived from crowdsourcing and peer production — MySpace, Facebook, YouTube, Wikipedia, Twitter — are also at the heart of the digital marketing maelstrom. The underlying forces are the same, and they feed each other. Where would Twitter be without the championship of so-called social media marketers? Where would social media marketers be without the latest infusion of attention to their mission that Twitter has brought?

This weekend, I was reading a presentation called The Metropolis Model: A New Logic for System Development by Rick Kazman of Carnegie Mellon University. Although it's written in the context software development for ultra-large-scale systems such as Facebook, many of the characteristics he describes should resonate with digital marketers:

• mashability
• conflicting, unknowable requirements
• continuous evolution
• focus on operations
• open teams
• sufficient correctness
• unstable resources
• emergent behaviors

At a certain scale, the closed-loop, top-down structures of old-school software — and old-school marketing for that matter — simply don't cut it. Instead, one has to design mechanisms for enabling broad participation and co-creation, yet do so without losing the core integrity of the system as a whole, its gravitational center.

Some of the principles that Kazman recommends make sense in software and marketing:

• egalitarian management — encouraging a broad swath of involvement
• bifurcation — tight control of the "core", openness in the periphery
• fragmented implementation — many independent participants contribute
• distributed testing — get many people involved in quality control
• ubiquitous operations — the system is "always on", always evolving

In reading Kazman's work, I followed a trail back to a 2006 report by Kazman and numerous collaborators on Ultra-Large-Scale Systems: The Software Challenge of the Future. Written for the U.S. Department of Defense (DoD), advising them on their goal of "information dominance", it has a funky yet fascinating twist if you substitute "systems" and "software" with "marketing" and swap "DoD" with "your company".

Again, the areas they recommend addressing in their research agenda seem highly relevant to marketing:

• human interaction around socio-technical systems
• computational emergence using game theory and digital evolution
• design that encompasses individuals, organizations, and software
• adaptive system infrastructure for massive decentralization
• adaptable quality control in the face of continuous change

Have computer science and marketing already converged, but they just don't realize it yet?

Marketing more human, more computerized

Two powerful and parallel trends are underway in marketing.

First, marketing is becoming more human. This is the social media revolution. Blogs, Facebook, Twitter, LinkedIn groups, etc., are thriving with real dialogues, between real people in organizations and real constituents in the market.

This pushes marketing as a whole to be more real — customers aren't abstract models on a whiteboard, but splendidly diverse individuals with emotions, opinions, and microphones. What used to be a novel concept, so-called "voice of the customer", is becoming an integral part of everyday marketing.

Second, marketing is becoming more computerized. A veritable tsunami of new marketing software has arrived — with more on the way — as a result of 7 converging forces:

• largely digital channels of communication to the market;
• cloud computing and on-demand software-as-a-service;
• an explosion data and analytics on customers and markets;
• open APIs for everything from ad networks to CRMs;
• the emerging semantic web (a.k.a., a "web of data");
• a new generation of AI-like algorithms at the intersection of computer science + economics + marketing;
• ever cheaper and faster computing power and networking;

This is enabling many of the ideas around marketing automation and computational marketing. These technologies change the scale at which marketing programs can operate — huge numbers of smaller atomic elements — allowing thousands of micro-segments of audiences to be tracked and engaged, in some cases down to one-to-one personalization and offer optimization.

But... how do these two trends affect each other?

Social media communities quickly reject "automated" participation (see David Armano's excellent post, how to be more human).

And marketing software still lacks human intuition and creativity — and will until the singularity.

Marketers must be the bridge between these two engines of innovation, and the wide range of possibilities for how to do that represents a wealth of strategic marketing choices.

A beautiful paradox of how marketing will be both more human and more computerized.

In search of computational marketing

Ever since reading that article in The New York Times about Wall Street-like data analysis being applied to Madison Avenue marketing — what I would call computational marketing, as a nod to computational finance — I've been searching for more stories about that idea.

It turned into a bit of a nomenclature expedition.

Computational Marketing

A search for "computational marketing" in Google brings back ~1,000 results, with the first page being dominated by a single SEMPO job posting from January 2007 for a computational marketing internship. ("These opportunities are ideal for mathematics, computer science, operation[s] research, statistics or engineer[ing] students".)

Most of the job description is grunt work that doesn't seem particularly computational ("growing a list of reciprocal links", "mailing list discovery and management", "reverse word of mouth advertising"). What exactly is reverse word of mouth advertising? But then the very last bullet is:

Implementing computational marketing campaigns using distributed architecture

You know, in between rote data entry, have the intern kick that out. Oof. This is a classic argument for why marketing needs a technologist on the management team.

But clearly "computational marketing" has not been a hot term. Bummer, I kind of like it. (Any chance of a rally to make #compmktg a trending topic on Twitter?)

Algorithmic Marketing

So next up I searched for "algorithmic marketing". A mere 131 hits for that term, although the first page of results has a little more substance. One is remarks from a GigaOM interview with Gail Ennis, Omniture's SVP of Marketing, where the interviewer asks if customers are ready for sites that optimize themselves. Ennis replies:

Some customers say, "I don't want the system to do it. I want to control it." And we love them to do that. Particularly as it gets to the keyword spend, it's so complicated to try to optimize keyword rules so people are much more reluctant to say, "I'm going to let an algorithm take that over for me." But some companies, and some marketers within companies, just get it right away: "We're going to throw everything into that autoserve, let the creative serve up, let the algorithms build." [At this year's user conference, Omniture CEO Josh James in his remarks] opened up with this very fact, that algorithmic marketing is ripe for automation right now.

Another is from a Scandinavian (?) company called Poets + Plumbers — isn't that a great name for a digital marketing agency? — that has a post on digital marketing "micro patterns". Talking about the different factors that play a role in how people respond to marketing execution (product, price, positioning, message, method, media), especially in a social network, they point out:

One of the most important aspects of new-school marketers is their ability to analyze data and identify consumer response patterns. This will give the marketer the opportunity to analyze data and identify consumer response patterns. This will give the marketer the opportunity to constant[ly] update and [improve] the marketing process. In an online marketing context, this model can be taken to an extreme by integrating the analysis with an automated response system, thereby creating algorithmic marketing. In the algorithmic marketing setup, the choice of message, method and media is constantly switching, depending upon which combination gives the best effect.

Okay, "algorithmic marketing" isn't yet a widespread term either — but the meaning seems to be heading in the right direction.

Marketing Engineering

My final search was for "marketing engineering" — which somehow sounds less cool to me than the previous terms, so I wasn't expecting much. Lo and behold, a whopping 210,000 results came back.

The top result is a site called Marketing Engineering, run by two Penn State management professors Gary L. Lilien, who apparently coined the term, and Arvind Rangaswamy. The pair have written a couple of textbooks on the subject, Principles of Marketing Engineering and Marketing Engineering: Computer-Assisted Marketing Analysis and Planning.

However, it appears that a lot of the focus of their vision of marketing engineering predates the web. It includes things like cluster analysis for market segmentation, multinomial logit analysis for customer targeting, conjoint analysis for new product design, etc. But the big picture concept that they're advocating seems right on:

Several forces are transforming the nature, scope, and structure of the marketing profession. Marketers are seeing increasingly faster changes in the marketplace and are barraged with an ever increasing amount of information. While many view traditional marketing as art and some view it as science, the new marketing increasingly looks like engineering.

Lilien and Rangaswamy's emphasis is on using software and techniques to help marketers "collect the right information and perform analysis to make better marketing plans, better product designs, and better decisions". They position marketing engineering as decision support, in contrast to the automated execution implied by algorithmic marketing.

The authors make the first chapter of their Marketing Engineering book available for download free, The Marketing Engineering Approach. It's a great read, although again, the application of this to state-of-the-art digital marketing is left as an exercise for the reader.

So I'd say "marketing engineering" has the most momentum behind it. But will it make the leap to the new new marketing, or will the term evolve with the discipline?

Are there other names that capture this intersection of marketing and computer science? (Be nice.)

Or is this simply marketing automation?