eMetrics NYC 2011 Recap - three major takeaways

The eMetrics NYC conference / data driven business week, is said to have attracted some 1600 people. The WAA Industry meeting coincided with it. It was a success.

Three major takeaways that stick out in my mind.

Janus Faces for Janus Audiences

I've now seen two versions of Peter Fader. The academic Fader and the industry Fader. The one you see at an INFORMS Marketing Science conference is the academic Fader. He's the kind of guy that'll smile as he tells his skeptical detractors to take their perspectives and reconsider them. His insights into models and the way theories have been constructed are original. He doesn't mess around with the bits at the edges. He's good at the comparative method among models, even if that's not quite how he'd describe it.

Then there's the industry Fader. He's accessible and engaging, even for a Friday afternoon. He presented a pretty good case of the way that marketing scientists in academia work with industry practitioners, and went on to highlight a variant of the shop-till-you-die model. It was interactive and participatory - the very opposite of INFORMS MS - and it was excellent.

This is probably the most radical example of the 'know your audience and adapt' that I've ever seen. It's inspiring - just how far a model can be contextualized and explained to a very different audience - to that extent. Incredible.

(Finally, for me, it's a key datapoint for the evolving partnership between academia and industry, for mutual benefit.)


Mobile has arrived

It's notable just how little hype there was for mobile this conference. Every year is supposed to be the year that mobile breaks out. That it's going to be the year of the hockey stick. Discussions around app and tablet usage were muted, and rare.

I saw loads of tablets. Loads of smartphones in use. There was tremendous engagement on Twitter and Beluga throughout the conference.

The reason for the lack discussion, possibly, was that mobile has already broken though.

Have we started asking interesting questions yet?


We're well positioned to take advantage of the next 5 years

For most of us in the East, the October eMetrics is our major collaboration. Not everybody made it. Those that did included long time collaborators and conspirators. And it was intense.

I was very happy to share a stage with the only other data scientist there - Michael Healy - and that panel led to a very spirited discussion with Jim Novo and one of SAS's leading text miners (IE: Richard Foley). The debate changed an important perspective I've long held on data quality. That discussion couldn't have happened anywhere else.

The sophistication of discussion was high. I spoke to a number of directors about their challenges with social. They were having the same problems I was having, and I was happy to share how I worked with developers to solve them. Many of the solutions reside within Syncapse Platform, and are getting better every day. I found that others were also really forthcoming with the problems, solutions, and mistakes they had made, in a range of fields. Again, those discussions couldn't have happened anywhere else.

Based on the quality, caliber, and energy I felt at eMetrics, I'm certain that we're well positioned to take advantage of the next 5 years. We have incredible challenges, opportunities, and technologies to take advantage of.

What's next

Patrick and I thank, again, everybody who came out to the 'Communicating to designers' preso. June Li remarked that the style is completely different from talking to executives. And I'm inclined to agree. It was very unusual to have a design discussion at eMetrics, and yet, I wonder - why should that be the case?

I'm up for another six months of contributing to eMetrics Toronto. I'm particularly interested in stories from the client side that follow the dramatic structure. No, seriously. It's great storytelling.

The next WAWTO is October 26 at the Wellington. I'm looking forward to seeing many designers, web analysts, developers, hackers, IA's, data scientists and marketing scientists out for the event.

Insight Revisited

I wrote a definition of 'Insight' on December 29, 2010 that read:

"An insight is:

• New information
• Executable
  
• Causes action
• Profitable

Or, more detailed, an insight is:

• A piece of information that you didn’t know before, which -
• Can feasibly executed, culturally acceptable and of a scale relevant to the firm, and -
• Causes a decision to be made that wouldn’t have been made otherwise, and -
• Results in profit or a sustainable competitive advantage"

If held to that standard, insights are incredibly rare.



Depending on who you talk too, an insight may mean:

• A bullet point factoid

• An element that combines all that is known about a segment, compressed into the foundation what motivates and inspires them

• A simple statement of cause and effect

So, any of the following have been called 'insights' in the past:

• Sales increased by 4%
• Young men 18-25 compete to get laid and look for edges
• Scoring better Google PageRanks increased organic traffic by 9%

It's all possibly all 'new to you'. The rest is all context. Bullet one isn't actionable. The second bullet point is actionable if a company can take action by solving that particular segment problem or aspiration. The third is getting there - what's the unit of measure for 'Google PageRanks' - and is it a lever I can push?



There's a lot of confusion out there in the market about just what an insight is. It's a problem.


Whenever you're asked for more insights, please ask, 'what do you mean by more insights'. It's the only way to get into their head. Yes, it's aggressive inquiry, but you won't be chasing your tail.

The Most Measured Era in History

We're living in the most measured era in history.

Are you the beneficiary of any of the data you're generating?

You optimize what you measure.

One of the most data intensive self-improvement projects I undertook was in 2005. I recorded everything I ate and every exercise I did. And did I ever optimize - to the point where my joints couldn't keep up with the muscle and bone growth. It was a massive amount of work to record all that detail, the weights of various things and then to cross reference with the USDA database. Then it all had to get loaded into SPSS for analysis. It was brutally time intensive. But it did generate incredible evidence-based insights about the way my body worked.

Those insights formed heuristics. The no-fry principle. The 100g daily protein target. The outer aisles principle for super market shopping.

The friction and time intensity ultimately meant that measurement receded.

It's 2011. The recording and the cross-referencing should be made better why way of mobile apps, so I'm looking forward to less friction on that front.

The transferability of data from that device into a format that can be read by SPSS or Python is a lingering problem.

Ideally, a machine would be able to make some sense of the data on my behalf. That is another, wonderful, opportunity.

In many ways, I should be the beneficiary of the data I generate. If I want to benefit other groups, like researchers, then so much the better.

It's my hope that more developers will partner up with statisticians to produce incredible data driven systems that provide real utility to people. That people can use to make themselves better and save a lot of time. There will always be other beneficiaries of that data - after all - if the product is free, you are the product. That's not necessarily a bad thing, especially with consent. It's a good thing.

We should all benefit from the data we're generating ourselves, for ourselves.

The fight for the Data Science Soul Begins

This is a pretty good summary of the definition of data science. Some statisticians seem to be incensed. Some people say that this whole thing is invented as an O'Reilly buzzword. And there's consternation, fear probably, over the devaluation of actual craft.

Sound familiar? Ah, the great Web Analytics debate of 2007. Yes. We've seen this.

Nothing like a fresh Gartner Hype Cycle in the morning, is there?

But lets consider what technology is causing, and the role that data scientist will play, in driving that cause.

Accessibility to data is expanding. What used to be the jealously guarded by people who didn't want to be educators, is now liberally spread. It doesn't really matter that most people don't know what the figures means, does it? At best, it's making big parts of the world smarter. At worst, it's merely reinforcing pre-existing ignorance about what people conveniently want to believe.

Pop-business literature is good. More people are aware of the potential. It's awesome.

And there's no resisting this market. Everybody wants data because they believe it will make them better. That it will make the smarter. That it will result in sustainable competitive advantage.

And it can. Data is preresquisite. Understanding, well, that's something else.

Data scientists will use the very best of computer science (computability, algorithms, scalability), the very best of usability (IA, UX, Infometrics) and the very best of statistical analysis (models, probability, learning). How they do so, and what are the best patterns for success, will fuel an entire generation of operations research. A sort of meta-meta study of the meta-meta of competing on analytics.

Too much meta?

Not enough. Not nearly enough.

While the fight for labels will begin, and then persist, for the better part of this decade - the proof will be in the experiences, and, one degree of causality out, the results.

***

I'm Christopher Berry.
I tweet about analytics @cjpberry
I write at christopherberry.ca