Lessons From Prediction Competition

What did we learn from predicting song popularity?

A recent data science competition focused on whether one could predict the success of a Taylor Swift song from less than half a second of audio.  One sample sounded like:




The objective was to predict whether that clip (and some 20k others) came from a popular song, or not.  As mentioned in a previous post this is a case where there aren't descriptive elements that "make sense" - no genre, gender, demographics, contact history, or anything that marketers typically rely on to help them understand what is going on.  Just 3,300 numbers - for the audio among you, down-sampled to 11k.

The winning models got to above 99% accuracy which sounds a bit to good to be true. While technically correct there are some interesting lessons to be learned.

First, by rethinking the problem as a segmentation problem rather than individual estimates the results got a lot better. That is, by grouping clips based on their similarity the accuracy improved - this is no different than targeting audiences as opposed to specific individuals.

Second, simple models tended to work as well as complex. In this case, accuracy mattered so effort was put on improving that as much as possible. But there are times when good enough is, well, good enough.  It turns out that a simple model of "how similar is this clip to its nearest neighbor" worked very well. With this challenge, that makes sense, pop songs are similar across their 3-4 minutes. After the competition I tried some really, really simple (and possibly stupid ideas) and did better than my official submission. Don't over think it.

Third, we needed to take a step back and align technique with problem.  The same data and the same logic resulted in different results based simply on the approach taken. Always use two or more methods.

A big thanx to Devin Diderickson for posting his approach and thought process.  (He finished second, I finished sixth).

Marketing Analytics Requires Judgment

What advantage do marketers have over machines?

There are lots of cool examples in the use of math, models and machines to predict, classify and recommend.  Some can be found on econsultany's site. Within the domain of marketing analytics there is the application of machine learning whereby we are simply trying to discern cues about our business from the data.  It works, we can predict content you might like on NetFlix or recognize a motorcycle from a car (most of the time). But it is hard to do; NetFlix offered a million dollar prize if someone could improve on their approach by 10%.

Why is analytics hard?

Computers can't see.

The more complex the problem and/or approach, the more time we spend on working out how not to generate spurious results. Something as simple as correlation needs human oversight. Here is one of my favorites from a great collection of spurious findings.

It appears that Nicholas Cage movies and drowning in swimming pools go hand-in-hand.

Then there's the one about the decrease in honey bee colonies tracking with the rise in juvenile arrests for marijuana.  Really?

Computers lack what seasoned marketers bring to the table - judgement. At present a lot of the work in the "big data" space is focused on how to best capture human knowledge in the analytics. In image recognition for instance how do we find that there are in fact cats on YouTube if we don't there are cats to begin with?

In short, we are trying to find ways to learn what we don't know we need to learn. This unsupervised learning can be a lot like kindergarten recess where we need an adult on the playground to keep us safe.

Thinking and judgement will never go out of style.

Always ask: Does it make sense?

Linking Instagram, Pinterest and Houzz to e-commerce

What products do my pictures suggest?

This is a continuation of an idea that arose from a conversation in a bar around the question: "How does content work?"

Recommendation engines typically work with data from one site. We've all seen examples of "people also bought..." on web sites. But an interesting idea would be to link my curated content (Instagram, Pinterest, or Houzz) to the shopping experience of an e-commerce site.

To illustrate the idea, here's a widget from Houzz narrowed to bedrooms:

Assuming for a second that is my idea book, I would appreciate the shopping site recommending the second bed below rather than the first.

To make this a reality requires a couple of things. First, we need access to both sets of pictures - this is a perfect use case for a widget. The benefit seems obvious: "Log on to your x account and we'll do the heavy lifting of sorting thru a couple of hundred thousand items for you."

Second, we need to identify what the pictures contain or portray. This argues for auto-tagging of images to classify and assign attributes. It is unlikely that traditional work-flow processes can handle this task because consistency across two businesses is required. For this to work we need the same method of classifying two sets of pictures (mine and theirs) to a common set of standards. The good news is there are web services for doing this on the fly. Although, if I were an e-commerce site or travel site I'd pre-process my pictures as part of the editorial/content management process.

Third, we need to define something that allows us to rank the pictures. And that means we need a measure of similarity. In market basket analysis it is easy - count the times each pair of products appears in the shopping cart and pick the most common. In this case though, all we have is a list of tags for each picture. In this case there are approaches to similarity ranging from a count of the shared tags among the whole set to computing how many steps are required to convert one set to the other.

The web services I looked at both give a 'confidence' score about the tags. This numeric value can be used to filter tags (clean out those we're not sure of) or be used to weight the tags. This gives the solution a new dimension and also helps to remove judgment from the picture.  

Next up. Working to get my own idea book and curated content classified and do some more experimentation.

Marketing Measurement: Lessons from Physics

How should we measure our efforts?

There is nothing new about the idea of measuring marketing. What is changing however is out ability to do so. Yet, we're not there yet.

Metrology, or the science of measurement, requires three things:

1. An internationally accepted unit of measure.
2. How to realize those units in practice, i.e just what is a meter?
3. Chains of traceability between a measurement and a standard

In the series "The Science of Measurement" Marcus du Sautoy describes the history and approach taken to nail down the seven fundamental things to be measured. (And I paraphrase the physics.)

• Time - the second used to be a fragment of a celestial cycle and now is 9B+ spins of an atom; we've moved from personal time zones (there were thousands in the US before the transcontinental train) to highly precise coordination. What good would Snapchat be without a common sense of expiry?
• Distance - starting out as the length of the pharaoh's arm and moving to 1 ten-millionth of the distance between the North Pole and the Equator it is now how far light travels is a very short period of time. Imagine an inaccurate GPS. 
• Mass - the weight of pure water, at sea level and at freezing point has been the reference point, this is the last measurement to based on an artifact of something else rather than a fundamental law of physics. Media impressions have been often described in "tonnage" to reflect reach and frequency. 
• Moles - measures how much stuff is involved without worrying about mass or weight. It makes conversions easier to understand and handle. For instance, two parts hydrogen + one part oxygen equals one part water. Gross/Target Rating Points are one way we try to standardize across channels. 
• Light - possibly the easiest to grasp and yet most peculiar, light is responsible for what we see and is defined in waves. The challenge is that our eye adapts to light creating two types of measurements - energy and "in the eye of beholder." What we observe in marketing is altered by both our observations and our biases. 
• Heat - is defined by how fast something moves with absolute zero being the absence of movement. Temperature is simply a measure of activity.  Word of mouth and influence would be analogous to positive heat. Poor customer service would be negative due to the friction it creates.
• Electricity - is all about the flow of stuff from one place to another thru time and space (lightning strikes for example). As evidence of the use of fundamentals, heat can be measured by electricity - the oven probe does it with a thermocouple. The customer journey and path-to-purchase deal with the flow of people with lots of insulators, leaky funnels and conductors.

A key point of the series is that with every increase in the precision of measurement comes a leap in technical disruption with new capabilities emerging.

In marketing, Ashu Garg of Foundation Capital recently wrote a nice white paper on the "Decade of the CMO" in which he articulated the seven most important metrics in marketing.

1. Marketing ROI
2. Customer Experience
3. Conversion Rate/New Customers
4. Overall Sales
5. Marketing-influenced Sales
6. Revenue-per Customer
7. Social Media Metrics

While all good, it would seem we are still in the middle ages of marketing measurement. There are no precise or accepted ways to define many of those metrics on unassailable reference points. Just what is an "experience" or an "influence"?   Even concepts like "sales" can be fuzzy concepts to nail down within one company let alone across companies.

Another takeaway from the series is: figuring this out isn't easy or quick. It often takes standing on the shoulders of giants to make progress so expect this to evolve over time. In the short term, we run the risk of measuring what we can, not what we need. It will take dedication from those who think differently to crack this nut.

Which leaves us with a fundamental question:

What are the building blocks of consumer choice?

Marketing Challenges: Updated

What's changing in marketing?

I stumbled upon a piece in McKinsey's insights section entitled "The Changing Face of Marketing". The author outlines six factors impacting marketing.

1. Customer - "the end users of almost every company’s products are shifting in makeup, location, and number at an ever-increasing rate."
2. Insights - "If knowledge about future customers is essential, and if the quality of the marketing output is materially affected by the caliber of the informational input" then we need more.
3. Technology - "the computerization of many areas of marketing is only a matter of time"
4. Testing - "more controlled experimentation to narrow the odds of an error in making marketing changes"
5. Sales as marketing - [the] "job is becoming less and less the presentation of the company’s product line, more and more the marketing of integrated systems."
6. Global - the problem is "determining how to provide most efficiently the marketing services needed—services that in many companies today are directed, if not executed outright, by a central corporate staff."

This summarizes what we often see in the marketing services world everyday.

It seems that the only thing new is history we haven't learned yet. The article was written in 1966. 

Predicting Music Popularity and Marketing Analytics

What do Taylor Swift's videos have to tell us about marketing?

There's a machine learning competition running here in Salt Lake City that deals with predicting the popularity of Taylor Swift videos from snippets of music. And by snippets I mean 1/10th of a second of a song. The goal is to classify whether a song is a 'success' or 'not' (if you can take 30m views as not being successful and 300m as success.)

Source

In marketing we try to understand what makes a consumer tick, but in this case an audio track has been converted to a series of numbers that look something like the following. Now imagine 3,300 columns of numbers for each sample...

  Success     V1        V2      V3        V4       V5      V6      V7     V8        V9
       0       -5787    9566    -511    -2274 18589   1170    2232   5073   -2578
       1        1067   -  521    1524     -209    -957    -777  -2666     716   -3273
       1          397     -314   -1701   -2568    -463   1123    1041     916      789
       1          784   -1017    2038     4134   1453    3731   3644    2759  -2033
       0     -19632 -13344 -14428 -13195  -3940  -1306  -1206   -1303  -1256

This is where you have to trust the machine and the team. There will be no conference room debate as to what impacts success - the math just works. And that raises an important question: how do we now improve the consumer experience and deliver on the brand promise?

Keeping with the music theme, a musicologist can deconstruct the signals into themes and styles much in the same way that Pandora was created from a music genome. As a result the experts won't be just the ones who can apply a Restricted-Boltzmann-Machine to a classification problem, but those who can translate the results. Interpreters will be in as much demand as the coders.

Keeping my Rosetta Stone polished.

Rainfall is a Path to Purchase

How does rainfall relate to sales?

The weather service now measure things with such precision that we can determine flat (rain) vs. elongated (ice) droplets, etc. A new project on data science competition site Kaggle focuses on predicting how much rain actually hits the ground from a series observations on the way down. The objective of the contest is to estimate the likely outcomes from an hour's worth of data. Such models have implications for agriculture, highway safety, and other resource allocation projects.

Weather now measured in two-dimensions

This sounds like marketing mix and the path to purchase.

Marketing is often faced with a similar problem: how to align out of store activity, digital or otherwise, with in-store sales. Like rain, sales vary with time and space so the analogy is apt.

Recently Yicheng Song and colleagues at Boston University have been working on identifying common path-to-purchase analyses that combine multiple touch points and discrete sales outcomes. Their approach can be summarized as follows:

1. There are patterns to the path to purchase (clusters of consumers emerge along common lines)
2. The path and outcome varies by initial stimulus (catalog, email > online, offline)
3. Paths traverse on and offline steps  (variance is a matter of degree, not one or the other)

What I find interesting is that the approach works with segments and plugs missing data (all too common in path to purchase work) to get a handle on what is likely to happen. Like most marketing analyses these days we don't know anything with certainty but can only talk in directional terms. This means we might be able nudge the pendulum one way or another with a change in the marketing mix but stuff still happens on the way to the store.

Maybe there's something more to be learned from the rain.

Image Recognition - a first look

Does auto-classification of images work?

There is a trend toward "Analtyics-as-a-Service" where companies can leverage specialized functions on demand. There are predictive modeling options like IBM's Watson and MonkeyLearn for text modeling. There are even image recognition options emerging that offer auto-classification.

And as part of a personal interest on figuring out how content works I'm experimenting with various ways of tagging content - images, descriptions, and reviews - to see what might emerge.

So as a test I submitted the following picture to a service I found: imagga.

Red Duvet

So, here's what came back as suggested tags.

• Shopping cart
• 3d
• People
• Man
• ...
• Jukebox

Not exactly what I would have picked...so I did it again.

This seemed to produce more useful tags (and had higher confidence scores).

• Sofa
• Furniture
• Bed
• Interior
• ...
• Relaxation

The tags come with confidence scores - not sure what they mean yet - but the scores for the second picture were higher. Definitely an interesting learning exercise. More to follow...

More Implications of Agility

How will agile development affect your calendar?

In the second and third installments of "what Agile means to the rest of us" posted originally on LinkedIn Tom and I discuss new behaviors you should expect to notice when working with an Agile team.

A new set of questions will emerge and they focus on the business value that a project delivers. The discussion is all about outcomes, not how things are accomplished. This is done to set the right priorities. Gone are the days where available resources defined priorities -- "we can't do that because we don't have {fill in here}, but we can do this because we have {fill in here}."

The only way in business to hit a target is to course adjust - frequently. Agile projects are dynamic and the team adapts as we figure things out. So, clear your calendar and expect to meet frequently.

The new normal

Search Trends, Football and Campaign Measurement

What does football tell us about how to measure campaigns?

A recent installment of Think with Google had a topical piece on football.  In it is the following chart showing the trends in "American Football" broken out by the terms that the search contained: What, Who, When, Where, and Why.

The annual spike is the playoff season.

American Football Searches: What, Who, When, Where and Why

Besides the obvious point that people look more for facts than explanations there appears to be something else.

Episodic events create step changes and the pattern is cumulative.

The latter is somewhat related to the network effect, but with Internet penetration probably close to saturation or at least hitting the second flat slope of growth one might not expect the same growth for the next five years.

What I do find interesting is that the level doesn't go back to status quo ante kickoff, there appear to be slight residual effects.

The marketing implication is two-fold. First, if you want 'superbowl' like results you need to create momentum thru repeated episodes. Second, the window of time in which you measure the success of a campaign should be broken into at least three parts: Pre-Campaign, In-Campaign and Post-Campaign.

Agile for the Rest of Us

What does it mean to be agile?

A friend of mine who has a lot of experience delivering software products based on Agile Development recently asked me about the training course he's developed. I thought it was a perfect setting to explain what the implications of Agile are to the rest of the organization.  If IT is doing things totally different, then we probably should be changing our expectations as well.

The first of several posts appeared on LinkedIN earlier.  The meat of the matter are five behaviors and their impact.

1. Business Value Focus - we need to focus on the what, not the how
2. Dynamic and Adaptive - we won't have a Gantt Chart on the wall, we'll be moving things around
3. Collaboration and Feedback - we can't just attend the kickoff meeting; we're embedded resources
4. Transparency - that wall in #2 is public for all to see
5. Continuous Improvement - we're not shooting for the moon in one step, but taking small steps to learn from

Each or subsequent posts will expand on the five changes above.
If you're interested, Tom Abshire can be found over on LinkedIn or on the AgileFluent site.

Step 1 in Generating Insights: Grab a Marshmallow

How should we approach developing insights?

The typical path to insights is lined with roadblocks of making sure the anticipated results are scaleable, generalizable and solid enough to base a business on. We typically don't go looking until we're sure that what we find is usable and workable. To be honest, we're handcuffing ourselves; maybe there is another path.

In The Marshmallow Challenge kindergartners routinely trounce MBA's at building the tallest structure possible out of spaghetti, tape and string because they do rather than plan.

From: source

Because they aren't riddled with fear of failure, they just try lots of approaches. In 18 minutes they cycle through ideas, learning along the way. In the meantime MBA students, and executives, are planning, allocating resources, and probably doing a bit of posturing to be seen as successful. By the time 'the experts' finally get to the objective, they've run out of time and have to bet the ranch on one tact. The results are predictable.

The rapid iteration of 'test and learn' needs to be applied more often in insights development..this can be in terms of thought exercises "How might it be different?" or allocating budget to learn rather than sell.

Insights are things you don't know you need to know, so how can they be judged a priori?

Touch Points and the Illusionary Path-to-Purchase

How do consumers end up where they do?

We try to understand the path-to-purchase by stitching together data from various touch points and appending them to an artificial identifier meant to represent a consumer.

I'm not sure if it is a love relationship or a hate relationship with technology, but marketers use a variety of data-related tools to connect with consumers and figure out how they came to buy. According to recent Winterberry research on "Marketing Data Technology" (registration required) firms use 12+ different tools on average in the course of marketing. And with that many tools attempting to measure the influence of marketing on behavior something is likely to be amiss.

At a recent conference on data and analytics John Pestana of ObservePoint illustrated the complexity of all this by listing the tags or the scripts on a single web sites that log consumer behavior.

This major media/news site had:

• 18 applications
• 29 tags
• 124 variables

His question: What are the odds that something is broken? Like finding correlations in Big Data, it is pretty much guaranteed. Breaks or discontinuities in data capture mean we have to guess about what happens in between what we can observe. It is worse if different definitions are used for the same event. We've all been in meetings where the sole purpose was to agree on who had the right numbers, when we should be talking about how to help consumers choose our products and services.

The biggest issue is not the quality of the data we do collect, but rather the fact that we're missing all manner of influences and behaviors. Integrating our touch points does not come close to providing a 360-degree view of the consumer.

Maybe we should ask a different question: What if we captured the path first?

Since we're likely to sleep with smartphones no reason we can't imagine passively capturing our path and then overlay marketing on that. Sure there are challenges - permission, revolving MAC address, proximity, etc. But nothing a panel couldn't solve.

I'd rather have gaps in marketing than blind spots on the path.

A Great Combination of Brand, Positioning and Packaging

What happens when it all comes together?

Some times you just smile.

http://www.boxedwaterisbetter.com/

The Millennial Skills Gap

Why don't the best and brightest work here?

This story is based on conversations around a panel discussion at the Utah State of Data, an event discussing the impact of data on education, finance, employment, government, and technology. While localized to Salt Lake City there are some general ideas about positioning and differentiation of a specific market.

Salt Lake City's biggest differentiator is the outdoors. "Life Elevated" is often aligned to what one can do within 20 minutes of downtown. We also have infrastructure - the NSA data center is here so the pipes got to be big, a hospital system that services 10 states and thus unique service delivery needs, and a work force with a decent ethos of helping rather than competing.

But it appears we cannot attract and retain the best and brightest young technologists. The logic woven from several conversations goes like this:

• A career is nothing more than a series of interesting projects and changing jobs to do something new different is common in strong tech markets (a year at Facebook, two at Twitter, a startup and then a new product launch and sabbatical at Google). Interns brought here don't convert because there aren't enough big cool things going on like the Bay, Seattle, Boston, Austin, etc.
• Job hopping creates a rapid rise in salaries for high-demand skills, and not only in technology, as companies compete for scarce resources with great credentials. Within a market it is business-as-normal, but across cities it makes "market-competitive compensation" surveys from your HR team irrelevant. If I'm willing to move, then the market is the country not your locale.
• SLC is not a "Friends" city where millennials conduct life 24*7 in coffee shops and other urban venues. We'll be lucky if we get to 18*6.  So, even if we get the best there isn't a lifestyle typically associated with high energy folks.
• We are good for families so we can attract and retain those wanting to settle down a bit, i.e., those in their 30's or 40's. They have experience.
• At that stage of our lives job hopping isn't viewed as a means of defining oneself. And without movement, salaries don't rise based on intense competition.
• If salaries aren't competitive we can't attract the talent we desire.  

And the circle begins again.

It would seem we're caught between our aspirations "Silicon Slopes is next Silicon Valley" - but based on deal flow funding so can Saint Louis, Munich, Seoul and Cincinnati (Beijing tops the list in terms of number of deals.)

So what kinds of companies should we create when the workforce is a) experienced and b) content? Rather than core game changing technologies, are there other classes of things we should focus on rather than the "next Google" (and good luck with that anywhere.) 

• Can we rent company founders and focus on later stages of the evolution requiring experienced management?
• Can we blend technology, the willingness to help and experience to solve human needs?
• Can we take what is unique and different and position ourselves better?

5 Problems with Marketing Analysis

What do we often forget to mention?

I missed an important birthday last month; the banner ad turned 40.  We're certainly not going to see the 40% CTR that this ATT ad achieved back then.  In fact we're not likely to see much above a single digit or even decimal place.

At the time I don't think anyone had an idea what the click thru rate would be, but in today's data-driven world we should have some clues.  We have all kinds of systems and analytic tools that predict what a CTR might be. But each of those come with one or more constraints that we often forget to convey.

1. Time, data and resources will narrow the options we can use to analyze a problem; need an answer today? or next week?
2. Explanations always rest on assumptions as we reduce the problem to one that is manageable; did we remember to tell anyone?
3. Our mental models are biased to begin with, and we may not realize it.
4. Data is left behind due to data collection procedures or application code; design choices made before we started the analysis limit what we can and can not do/say.
5. Our use of data may be different than how it was intended; this is a common question when creating mashups particularly with public data.

So, say out loud "this is what we don't know"....

The Need to Link Data to Marketing

How do we come to understand things?

In the era of "big data" this and "big data" that, we have a problem. Our minds are wired for language not data. Rarely do we have direct intuition about data, what it means, and what we ought to do with it.  We need to invent the Rosetta Stone that bridges the hieroglyphics of machine learning with the ancient language of marketing.

Consider the following examples:

• What is the cost function of loyalty?
• What does unsupervised learning tell us about branding?
• Where does gradient descent take us in terms of understanding engagement?

On the left side are data terms; on the right are marketing terms.   Each takes a common, yet ill-defined marketing idea (loyalty, branding and engagement) and maps it to an aspect/method of the big data world.

• Loyalty has a cost (or a profit) that comes at a price. What is the allowable marketing cost of achieving it?
• Branding is often described as being in the mind of the beholder; thus we should not label things as A or B but rather be surprised by how consumers cluster things.
• Engagement is a cumulative set of interactions that may (or may not) move us toward conversion; we need to understand the marginal contribution of each step.

Only if both audiences, data folk and marketers alike, can ask and answer these and a myriad of similar questions will we be able to translate from one world to another.

Digital Differentiation and The Chore of Choice

How does differentiation apply in a digital world?

This post was inspired after reading "Digital Brand Differentiation - 3 Frameworks to Follow and asking the question above.

Differentiation is required when we are faced with the chore of choosing.  It works because it provides a focused benefit that we can internalize and compartmentalize as "that does this".  So when we have "this" need we lean toward buying "that" product.  It is a step toward branding.

The qualifier 'digital' raises some questions to consider - some apply to digital products others apply to digital tactics for analog products.

1. What new dimensions are available upon which to differentiate 'digitally'?  (Which old ones don't matter any more?)  
2. How does digital information access and transparency change the relative importance of 'being different' in the decision process?
3. Are there categories where it is more/less important to be different because of digital?  Why so?
4. With the digital consideration set now being defined by the amount of time we invest, is it more important to be the first to be perceived as satisfying the basic need rather than doing it differently?

Jack Trout's "Differentiate or Die" is a good place to start on the first question since it highlights 20+ dimensions upon which create that distinction in people's mind.  Written in 2001 it probably doesn't cover the full implications of digital so could use an update from a digital perspective.  For example, I don't recall it using communal content (social networks or reviews) as the basis of differentiation.

A counter argument to the differentiation argument was offered by Killian Branding in a post entitled ""Differentiate or Die" is dead. RIP, USP".  That post concludes with the statement "Visibility is now more important than differentiation. Mission One: Get into the selection set."   To paraphrase the logic, if you're not at the top of the list (Yelp, TripAdvisor, Google, etc.) it doesn't matter how you are different.

What makes digital different is that the consideration set emerges/contracts/grows over the decision process - it is no longer an all or nothing proposition of being in front of the toothpaste shelf or looking at Thai restaurants in the yellow pages.  This would suggest that the emphasis on differentiation may rise/fall across the journey based on inferred intent.

The 5 Persona of Big Data

How do we decide WHICH Big Data is right?

The stories about Big Data talk about it as if it is one thing; it isn't. Big Data actually encompasses several different universes based to some extent on the fact that there are different underlying technology that can be deployed. And since technology is versatile but not omnipotent it makes some sense to first try to articulate what one is trying to achieve rather than just saying "let's do Big Data".

And any time there is hype like we're seeing today a gap emerges between what the business thinks it wants and what the technologists believes it can deliver.

To help bridge that gap it might help to describe the various approaches as people with skills and characteristics. To provide context, I'll start with the old guard and then introduce the new characters.

• Rachel - is the oldest with a very structured and ordered view of the world, her motto is "everything has a place, and everything is in its place" and put there immediately. She doesn't handle ambiguity very well and her rigid out look makes it difficult to change things as the world around her changes. She struggles when either too much complexity exists or the scale of the problem is too big. We all know Rachel as "relational databases" introduced in the '80s/'90s when computation, storage and memory were all expensive. She is showing her age. Think business intelligence and MySQL or Oracle.

• Chloe - views the world thru a different lens. Rather than looking at everything as well-structured events like Rachel, she is more interested in how groups of those things overlap on key attributes. She is less interested in "how many" than in "what are the commonalities/differences" in a set of things. In addition she clearly recognizes that every thing isn't known about everything so is quite comfortable with uncertainty. These holes in the data create a sparse matrix that is often best handled by "column-oriented databases". Think highly variable data elements and GoogleBigTable or Hbase.

• Kayla - looks at the onslaught of data and says "I can manage this if I spread it around across as many buckets as possible." Her goal is to preserve things as quickly as possible and to make sure they are retrieved upon request. She defers understanding and definition to others. In the purest sense "key-value pairs" make the data opaque by design, trading insight and thus latency for scale and availability. While excellent at writing information out her challenge is trying to organize and integrate data from several processes that all use the same term to mean different things. Think extreme write and access of user-supplied data and Apache Cassandra or Amazon Dynamo.

• Danielle - has probably the most unstructured view of the world "everything is a chunk of stuff and tags". Success then is dependent on the completeness and accuracy of the descriptors assigned to a chunk since searching its entire contents can be costly. The unstructured view applies to connectedness as well as content. In the "document store" world there are few relationships between the chunks beyond the tags. Consider posts tagged with #BigData - the hashtag is the only commonality. There are no implicit relationships beyond that fact. Think content management with mongoDB or CouchDB.

• Grace - is the most social and believes relationships are everything. She thrives in scenarios where the existence or nature of a connection matters more than what is connected. This focus makes it very easy to extend the universe in new ways as new things happen or new actors emerge. Grace exists as "graph databases" that have been around for a while, but most noticeably in social networking and social sciences. Think recommendations, communities of interest and Horton or Neo4J.

Which persona we adopt depends a lot on what we want to accomplish. And it is quite possible we'll end deploying more than one solution in our quest to "do Big Data."

{re}thinking with Data

How do we go about finding insights?

Let's start with what we mean by 'insights' -- here's my working definition in the context of marketing:

the identification of a previously unknown connection between marketing activities and consumer behavior that changes how we align our solutions we human needs.

Or, simply the finding of something "I didn't know I needed to know."

The typical story line for doing analysis consists of the following chapters:

1. State your objective
2. Define your strategy and tactics for achieving them
3. Create a measurement framework of KPI's that defines success
4. Articulate the dimensions of the business/behavior you need to understand
5. Set targets or goals that you need to compare to in order to track profess

And along the path you're defining a technical implementation and data capture plan to get from #1 to #5 and then refine.

If we're not careful, there are several risks in that plan.

First, we don't do all the steps...there is a tendency to jump in and build something; we forget to ask "why are we doing this in the first place?" enough times.    Think about the difference between Cliff Notes and the real thing.  A good essay requires reflection.

Second, we prematurely narrow the list of potentially valuable options.  This "focusing illusion" creates a bias because we tend to look no further than the first idea.  This is akin to judging a book by its cover. 

Third, we often view the problem in terms of outcomes related to our current business model not what might have caused them. While it is imperative to have metrics to track, they are simply links between behavior and performance.    Conversion rate is not a behavior.

The remedy to these risks is to spend more time thinking and that requires deep domain knowledge as well as the ability and willingness to explore.

For a good read try "Thinking with Data" by Max Shron