Master Marketing Profile: New from Adobe

Just how they do that?

At the Adobe Summit this week several core services (much better construct than 'shared services') were introduced.  While not terribly sexy, they do go to the continuing evolution of an enterprise platform.   One of the new elements is "Master Marketing Profile" described as a single identifier or a 'ring to rule them all.'

Conceptually I'm intrigued since we have a need to look at audiences from the perspective of a publisher as well as a marketing services firm.  And as with any new service the gap between the idea described in the opening session on a 50' screen and the explanation of it on a monitor at the booth was quite noticeable.  We had questions and it felt like the answer was "I know a guy who knows a guy" who wasn't around at the moment.

So I thought I'd post my questions.   And to provide context, let's start with the following scenario.  I want to access an audience based on the following:

• Activity on our owned and operated sites using only content about a category of interest
• Traffic and events on the networks we represent that fit an 'audience extension model'
• Email and print subscription information for that category from a related line of business
• Loyalty and transaction based information from a retailer who promotes items in the category
• Social graph and interests from an advertiser's customer base that has expressed interest in a brand in the category

While going from site activity to testing on the same site seemed to be the emphasis in the beta testing, the above real-life problem generates a series of questions about the Master Marketing Profile.   I've tried to frame then in general terms, but they are biased to specific Adobe products/constructs.

1. Is the identifier built bottom up, e.g. a cookie, or from top down, e.g. total business?
2. Is this the commercialization of the 'universal ID' concept that existed in professional services?
3. If it is the master, is it safe to assume the existing identifiers are slaves and still remain active?
4. Is there one or minimum set of IDs required to build the master? 
5. Must all child identifiers be an existing Adobe ID?
6. As a core service, can I use it without analytics?
7. Can I define the master to be the data management platform ID or even a client or advertiser's identifier?
8. How are conflicts of information handled when dealing with multiple touch points?
9. What is the logic sequence for aligning the above sources into a master?
10. How will we know whether the process works or not?  
11. What percent of a given source can we expect to be mapped?
12. How do we ensure orphans remain viable entities over time if they are not in the 'audience library'?
13. How do we know which sources were used in a given profile or segment?
14. Is there a probability of belonging associated with each source of contribution?
15. How does the mapping process impact the creation of downstream business rules and event triggers?
16. If a map is built with incomplete information, how do we avoid garbage in gospel out?
17. Is membership in a segment based on this identifier a "Yes" "No" proposition, or "Maybe"?  (What is the likelihood that this identifier really does represent what we think we want?)
18. Is this a precursor to another consumer pool and compete with Axciom, KBM and Liveramp for matching services?
19. What information is in the payload when shipping segments?    
20. Does the process provide an audit trail of the mapping such that it can be defended?
21. Where is PII used and then masked in the process?
22. How are cookies, digital fingerprints, and device ID's utilized in the process?
23. How confident should we be we won't piss off consumers because of a technical black box?

If I get answers, I'll post them...

Impact of Store Brands on Store Loyalty

Why and when would I be likely to choose a store brand?

A recent article in the Journal of Marketing analyzed the relationship between the amount of money a consumer spends on private label brands and her loyalty to the retailer.  Store brands now represent 15% of global retail revenue according to AC Nielsen.  The idea that as I add more store brands to my basket (larger share of wallet) the higher my store loyalty will be is not a new one.

But as with all aspects of marketing, different segments and different categories work differently.  The relationship between private labels and loyalty is a bit more complex than 'the more the better.'

The research identified several factors that impact the relationship between the two consumer metrics: Private label share and Store loyalty.  In order of importance:

1. Low price seekers exhibit a stronger relationship between store brand share and store loyalty.  This is in part due to the tendency that I think of private label brands as less expensive to begin with.
2. When a retailer occupies a lower overall price positioning in my mind then as store brand share raises so does loyalty.  A collection of store brands from a low-price retailer makes it easier for me to decide.
3. Categories with high-involvement, or information seeking, are areas where the relationship between the two is also stronger. This suggests that unique content development should focus on categories where I need to find out things.
4. Highly commoditized categories, i.e., lots of acceptable alternatives available, mitigate the relationship between share and loyalty.  Basically, if all things are equal I might as well just pick one.

Thus, when it comes to distributing promotional content both the audience and the offer matter. 

5 Steps to Accessing Audiences

How do you find me on the Internet?

There is a lot of talk about audience extension, or the desire for a publisher to increase its value to advertisers by reaching site visitors elsewhere on the Internet.   However, the real business need is accessing an audience - not just reaching them off-site.

The idea of accessing an audience requires thinking differently about traditional content and visitor concepts.   Since we're no longer interested ONLY in what visitors do on our site, we have to rethink tagging, signal detection and the organization of traits/segments.

To illustrate, we (and yes I work for a publisher) have tags related to the existing content and its purpose for a given site.  But when looking across sites similar content may have very different contexts.  Add to the mix that similar looking tags may also mean very different things and you have a recipe for confusion.   And speaking of recipes, here's a set of tags that make perfect sense when the content is in fact a recipe:

meta property="article:tag" content="Kid-Friendly"
meta property="article:tag" content="Spring"
meta property="article:tag" content="Vegetarian"
meta property="article:tag" content="Bake"
meta property="article:tag" content="Pitas"
meta property="article:tag" content="Cheese"
meta property="article:tag" content="Tomatoes"
meta property="article:tag" content="Cabbage"
meta property="article:tag" content="Pineapple"

I can almost picture precisely what those tags mean.  Something like....

Rainbow Veggie Pizzas

But what would those tags mean if used elsewhere on the Internet?  Without more context the tags "Spring" and "Bake" are terribly ambiguous.

The challenge is ensuring content and tags are both used in a way that can be useful to reach audiences.  

From a text book perspective, the five-step approach for accessing audiences should be:

1. Develop the segmentation strategy – which audiences make the most sense to access across touch points.
2. Standardize the concepts – what ideas are core to the audience and how common are they across the entire spectrum? (dare I say, build a taxonomy)
3. Modify the tagging – allow the tagging of content to be both dynamic and link to the common concepts.
4. Capture the signals – actually, it is the classification of the signals that is more important.   We need to understand audience signals rather than site-specific signals.
5. Organize the traits – being able to build segments from pan-site signals requires a new way of thinking and organizing things.

For instance, imagine creating a 'fashionista' segment across food sites, style sites, and financial news….one can imagine linking 'formal dinner parties', 'walk-in-closets for shoes' and 'fashion week' as potential evidence of interest in fashion.   In a site-centric world the tagging strategy would be obvious and literal; in an audience-centric world the tagging is a little more about concepts and human judgement.

The reality is that we actually have a chicken and egg problem:  We need to collect data first in order to determine if we should/should not combine different traits into a valuable segment.  This is very much a test and learn environment.  And I'll save that discussion for a later post….

There are some good resources that explain more of the mechanics of audience extension; for instance AdMonster recently issued a of playbook on the topic.

Tim's Vermeer - Go See It

How does a painting come to be?

Last weekend in Toronto we went to see "Tim's Vermeer"...the story of Tim Jenison's quest to figure out how what could be the best painter - ever - did it.   What made his paintings stand out and stand the test of time (as in several centuries)?

The answer to 'how' is that he most likely had help in the form of mechanical aids - a simple lens and a simple mirror.   This combination allows the painter to ensure that the color he paints matches the color of the setting.   Now, there is no letter from one of his contemporaries like Pieter de Hooch saying,

"Dear Johannes...can I borrow your studio and lens/mirror contraption for the next 18 months to paint, can't seem to get the light quite right here across the river."

But as the scholars in the movie discuss, the painting itself gives clues on how it might have been done.  The light raking across the wall, the curve in the seahorse's tails most likely wouldn't have been painted the way they were - without help.  Using tools of the era, including grinding/polishing his own lens, Jenison takes us thru the five years it took to go from concept to final painting. 

The painting in question is "The Music Lesson" hanging in Buckingham Palace.

Is this the original, or the one created by a man who didn't paint?

ABCs of Successfully Distributing Promotional Content

How can we improve the odds of consumers paying attention to what we have to say?

In the development of anything dealing with the marketing of promotional content (disclosure: part of my day job), there are three challenges to overcome:

1. App Abandonment: with less than 25% of mobile apps being used more than once, there is obviously no silver bullet from a functional point of view.
2. Banner Blindness: recall, relevancy and results are often in the realm of single digits or lower making the shift from media tonnage to targeting an imperative.  We simply don't remember what we saw.
3. Coupon Clutter:  the impact of 329 billion coupons being distributed in the US alone has to drown some segments to the point of not even wanting to deal with coupons.

Since a PhD dissertation in 1999 first described the irony of consumers skipping the creative designed to help then find the information they were looking for there has been a lot of research into what makes for engaging experiences.  There is even an organization dedicated to banner blindness that quantifies the impact tuning out.   (Great infographic here.)

All this suggests consumers (or a least a sizable segment) don't want to deal with the hassle - they just want the benefit.  This point is confirmed in Inmar research (free registration required to download.)

There seem to be so many rules to coupons. I don't want to have to carry around the store coupon policies in order to get a deal.

So, entering the market are digital coupons of two types - Print-at-Home and Load-to-Card - to help remove the friction in the process of saving money.    The research ends with a consumer survey stating the obvious: Consumers want "easy" couponing.  More specifically, they don't want to do it -- they want somebody else to ensure they get the benefits.

• 65%  "I want coupons loaded to my store loyalty card for products that I normally buy."
• 65%  "I want stores to email me with coupons for products that I normally buy."
• 58%  "I want all available manufacturer's coupons to be loaded onto my store loyalty card."
• 63%  "I want all available store coupons to be loaded to my store loyalty card."

Q5.10: Now I want you to think specifically about coupons that you can access online using your computer, mobile phone, or tablet.  Please rate your agreement with the following statements (top 2 box -- Agree/Strongly Agree).

The answer to the opening question appears to be another question:  Should we identify a loyal segment where we can simply take coupons out of the equation?

Viral Marketing of Apps on Facebook

What form of sharing should I use?

A recent article in the Journal of Marketing analyzed the success (or failure) of 750 Facebook apps in an attempt to understand how various mechanisms of social sharing impact acceptance or reach.

While a lot of ink has been given to influence (Gladwell) and seeding-strategies (Watts) this research looked at what tools marketers can deploy to facilitate sharing.

Specifically, marketers can choose to use one or more of the following tactics

1. Unsolicited Messages vs. Solicited Message 
• A message appears in the inbox about a new app vs. see what apps appear on a members "About" page
2. Messages with incentives vs. those without
• "Try and get a month free" vs. just "Try it out"
3. Direct messages from friends
• Communication among 1st degree contacts (similar to forwarding an email to a specific individual)
4. Broadcast messages from strangers vs friends
• Timeline posts viewed by 2nd to n-degree contacts on others vs. own timeline

The analysis focused on how those choices impact the adoption rate of leisure vs. business apps (or in the language of academe - low utility vs. high utility). 

The key takeaway: the techniques that garnered 100m users for FarmVille in roughly 40 days would be counter productive (and possibly detrimental) to a business-oriented apps.    From the research:

The very mechanisms that made FarmVille so successful is a recipe for failure when used in a different product context.  Unsolicited and incentivized broadcast messages from friends are the least effective sharing mechanisms for primarily utilitarian [business] products.

You rarely see such strong language in journals.

And speaking of context, the results would theoretically be different on LinkedIN because it is by nature a business/high-utility social network.  The mechanisms of how people choose, and what information cues they use, differ depending if they're looking for a job or playing Candy Crush.

So, as we design campaigns with "Share This" functionality, we need to understand what the usefulness of the app/content is as well as the distribution platform and choose our tactics accordingly.

Shift to Digital: Part 5 - Operational Savings

Where does the money we want to shift come from?

The previous two posts in this series focused on Driving Growth and Building Leadership, but at some point we need to think about how to free up resources to dedicate to the new mix.   It is great to have a vision of two pie charts that split out the entire marketing budget into completely different approaches.   But we do need to find ways to allocate a reasonably fixed set of resources - moving 50% of a budget is vastly different than growing a budget by 50%.

So, here are a list of questions focused on operational and tactical topics.

1. Where is marketing needed vs. not needed?
2. Where is the ROMS (Return on Marketing Spend) strong/weak? 
3. What business and geo-demographic elements explain the variance in ROMS?
4. What is the value of each action leading a sale – by channel/source of traffic? 
5. What is the value of social sharing? (Interesting facts for e-commerce here.)
6. What is the value of an “Offer View” – by channel/source of traffic?  (single offer centric)
7. How do we align the digital footprint with terrestrial sales activity?
8. What value would creative optimization bring to improving conversions?
9. How can we leverage targeting, bidding and interaction history to reduce ‘wasted impressions’?
10. What is true cost of execution of a digital budget?
11. What is the financial contribution of individual touch points?
12. What are the reach, frequency optimums by channel and segment?
13. What is the difference in cost models of reaching consumers on-site and off-site?
14. To what extent does spend on loyalty programs offset ink and airwaves?
15. What is the most effective means of reaching a consumer/segment?
16. Where else can we amortize fixed costs of content production with the most bang-for-the-buck?
17. Which data sets should be integrated in which order to gain efficiency?
18. What is the dollar value of what we know to our suppliers, vendors and market in general?
19. Which loyalty or transactional segments work best for targeting?
20. How many digital impressions does it take to replace a TV ad or a page in a flyer?

I don't know if the answer to that last question is 100,000 or 1,000,000 or even 10,000,000 but it seems to be a worthy goal for an analytic program to focus on.

There are a variety of ways we can approach the shift to digital, but hopefully these 60 questions help provide some guidance on how the analytic program could be developed and managed.

Shift to Digital: Part 4 - Brand and Category Leadership

What position do we occupy in the mind?

This is the continuation of thinking about 'the shift to digital'.  The last post on the topic listed twenty questions that need answers when thinking about Driving Growth.  This set of questions focuses on what it might take to establish brand and category leadership.   As such, they focus a lot more on understanding consumers and customers because the mind is the most difficult thing to change.

1. What segments emerge from existing marketing activities?
2. What roles do existing product categories/segments play in driving store traffic?
3. What categories impact share of mind, share of wallet and cart size?
4. What segments exist based on how people shop?  (touch points, timing, content consumption)
5. How much and what type of information is needed to impact choice?
6. What digital metrics correlate with or predict sales?
7. How do consumers use shopping channels (store, catalog, ecommerce) differently?
8. What level of penetration is required for a loyalty program to impact store sales?
9. What type of cadence generates the desired behavior by segment by category by channel?
10. What consumer segments based on loyalty are a good proxy for the entire franchise?
11. What level of personalization moves the needle at an acceptable cost?
12. Does the pattern of consumer interactions mimic merchandising allocations?
13. What jobs need to be done by the consumer that have the most impact on transactions?
14. Do some categories do better in printed flyer, and others in the digital channels (flyer, email, mobile, display)
15. What is the impact of the gap between “Say – Do” on sales?   (eg add to list, but not buy)
16. How much of the journey do we need to see in order to impact choice?  
17. What is the lift associated with personalized offers based on transaction history?
18. What is the value of own-brand vs. manufacturer brand in driving behavior?
19. What is the contribution of owned, paid and earned media?
20. Where do out customers go to find information?

For each question there is a set of data than can be cobbled together to help answer it.  A key element is being able to reduce digital interactions to a point, i.e. geo-locate the activity.    And that space will be the subject of another post - How do we know where it happened?

Attribution: A Mixed Model

How do you attribute offline AND online marketing at the same time?

Giving credit where credit is due is hard, particularly in the the world of marketing.  To help me think thru this I mocked up the classic 2 by 2 matrix.

There are two broad approaches to attribution: Top-down and Bottom-up: as well as two different purposes: Tactical and Strategic.   This results in four ways to approach the problem; this doesn't mean there are four techniques - there are a myriad of them in each cell.  The example I'll use is the flyer or circular (disclosure: my day job focuses on this) in the context of reallocating budget from the printed version to the digital world.

Top-down approaches looking at the tactical level grew out of the Marketing Mix Modeling world.   Econometric models try to tease out and control for the effect/variability of marketing activities.   While some testing is often done, this approach won't support the needs of an organization looking at bold strategic moves (like re-purposing 50% of its budget) to something new.

In the digital realm, and in particular the e-commerce, bottom-up and path analysis of a specific consumer is often the basis for attribution.   Whether it is 'last click' or a more advanced form of attribution, it is still limited to a limited domain.

The 'shift to digital' series (posts before and after this one) focus on a big change where both styles of attribution are required.  So, the challenge is figuring out a way to link marketing mix models which can handle flyer distribution, OOH, and TV with their digital equivalents that work with anonymous, PII and segment data. 

It seems to me that the variance of one might be explained by the other....

Shift to Digital: Part 3 - Driving Growth

What questions need answers?

In the continuation of a series of posts on shifting major portions of a marketing budget around, this post looks at the goal of "Driving Growth" and lists out a series of questions that could form projects.

Since growth focuses on results, the questions in this post tend to focus on monetary issues.  I'll save the consumer/customer questions to a post on building brand and category leadership.

The list:

1. How do sales change with respect to specific marketing activities?
2. How much of the effect of a tactic is offensive (lift/incremental) and how much is defensive (baseline/erosion)?
3. What is the sensitivity in sales to different marketing mix allocations?
4. What is the source of promotional sales? Brand, category, store, net-new?
5. What level of digital air cover is required to replace offline air cover?
6. To what extent to national decisions impact local sales?
7. What is the interaction between tactics?
8. Where is there headroom to actually grow the business?
9. What role does competitive presence play in our sales?
10. How much of a store's growth is controlled by the organization?
11. How can content be repurposed and distributed to impact sales?
12. What friction in the current business model prevents sales?
13. Which current non-digital steps could be (should be??) made digital?
14. How can decisions in-store be facilitated with mobile content and features?
15. What is the role of communal content (reviews, social) in making/speeding the choice process?
16. How should budgets be allocated geographically?
17. How do we forecast or extrapolate from a test to a broad roll out?
18. What digital/retail trends will work for us over 3-5 years?
19. What is the allowable marketing cost of each tactic in each area?
20. Do the puzzle pieces form more than one marketing plan?

All of the above require thinking about the holistic environment in which a store operates rather than looking at it from a channel perspective.   In the end, this is a problem of allocation where not all the information is at the same level or potentially not even capable of being integrated.  Our job is to reduce the risk of moving 50% of a budget away from one tactic to a collection of others by quantifying as many of the variables as possible.

Open Data and the Organization

What might open data do to the organization?

The trend of using accessible data from outside the organization continues.  McKinsey recently quantified the economic opportunities of using what is termed 'open data'.  They're big, but you'd expect that from a management consulting firm.   Tim O'Reilly describes open data this way:

There’s a pragmatic open and there’s an ideological open. And the pragmatic open is that [data is] available. It’s available in a timely way, in a nonpreferential way, so that some people don’t get better access than others.

Some implications:

1. The lack of control, and the potential for change, means those providing business requirements need to more like mentors and docents than hardliners and dictators.  Much more emphasis on thinking about 'why we will be successful' (strategy) rather than 'how we will accomplish it' (tactics.)
2. The value proposition may shift as organizations find they may have data that they want to contribute to the community.  These may be byproducts of processes that spin off data, e.g. geo-location and timing of distribution activity, that others may find beneficial or acting as a broker for a consortium of data used in benchmarking performance.
3. The skill set of marketing department will include a collection of hackers responsible for finding novel ways of identifying and satisfying market needs by combining internal core competencies and any/all external supporting data.

In short, open data will force organizations to clearly understand what its purpose is...others, will be using the same data to do similar things.

5 Analytic Steps in the Shift to Digital

How should marketing approach shuffling the budget?

Business objectives haven't changed a whole lot over time - drive growth, build brand/category leadership and create operational savings.    Broadly speaking these goals end up requiring shifting money around based on three steps.

1. Freeing resources from non-productive programs,
2. Identifying how marketing impacts key metrics among key segments, and
3. Re-purposing marketing dollars to more productive programs

Transferring money around at a tactical level, e.g. spot to national TV, display to search, print to tablet, and understanding the impact is fairly common and a tractable analytic problem.  But in a world of stagnant same store sales imagine that 33% or more of the largest budget item is focused on a single offline tactic, e.g television, flyer/circular, or out-of-home, and that money is put on the table as the foundation of 'doing things differently' and moving to digital.

We're not talking about our discretionary spend or bucket for experiments here, we're talking about betting the farm. 

In this case, there are numerous known unknowns, like 'what is the sales impact if we don't do that anymore', but also a very real potential for unknown unknowns - things we simply don't know to ask about yet. This begins to sound like a three-to-five year program, not a single campaign, that has as many cultural and process changes as it does tactical ones.

To help break down such a large problem from an analytic perspective, it is helpful to identify specific programs that shape our understanding and thus our planning.  Here's my initial categorization:

1. Document "Cost of Sales" attributable to each key tactic
2. Identify productivity of each tactic by geographic zones, e.g. store trading area.
3. Develop consumer segmentation model(s) based on how they choose and decide
4. Build a model for how different types of content work
5. Define how wholesale changes in the media mix impacts sales

Each of the above categories can then be further broken down into analytic projects, data-related activities as well as in-market pilots.   These will be the focus of the next set of posts that will list out 15-20 specific questions per goal marketers should be asking their data teams whose single biggest contribution to the transformation will be reducing risk by quantifying uncertainty.

(Disclosure - this problem is something we're facing at work and I'm sharing my thinking process of how we're helping clients get from A to B.)

This is the second part of a continuing series about the 'shift to digital'.  It started here.

The Shift to Digital

How do we get there from here?

Boston Consulting Group's 50th anniversary includes a survey on what troubles leaders the most.  (Can't find the source document, just the press release - so using the eMarketer blurb.) And for a 'consumer insights' professional it is good to see "Leveraging Customer Data" and "Digital Channels" in the top five, right behind "Open Innovation" and "Large-scale Transformation" and above "Distinctive Business Models."

There is nothing prescriptive about those items.  In the context of strategy development, they are broad constructs that shape or focus thinking - they are goals, not defined objectives or outcomes - those have to be company specific.

The eMarketer article highlights the challenges by industry and there are some differences.

• Consumer and Retail companies are most interested in growth in their current markets whereas Industrial firms are looking for new markets.
• Consumer and Retail firms also rank Leveraging Customer Data the highest.
• Digital Channels shows a wide swing (Technology is highest, Energy and Environment is very low).   C&R is above average.

So, weaving things together one can start to imagine a program-level plan to "Shift to Digital" for consumer/retail firms that is constrained by competitive activity, lack of customer insights, and a marketing budget that includes both on and off line tactics.  

In my mind, such a plan would have three parallel objectives:

1. Drive Growth: re-purpose marketing dollars to more productive programs.
2. Build Brand/Category Leadership: Increase penetration, share of wallet and loyalty among key segments for key categories.
3. Create Operational Savings: free resources from non-productive or sub-optimized activities (and not just marketing.)

Achieving these objectives based on what we know is hard enough, doing it while we are still learning what we don't know will prove to be interesting to say the least. 

So, this post is going to be the start of a series around the questions that need answers as firms make the shift.  It may even provide some ways to answer them along the way.

Stay tuned....and hope it helps.

Mad Math Men

Why do we need more of them?

The newsletter, conference, webinar circuit is full of content about Data Scientists.  There are articles on how to hire them; how much to pay them; how to organize them.  There are surveys (and client conversations) about the unique value of data - "it is a great asset" and "it is the lifeblood of our company" is heard in every board and conference room.  But there remains a disconnect in connecting the bits in order to create value.

So, what should we be actually looking for?

The nature of the work often described as 'big data' focuses on finding answers to ambiguous questions: How can we turn planes around on the ground faster?  Can we identify life events from shopping patterns?  Can we discern intent from digital signals?   To re-purpose a phrase from Jim Cooper's thinking around building companies these are the BHAP's or big, hairy, audacious problems.

To solve them requires a team that covers data, computer science, and math - but also, and possibly more important, curiosity, experimentation, and creativity.  The first set of skills is what people typically think of in terms of 'data scientist' or Math Men.  The latter set of skills come from a wide variety of non-traditional disciplines.  Art history, psychology and music backgrounds are just as valuable in providing insights into these problems precisely because they've never been solved before.  The ah ha moment that changes how we market is analogous to the big idea of creative agencies and the Mad Men like Don Draper.

In marketing this combination appears most often in the digital realm - understanding the confluence of context and content and how those interactions shape response.

Add to the mix subject matter or domain expertise and you have the makings for a Mad Math Man.

A title I aspire to....

Segments are the Lingua Franca

How can digital and direct marketers see the same 360 degree view?

The tracing of a consumer's journey across the shopping journey is as much a guessing game as it is a sure bet.   The idea of creating a 360 degree view actually requires making a lot of guesstimates.  To help think thru the idea I look at several different layers of the problem.

• Anonymous - the bottom of stack where we know something happened, but have few (if any) signals that can help us link things together.  Online this could be visitors with cookies blocked; offline it is store browsers.
• Recognized - there are enough signals present in an interaction to begin thinking about profiling, segmenting or linking them.  Web visitors, device IDs, shopping baskets, etc. form the basis of recognition. 
• Credentialed - a semi-persistent identifier exists that increases the odds of knowing it is the same person.  Account information such as email, log-ins and tokenized credit cards fit this model.
• Identified - the persistent identifiers, e.g. name, are known and usable. 

Thinking and working in all four layers helps with creating the desired view.  So far, there is no silver bullet or a right order of approaching things.  Sometimes we start at both ends and work toward the middle; other times we work on one area in particular.   Campaigns may be designed simply to help make some linkages, e.g. newsletters or white papers to trade value for value in order to improve Credentialed. 

Digital marketers tend to understand this stack quicker than the traditional database marketing crowd. The common ground between these two worlds is the use of segments.  While done for very different reasons (eliminate PII vs. allocate scarce resources) the marketing thinking is very similar - find a homogenous group that responds similarly to marketing messages.

In fact, shipping segments will get you to an integrated view faster than trying to work at the lowest level where natural constraints about what you can and can not do throw up roadblocks.

An interesting exercise would be to think about mapping and modeling consumer vs. customer segmentation schemes.  (More on that in a bit.)

Customers are Simply Consumers with History

Is she or isn't she a customer?

Very often we think in term of consumer marketing versus customer marketing; one focused on acquisition and the other focused on retention.  We even allocate budgets as above and below the line further separating one idea from the other.  However, in the age of showrooming, the distinction between being a consumer and a customer is more often down to a fleeting moment in time and a choice.

Will she buy?

Similar to Schrodinger's Cat of quantum physics that is both alive and dead until we open the box and look in an individual exists in both states until she actually buys.  Since it is just one individual, we might want to consider 'customer' as only an attribute or descriptor rather than a separate and distinct class.  Clearly campaigns can select an audience based on that knowledge, but in the world of 'omni-channel brand experiences' the messaging and marketing needs to be integrated.   One of the worst things a brand can do is create cognitive dissonance thru conflicting programs - like offering discounts to new customers while not rewarding or recognizing loyal ones. 

There are only consumers, some of them just happen to have a history of transactions that we can see in the rear view mirror. 

Old and New School Marketing

What can policy wonks teach us about big data?

In a great essay in Foreign Affairs on "The Rise of Big Data" the authors describe the implications and meaning of it all.  Since that is a premium article, I'll summarize a few key points. 

The sacred cows of the analytic kingdom from research to finance have rested on three tenets.

1. Quality of the data
2. Representativeness of the sample
3. Causation of the outcome

For someone who started out in market research, has taught statistics and has an MBA in Finance, these were the inviolable crown jewels - something debated as much as the findings themselves.   However, in the world of 'big data' none of that really matters any more.  In fact,  we now use the terms..

1. Messy
2. All
3. Relationship

as the lingua franca of the realm.   The reason: damn near everything has been datafied - a term the authors use to describe process of reducing everything to a stream of data.  Correlations of events based on all possible data, even if some is messy, is better for a business than a well selected sample from which we try to prove a hypothesis.  Data is now an operational function.

From your butt's imprint on a car seat (think anti-theft) to the spread of flu based on search terms to serving eviction notices based on the risk of fire, data now serves the role of providing the basis of taking action rather than just recommendations.  

Marketing, like many other functions, has been datified. The path-to-purchase is riddled with opportunities to leverage intent signals from one touch point in the business rules for the next.  We should now be asking questions like: 

• What do consumers do before they do something next?
• What sequence of content consumption relates to making a decision?
• Where and when is the best place to facilitate choice?

These are the kinds of business requirements that marketers should be stating.  As marketing technologists, it is our job to architect a solution that provides the means to find and implement the answers.

If we're thinking about a report or a 3" research binder as the output from the data team, then we're thinking old school. 

Top 10 Digital Thoughts from LiveRamp

What were the CEO's talking about last week?

A conference in San Francisco sponsored by LiveRamp on digital marketing focused a lot on the display advertising ecosystem.   Since this was held at the Computer History Museum it was appropriate to see a lot of panels staffed with the CEO's and senior executives of technology and data firms. 

Here are the top 10 things I took away.

1. The industry is not as advanced or as rock solid as the sales rhetoric suggests.   Expectations (and investment) suggest that several of the challenges will be addressed.
2. The worlds of "brand advertising" and "direct response" are merging as longer term goals align with short term tactics.
3. Buying audiences (and creating them) has replaced buying sites where audiences may congregate.
4. Programmatic and RTB (real time bidding) is moving up the inventory ladder from secondary, remnant levels to premium as publishers get comfortable and see appropriate CPMs.   The display media buying process is filled with inefficiencies and the technology platforms are eying that world.
5. Leveraging cross-channel data in real time is not yet a reality, but a lot of attention is being spent their in order to improve the consumer experience.   This puts the 'omni-channel data warehouse in cold storage' because it doesn't fit the consumer model of respond now, not next week.
6. Attribution is 'directionally correct at an aggregated level' - and this from a guy who should know: the VP, Display at Google.    There is no certainty in any of this data – at best we can improve our confidence and that sounds like a service offering from the marketing service providers.
7. Marketing would pay for a single anonymous identifier that deals with device, browser, OS, carrier, DSP/SSP, email and offline.   Device ID and IP remain suspect, but they're the best we have at the moment.
8. "Intent signals" the art of separating out what's actually important from a business point of view.   The context of location (mobile) is interesting as services with 700 million devices come on stream.
9. The money (VC and and advisers) sees opportunities and carnage on the horizon.  Platform proliferation and channel fragmentation is creating unsustainable market for this MANY companies.  
10. The idea of a 'media plan for one' that is then rolled up into a buy was repeated a couple of times; SVP at dunnhumby made the best case for this.

As is often the case, the sidebar conversations and chance meetings were the most interesting aspect of a conference. 

Get Closer to Customers

What is being taught on the conference trail?

I heard from the attendees at the recent Optimization Summit a couple of interesting points; one very tactical and one strategic.

• "Get" may be the most powerful verb in the English language, at least from a conversion point of view.  Seems that the term entices consumers to act a whole lot better than the IT-centric "Submit"
• "Data-driven" is the wrong message.  We should be consumer focused, and it is data that gets us much closer to an understanding.

 Good points to remember....

Off to RampUp and the integration of on and offline data this week.

Reports Don't Produce Opportunities

How does 'data-driven' actually work?

In a post on All Things Digital, Ben Elowitz of Wetpaint makes the case that it is the phobia of being shown up by the data that contributes to the slow adoption of data-driven {marketing.}   In fact, he quoted a friend's dirty little secret. 

"Nobody wants to use the data."   He goes on to argue that collecting data is the safe AND easy part.

Re-imagining the world is the prescription for the problem, not justifying a set of HiPPO decisions.   To that end Ben offers five great questions to consider:

1. What does my audience love?
2. How do they want it?
3. How can I best relate to them?
4. What secret signals is my audience sending?
5. Where is my sweet spot? 

So, how does a marketing organization get to this point?  Tools and humans - split 10% and 90%.  Too often we hear of technology as the saving grace to the problem; it isn''t.  Creating new solutions that satisfy consumer needs is a high-risk business (just look at the failure rate of new products).  Data and technology are enablers that allow exploration to happen.  In both the NetFlix and Target examples, there was a general direction stated by business executives:  "Should we buy the rights to the British series House of Cards?" and "can we identify pregnant women?"  These questions provided the compass by which people found the answers - 'yes' in both cases.

Note that none of these questions have to do with metrics, conversions or optimization; nor are they the kind that can be programmed into a report or dashboard.  They are the best kind of question:  ambiguous and in need of human thought.    This is where opportunities are found.