Methods in Customer Lifetime Value

What is the future value of each customer?

Sales and Marketing needs to make choices about where to invest: whom to encourage or discourage, whom to offer premium support to, and a myriad of other decisions.   The scarcity of resources or simple misalignment means that not every customer is necessarily a good one.   So, how do we define good?

In this context value is defined in terms of profitability – gross, net, or contribution – and not sales.  Total revenue masks too many costs to be useful in investment decisions.  The objective here is to explore ways that create differentiation or spread between customers so that different strategies and tactics can be implemented.

A customer’s lifetime value (CLV) is often used as the basis of discriminating between customers.    The premise is simple: profitable customers yield a profitable business.  While we may know what a customer has done for us lately, the important question is:  what will she do going forward?

Rather than dusting off the crystal ball, this series looks at three different categories for estimating the future value of a customer and describes a total of nine methods using customer behavior and interactions.  Each of the following three areas are described in separate posts.

1. Time Series – the simplest, and often overlooked, type of forecasting
2. Segment Migration – the idea that some people migrate from one band of profitability to another over time
3. Predictive Models – the use of explanatory variables to identify the specific levers that drive value

I'll save the idea of predicting the Stanley Cup and NBA finals for the Ouija Board....

CLV Episode 1: Time Series

What can the passage of time tell us?

Working under the dual assumptions of “history repeats itself” and “we are creature of habits” time series approaches leverage the pattern of historic sales or transaction data to estimate the future.

With these methods you are driving down the road looking out the rear view mirror.  They are based on simple assumptions and make no attempt to explain why anyone is or is not profitable.  That said these techniques might out perform more sophisticated models. 

Let’s imagine a scenario with sufficient historic information to estimate the profit of each customer for each month in the past two years.  That is we know what he bought and what our cost to market, sell and service has been.

NPV
Given either average or most recent profitability we can create the present value of that stream of cash over the foreseeable future.     It looks like this:

where

• N is the number of time periods
• R is the profit at time t
• i is the discount rate that addresses uncertainty about the future.

An interesting twist is to vary both time and risk on a per customer basis.

Since customer lifetime value is predicated on transacting in the future we can adjust the current value accordingly.  In certain categories a person’s age is a surrogate answer to the question:  How long will she be a customer?   And if there are insights in terms of attrition, then the risk factor can be raised or lowered accordingly.

This approach works well when average or current profit itself isn't expected to change that much.  It also fits expected ideas about customers:  Young, active customers will be more profitable over the long run. 

Trend
Given sufficient history of profit, one could apply time series techniques to predict the trend going forward.  There are two approaches possible here using regression:

• The simplest way would be to estimate profit as simply the passage of time in order to isolate the trend component.  

Yt = a + bt + et

where the value at time t is some multiple of the monthly change in a customer's profit (b) plus a baseline figure (a).    Future profit is simply the number of months since the start of the measurement window times the rate of change.

• A slight extension would be to estimate profit based on the previous profit levels themselves.  It looks like:

where the value at time t is a weighted value of previous profits.   Future profit is based on estimating t+1, t+2, etc. up to the expected end of the relationship.   

As with NPV, both methods can create the desired spread by varying the length of time for each customer.

This approach fits well over the short-to-medium term when customer behavior, and thus profit, trends one-way or another over time.  However, at some point there is a bend in the road that we won’t see and this type of trending can produce surprises if used too far out into the future.

Depth of Repeat
“Once a customer, always a customer” is a big fat myth.   Consumers defect and do so at a surprisingly predictable rate.   We can take advantage of this fact and develop a depth of repeat model that estimates the number of purchases over a given time span.

The date requirements are slightly different than above and focus on using recency and frequency type metrics as the first.   For each customer we need to answer the following questions:

• How many repeat transactions did they make?
• How long did they have to make those transactions?
• How long since their last transaction?

Rather than the regression or econometric approach typical used by marketing analysts, this method uses the distribution of past behavior to estimate the probability of a customer doing something.   To provide a basis of discussion, we’ll start with a picture of the results first.   The following comes from a published analysis of repeat purchases for CDNow and it shows good fit with the number of future transactions. 

And since this is a more technical piece, here’s how it looks to the analyst (and yes it can be implemented in Excel).

The above simply turns things marketers believe into useful planning information.  The probability that a customer will make a certain number of transactions given her history is based on two key assumptions.

• The number of transactions a customer makes varies around their historic average.
• Customers may stop being customers after a transaction and thus drop out never to be seen again.

The result is a series of four parameters that generate the best fitting depth-of-repeat curve.   Thus we can now estimate the number of transactions each customer will have in the future.  And given the profitability of a transaction, we have future lifetime value.

This approach (BG/NBD) and its variants have been used for years in estimating the success or failure of new products from survey data so has a lot of validity.  But as with other time series approaches it is a bit of a black box from a tactical point of view.

Next up:  Segment Migration
Then: Predictive Models

CLV Episode 2: Segment Migration

What bucket will you be in?

The Time Series methods discussed previously compute lifetime value based on individual performance and behavior without regard to either similarity between customers or specific drivers that impact profitability.  Sometimes we can gain more insight by pooling customer behavior and segmenting customers into cohesive groups.  This allows us to either use more information in estimating future profitability or reducing uncertainty by focusing on sameness.   These approaches also begin to help marketers think about what strategic and tactical choices they might want to make from an investment point of view.

Segmentation offers two specific advantages for estimating the future:  stability and migration.

• Stability comes from the safety of understanding how a group behaves rather than a specific individual.  And while ‘average’ is the most dangerous word in marketing; segments provide multiple averages to use.
• Migration reflects that people can change over time and move from one segment to another.  Life-changing events are clear boundaries between segments. 

A market segment is a group of consumers with a common need and can be reached effectively using a common set of tactics.   Similarly, a profit segment is a group of consumers whose profitability, behavior or characteristics are likewise common to one another.  Since the time series methods can be applied to segments as well as to individuals, we’ll focus on potential ways to segment and then think about migration.

Profit Only Quantiles
The concept of dividing customers into equal sized groups based on some metric is common in response analysis.   We often set campaign selection rules that ‘target the top two deciles’ – or that 20% of our database most likely to respond.

In the CLV scenario, we can create segments based simply on the distribution of individual profitability across customers.   Two commonly used methods would be deciles and quartiles. 

• Deciles: create 10 equal sized segments of customers, each with 10% of the customers
• Quartiles: create four equal sized segments, each with 25% of the customers.   Sometimes this will be simplified into three segments by combining the middle two.

The boundaries of quantiles are driven purely by first ranking the customers and then dividing them into equally sized buckets.   This is where the Pareto rule drives from:  20% of my customers account for 80% of my profits.  This approach works when profit is a direct result of behavior, product margin is fixed and customer characteristics don’t discriminate between profit levels.

Descriptive Nodes
Rather than using profit as the sole segmentation measure, we can use all available behavioral and descriptive data to differentiate between customer profitability.   Since we’re interested in assigning people to segments, as opposed to predicting a specific outcome, this approach leverages classification and regression trees.

The following example looks at survival rate of Titanic passengers according to demographic characteristics.  If you were female or a male in a large familial group (siblings or spouses) the odds of survival were in excess of 70%.

Descriptive nodes have a couple of distinct advantages:

• We begin to understand what characteristics and events relate to profitability because the approach splits customers according to the attribute that does the best job of discriminating between high and low.   
• All customers are assigned to one of a set of end segments that can be reduced to simple business "if then" rules.   While often mapped to deciles, there is no reason that the business has to fit the data to the 10% rule. 

This approach works when customer attributes or life-stage relate to behavior or when contribution margin varies greatly across the product line or channel.  

Clusters
Rather than splitting on levels of a specific attribute like profit as Descriptive Nodes does, clustering is a machine-learning approach that groups consumers based on their similarity to or distance from one another.   Those who exhibit similar patterns across a host of variables are grouped together while other consumers are placed into different groups with their neighbors.  

In clustering we’re trying to find homogeneous groups and then see how they differ in terms of profitability.  There are numerous techniques to choose from, but they have all the same objective: reduce customers to groups.  This approach does carry the  risk that while distinct segments may emerge they may not have different profitability profiles.

Clustering is appropriate if unusual or complex ways of identifying 'sameness' are required.  And it is particularly useful when one should segment based on implicit or explicit consumer needs.  Because needs come and go or shift in importance, the size of the segment is critical to understand over time. There may be very profitable customers in a segment but if the underlying need evaporates, they will go the way of the carrier pigeon.

Migration
Since we are interested in the future we need to come up with ways to understand where customers will end up.  All of the segmentation approaches allow us to not only to define to which segment a customer belongs, but also the likelihood of customers migrating from one segment to another over time.

If we look at historic data we can apply a segmentation scheme at the beginning and then again at every subsequent planning period.  This allows us to understand how consumers change over time and it is this information we’re most interested in.    The following is a simple view of what we’re looking for using quantiles. 

The columns represent where someone is today and the rows are the probability she will be in a particular segment next period.  This includes the idea that someone might no longer be a customer, thereby accounting for attrition.  For example, a "Bottom Third" customer is likely to still be in the "Bottom Third" but there is a 10% chance they'll improve and a 20% chance of no longer being a customer.

Because we’re dealing with uncertainty, migration should be simulated a large number of times in order to arrive at a stable estimate for each customer.  If we didn’t simulate the answer would simply be the segment with the largest probability – which in this example is the status quo.   And if we're not sure how stable the migration rates are, we can simulate them as well.

The key is to focus on the odds of changing segments over time and use that to guide the estimate of profitability.


Previous episode: Time Series
Next episode: Predictive Models

CLV Episode 3: Predictive Models

What drives profit?

The previous two sections, Time Series and Segment Migration, address the problem of future profitability through trends and migration.   Predictive Modeling approaches tackle the problem head on.  That is, they use all possible information to predict the profitability figure itself.  This class of techniques requires the most data and has the greatest risk/benefit profile.   It can be very good or be very wrong.   Predictive modeling is excellent for interpolation within the boundaries of a problem, but it can be suspect when trying to extrapolate to areas – like the future – where it hasn’t seen any data. 

All the ideas presented here look like the following:

where the x's are the different variables used and 'B' is their respective contribution to profit 'y'.

These approaches can be applied to individual customers, segments or across the enter customer database depending on objectives and business needs.

Direct Estimation
The easiest method to imagine is to predict profitability directly from a combination of factors – customer characteristics, behavior, market conditions, etc.   This is an extension of the Time Series model shown in the first chapter.

Since we’re dealing with time series, a subset of variables should be the consumer’s previous behavior and profitability.  It is common to see ‘lagged profit’, ‘lagged growth in profit’ or ‘lagged transactions’ in such a model.    Future profit is simply a recursive set of estimates each one period further in the future using the last period’s estimates.   

This approach works given sufficient history of profitability data.  How far in the future one can predict is often a function of how far back one can go to create the baseline trend.
 
Drivers First
Profit is often an outcome of a distinct set of behaviors and characteristics.  Thus, it may make more sense to estimate those factors first and then derive profitability from them.   This is particularly true if product margins and the cost to serve vary substantially over customer segments and product lines.

Profit only exists if a consumer is still a customer.  So, if retention or churn is an issue then focusing on the probability of future purchasing makes sense as a first step.   In this scenario, factors like cross-buying might help since it is often assumed that the more products a customer buys, the more loyal they are.  

An issue with driver estimation is that one model begets another.   For instance, to use cross-buying we need to estimate not only the likelihood of doing so but its relationship to overall profitability.  For instance, it is known that consumers with high customer service costs; heavy promotion usage and a history of revenue reversal actually are more unprofitable the more they buy.

This approach works when profitability varies greatly by things the business can influence.  The challenge then becomes one of estimating all the inputs in the future and keeping those relationships straight.

Marketing Influence
A special extension of the driver estimation procedure is to explicitly model the impact of marketing on the propensity to purchase in the first place.  Whether this is a marketing mix model approach or an assessment of direct contact is a function of the kind of business involved.

This approach fits scenarios with the following logic:

•    Marketing is allocated to people deemed worthy of investment; i.e. it isn’t random.
•    Marketing directly impacts the probability of a purchase
•    Profitability is a function of purchasing and level of marketing conducted

These statements imply that three separate models are required using transaction and marketing contact information.  Each builds on the previous one by using one set of predictions as inputs to the next model.  

This approach is the most theoretically sound and directly accounts for marketing activity. On the flip side it has substantial data and analytic requirements to implement. It has been shown to do a better job than some of the time series suggestions. 

Dealing with Uncertainty
Usually in predictive modeling we’re interested in interpretation rather than estimation.  As a result we tend to ignore the little ‘e’ at the end of the equation listed above.   The error term represents variance we can’t explain and is assumed to average out to be zero and not related to anything else in the model.  Thus, it has little value to the business in terms of explaining what’s going on.  

However, in forecasting the future that little e implies that there isn’t one predicted profitability but rather a whole bunch of them that vary in amount depending on the size of the error term.  Plus, if we do use some nested form of models where one set of outputs is used later as inputs it is quite possible that the individual error terms play off one another creating huge swings in profitability. 

So, like the migration suggestions above it is recommended that any predictive model that estimates the future be run thru a number of scenarios that change the error term.  Here’s an example of the distribution of Customer Lifetime Value from a research paper entitled “Will the Frog change into a Prince."

While the most likely lifetime value for this individual customer is around $200,000 the long tail of high values just might influence the marketing investment or customer support decisions.

Summary and Recommendations
Given scarce marketing resources, it makes sense to focus them on the most important or valuable customers.   Targeting customers who will be profitable is probably the most efficient way to spend marketing dollars.   To do so requires the creation of Customer Lifetime Value based on future behavior and potentially marketing investments.

This series covered a wide range of methods from the simple, naïve approach of ‘most recent’ to contemporary advances from academics.   Since the future is unknowable sometimes the simplest models work best.  There is no single, best approach – it all depends.

1.    Define the objective, appetite for risk and uses of Customer Lifetime Value

• Time Series: stable, abrupt changes not expected, direct marketing not in play
• Segment Migration: different groups exist that behave very differently, marketing influences medium to long term, market structure may change
• Predictive Modeling: large number of known drivers, one of which is direct marketing contact

2.    Define multiple techniques for estimating and simulating the future and ‘average the results’

And remember: Forecasts are always wrong, what is interesting is why.

Expressive, Editorial and Intentional Content

What kinds of ads work where?

Julie Shoenfeld of Perfect Market updated her summary of the LUMAscape for display and discusses the directions display advertising will likely take as the industry shakes out.   One key point at the end is that display advertising may be served with emotion (Facebook) or with intent (Google).

Thinking about that point from a content perspective suggests there are three environments where ads appear.

• Expressive: the communal world of sharing and social networks
• Editorial: the curated world of brands and publishers
• Intentional: the direct act of searching for something

Given those different contexts one can begin to imagine different creative briefs, media plans and even measures of success.  The discussion about GM pulling $10m from advertising Facebook revolves around understanding expectations.  And their pull-out from the Super Bowl clarifies how they view things a bit.

If we think ads served in the world of self-expression surrounded by the banal and the inspiring are going to work the same way as those served in a shopping scenario then we're not digging deep enough into how content works and people decide.  

The proof is in the click-thru rate of all ad types:
Facebook: 0.051%
Google: 0.4%
Average: 0.1%

As the worlds largest broker of destinations, Google's number should be the highest.   And the drop down to Facebook reinforces the notion, as does Internet minutes, that people aren't willing to go anywhere else.   In the middle is the traditional, editorial context where brands have a broader range of options to work with.

So, given those performance figures how do we use these media properties and what is the best use for them in our plans?    Some ideas on what attributes of might work in each context.

• Expressive: interesting, entertaining, humorous, startling
• Editorial: affinity, knowledge, reputation, positioning, differentiation
• Intentional: features, advantages, benefits, pricing, promotions

Not sure what attributes to use yet for the home page on Kindle where Amazon will now sell ads.

Foresights vs. Insights

What's the difference?

Insights are typically focused on the here and now looking for opportunities to take advantage of or activities where course corrections are needed that impact the current planning cycle.   

In the case of foresight we try to imagine a new future that might affect long term planning.  Since there are things we simply don’t know the thinking and analysis style is quite different.  Two questions that help frame the future:

What macro events or trends create step changes in consumer needs that we’re attempting to satisfy? 

For instance the Great Recession fundamentally shook our confidence, which in turn reversed the trends in savings rate and disposable income.  This translated into a need to be frugal which in turn led to ‘deal chic’ where promotions are a badge of honor.  
Who would have thought that sub-prime mortgages would give rise to Groupon?

What needs and solutions could emerge that alter the fundamental positioning of our brand in consumers’ minds?  



For instance, disruptive innovation or blue oceans completely clearly reset the landscape, resulting in new winners and losers.  Given that the mind is the most difficult thing to change there is a tendency to be blindsided.  The idea of taking an unimportant feature and making a market out of that is the basis of being blindsided.   Who would want a small drive that is slower, holds less, and costs more per unit stored?  The folks that wanted to make computing distributed rather than efficient; they created totally new markets.

Insights: how do we improve the existing marketing plan?
Foresights: what marketing plan should we have?

Note: this idea came from discussion on "Growth and Foresight Analytics" on LinkedIn.

Improvisational Insights

How should marketing play insights?

The purpose of finding 'ah ha' or 'oh no' moments is to change the business; and those kind of insights come from the application of creative and analytic thinking.   I've heard it said that jazz musicians make the best analysts so the idea of improvisation within a structure or framework seems a good way to think about insights.

Consumer self-expression and dynamic connections mean that channels and platforms should be viewed more like suggestions rather than prescriptive things that marketing plays repeatedly over and over.  As a participant we have to adapt our approach in real time by taking cues from others; the result is that no performance is necessarily the same.  The notes in the media plan score become merely a jumping off point.

But to orchestrate something beneficial in a fluid environment still relies on some form or structure.   From a marketing perspective, that focus could come from several directions.

• How should messages be constructed?
• How do we arouse interest?
• How do we build trust?

Within these questions insights provide the ability to use tools and platforms to land back on the important tones of the brand's chord.

Bursting the Bubble of Cross-Buying

Is cross-buying profitable?

We all dream of effect cross-selling opportunities to deepen the relationship with our customers and typically think that because of no acquisition costs and the likelihood of better response, certainly compared to cold prospecting, cross-selling is a profitable endeavor.  And on average, it is.   But as we know, "average is the most dangerous word in marketing."

So are there customers to whom we should NOT be cross selling?

In a recent article in the Journal of Marketing the question was put to the test.  It turns out there are bad customers and they are habitually bad.  In fact, across the five industries reviewed between 1 in 10 and 1 in 3 customers are unprofitable buyers and that they account for a whopping 40-90% of the cumulative total loss from customers.

The research suggests staying away from customers who exhibit one or more of the following bad habits:

1. Don't spend more in total as they buy more categories
2. Rely heavily on customer service
3. Cause revenue reversal (cancellation or product returns)
4. Deal hop and make disproportionate purchases of promotional items

The use of these four metrics led to identifying nearly 90% of the unprofitable customers that should be filtered from any outbound cross-sell program. 

So, we have a new tool in our arsenal.

Marketers as Parents

What do we need to believe in to be successful?

The speed at which information flows and consumers use it and discard it is daunting from a traditional planning cycle.   The typical linear approach of Plan - Execute - Track runs causes us problems - by the time we assess and make recommendations the world has changed and passed us by. 

As a simple example of this speed, consider the half-life of a link to content:  50% of the clicks it will ever receive occur within hours of its posting.   There simply isn't time to go back and adjust; there is only time to move forward.   In fact a recent DMA presentation it was reported that now about half of all marketers track performance during execution as opposed to after the campaign runs.

This suggests that we need to think in terms of Track - Execute - Adapt

And to do that without creating total chaos in the business we need to be sure of a few key things:

1. Be clear about our brand promise as this creates the guiding principle for working in such uncertainty
2. Be clear about the objective we are trying to achieve and what is really important
3. Be patient during the learning cycle and be accepting of false starts and missteps
4. Be trusting of our data and technology teams that they can translate digital footprints into intent

Sounds a bit like being a parent as we let go...

The Curve: When to thank, encourage, and reward

What can we learn from transaction history?

The advantage of working on the 'agency' side is that we can see across several different industries.   When comparing how frequently a customer shops across categories one can get a sense of the range of possibilities.

The following chart plots the 'depth of repeat' curve for several different industries - from fast moving retail to 'once in a blue moon' service organizations.   To make comparative statements, each was normalized to the first visit, i.e. 100%.

There are several points to be gleaned from the chart based on looking at the odds of coming back for one more visit.

• The drop from 1st to 2nd purchase is a steep fall off for all five examples.   This suggests a strong "Thank You" program to help consumers decide to come back.
• There is a range of transactions where consumers need to be "Encouraged" to return.  The group of customers with 2-5 transactions may need to be treated to a different set of offers.
• Loyalty can be operationally defined at the point at which the odds of coming back are greater than some point on the curve, e.g. 50%.  These few, rare individuals need to be "Rewarded"

To a large degree the shape of the curve is driven by purchase cycle.  For instance, the average interval for the retail company is much shorter than the Service and Auto 2 companies.

But can this information be predicted based on knowledge of past behavior?   In a word: "Yes"

Here is the customer profile for one of the auto/service companies; it shows the number of consumers by number of repeat transactions over a two-year period against the estimated count of customers.   The estimated number of repeat transactions are based on a combination of Recency and Frequency type information alone.

While possibly not as valuable as a predictive model due to the lack of explanatory variables; this does provide a benchmark method of defining what the expected number of purchases a person will have.  Against this backdrop we can see if direct-to-consumer programs move the needle.

Organizing Digital Marketing

Since consumers don't care about channels, why should we organize that way?

In an era when everything is interactive and chunks of content are consumed on any of four screens as appropriate; what does a marketing organization need to look like to be successful? A couple of scenarios come to mind:

• Direct-to-consumer: the delivery of messages aimed at facilitating a decision by an individual or a group.   This covers targeted media whether traditional direct marketing, consumer promotions, pure Internet/Mobile plays, or custom content served up based on behavior. 
• Engagement: the provision of applications, leisure content, or information with the desire of establishing rapport by helping or entertaining people.  This covers aspects of marketing where the focus is typically on brand and relationship building, e.g. consumer experience. 
• Insights:  the identification of the "Ah Ha" or "Oh No" moments that come from connecting all the dots that digital self-expression leaves behind. Plus consumer expectations for real time response means we're entering the era of Track>Execute>Adapt where the data must define the plan.  

This structure rises above the challenges that arise with the proliferation of channels, technologies and tools. If fact, tools like social technology can be used for both objectives - influence and support. As an example twitter can deliver offers direct-to-consumer or engage them with links to interesting content.

It also provides a better platform for career growth since it is highly likely that a large number of CMOs in the future will come from this part of the business. 

Notes:
For some comments on digital structure and issues eConsultancy's produced a video summary of their conference on the topic. 
For a look at how the total communication spend is broken out, Veronis, Stuhler, Stevenson (VSS) released their updated spending forecast.

Digital Behavior vs. Digital Identity

What do we see when we only have bits?

Due to the nature of technology, identifying actual people based on their digital footprints is extremely difficult.  In fact it often presents a fleeting, alternative reality. 

Cookies, IP addresses and other technical devices are at best a proxy for a real person; at worst they are an inaccurate puzzle with missing pieces.   To make matters worse these methods are limited in terms of both time and device.   Just try to identify me across four screens and six months.   Add to the mix the response time that digital marketing allows and we must focus on the here and now.

The result is that we have to balance our focus of what a person does over time with what is the behavior being exhibited on a device right now.   The rise of retargeting technology seems to be a good example of this new duality.

AhHa or UhOh Moments of Analysis

Just what is an insight?

On LinkedIn there is a long running discussion on defining 'insight' in one word.   My favorite pair is 'AhHa!' and 'UhOh'.    These two words sum up the essence of insights - tell me something I didn't know I need to know.  

Both reactions suggest that a flash connection happened that allows the analysis to turn into action.  In one case it is opportunistic; in the other corrective.  

• Opportunistic: finding new markets, product uses or ways to connect with consumers
• Corrective: adjusting plans to better reflect reality 

In both cases analysts have done their job - help marketing change the future.

Estimating Customer Profitability

How can one predict the future?

Given that customers are fickle and firms are inconsistent the idea of predicting future behavior, much less profitability, is fraught with perils.   In fact, it is often the answer to 'what have you done recently?' that is the best predictor of future profits.   While current or average profit is a naive assumption it often out performs much more complex approaches.   Recent research took a stab at addressing the question again focusing on a series of questions.

1. What marketing activity is the customer likely to encounter in the future?
2. What is the likelihood of purchasing given the contact plan?
3. What is the probable profit given that a future purchase is likely?

Each question results in a series of estimates for every customer across the planning horizon of 12 quarters.  To assess whether it works, the summary looks at customers at two different points in time and plots the migration of people across profitability segments. 

Since the answers are all estimates and subject to error the authors added a nice twist; they simulated a 1,000 different futures for each customer based on the uncertainty in answering the above questions and then pooled the results.   This approach of creating a myriad of potential outcomes worked better than the benchmark models.

Like so many marketing problems, the future is all about best guesses based on alternative reality rather than the certainty of a point estimate.   

Content is Not A Device nor a Channel

What contributes to conversion?

The other day there was a good post on multichannel attribution on Occam's Razor that covered three different ways to think about the problem: Stimulus to Store, Screen Experience, and Channel Usage.  Each approach answers different questions and throws up unique sets of challenges.  Given that attribution analysis often results in the allocation of marketing spend to silos we continue to reinforce the bad habit of thinking in ways that consumers don't.  The examples in the post of changing channels and switching devices along the path to purchase reinforce this notion.  It also makes the case that the problem may actually be intractable in the long run because we simply don't know who the person is.

So, if that line of thinking is a potential red herring is there another question to consider?  Maybe we should think more about what consumers are consuming rather than how they are doing it since it is a given we should be creating content for multiple channels and devices. Shouldn't we be more interested in the fact that a person might be looking for aspiration, deals, feeds & speeds, or validation than what version of an operating system he uses? 

• Emotional content sparks interest and satisfies need and can be in the form of TV, print or branded content reached thru organic search.
• Promotional content satisfies the need to 'get a deal' and reduce the risk of trial and can be a direct response email, coupon code aggregator, or store flyer.
• Informational content provides the rational basis for defending a decision be it a feature-laden fact sheet or 3rd party comparison report.
• Communal content provides the validation of other people be it reviews on an e-commerce site or a quick response to a Tweet. 

A segmentation strategy based on content consumption habits is likely to be more useful to a marketer than delivery preferences. To be sure, this is not how most attribution discussions go but in reality both channels and devices are simply delivery vehicles for content. 

Analtyics Were at the Beginning of Cinema

Was it a creative idea or data that gave us motion pictures?

While associated with the creation of motion pictures, Eadweard Muybridge basically used data to settle a bet.  The question as to whether a horse's feet were all off the ground at one time while running was apparently the source of much discussion (and likely a wager or two).

To settle the bet for his patron Muybridge invented a system to trigger a series of pictures as a horse ran by - he also later created the zoopraxiscope to display images of what was to become stop-action or slow-motion photography to audiences.

Today Google honored him with a doodle....

Marketers Should Embrace Hadoop

Why is technology named for a stuffed elephant important?

As marketers we've been conditioned to do the following:

• To ask only really important questions because the half life of the others is much shorter than the time it takes to get answers.
• To view most data as having little value because we can't adequately see or filter it into anything useful.
• To accept answers to questions that technology can support based on yesterday's needs rather than what the business requires today.

We got into this situation for two very good (at the time) reasons.   First, the cost of data storage and processing was such that we had to be judicious about our requests - constantly making trade-offs between known costs and uncertain benefits.  Second, the need to report results upwards ended up with a belief that there is only one true number.   The result was a lot of time and effort spent on collecting and cleaning only the best data, putting it in a data warehouse, and judiciously handing out the keys to the kingdom.

An unintended consequence was that we tended to focus on structured data - things that could be represented in rows and columns like customer transactions - rather than the amorphous data of comments and images whose value was unclear.  Because the marginal cost of implementing changes to data and reports was high we were left with a situation where the addition of any new data source or question would be delayed by the process of getting funding. 

Well, that approach won't work anymore. 

Digital interaction spins off data in real time that we have to leverage in real time if we are to respond to consumer interests and actions.  Businesses need to consider that...

• Self-expression leaves a myriad of opinions and objects to be shared
• The infinite paths-to-purchase leave breadcrumbs at each and every step
• Human interaction influences opinion and choice more than the paid placement of messages

And a world in motion is very different than a static one in terms data volume, variety, and velocity.  This is the V3 world of Big Data where everything is simply more. 

In the past we could plan, execute and track; today we have to track, execute and plan.

It would probably be insane to try to implement a consolidated system that captured and stored all of those events with a customer loyalty program. So, if we can't build an uber data warehouse, what are we to do?  Well the folks at Google and Yahoo! solved that problem by distributing the question rather than centralizing the data in order to improve the indexing of web sites for search.  And out of that comes Hadoop a framework for data-intensive applications that leverages distributed processing designed to handle the type of data generated above. 

Consider the simple idea of presenting a customer with the next best product while she browses your web site. This requires he tight integration of real time events (what are you looking at), historic transaction data (what have you bought) and predictive analytics (what should we recommend).

Hadoop is not a replacement for existing infrastructure but represents a way to think about three key marketing needs.

1. How do we transform digital self expression in order to leverage it with historic transactions?
2. How do we do scale real time, event based analysis and deliver them where the consumer is right now? 
3. How can we overlay content consumption habits with our traditional segmentation schemes to deliver a better experience?  

The new schwag for both your CMO and CIO....

(source)

Building a Degree from the Data

What should we study?

It is not unusual for people to change colleges or even start-stop-start their education.   In certain segments it is common for applicants to come to the table with a substantial number of credits.   So, an interesting analytic problem would be to take those credits and come up with two to three alternative courses of study.  Maybe one that is of expressed interest and the others are simply the shortest distance between two points - now and the degree.   So, what would be required:

1. Courses taken: what credits have been earned to date.
2. Course map: articulation agreements align courses from one institution to another
3. Degree requirements: the core and elective courses for potential degrees

It becomes a database alignment and probability exercise to create several alternatives given a set of inputs.  To add some spice to the solution a simple way of flagging interests or using collaborative filtering or recommendations to make the solution more social.

With the proliferation of degree options it is no longer a simple choice of A vs. B.  It is more like A with some C and maybe a D thrown in because I don't like B.  And all these options actually make it hard for people to choose one university over another; a simple tool that provided some alternatives might just help conversion and win the race to rapport. 

Digital vs. Analog Marketing

What changed when we went digital?

There have been at least two fundamental shifts in how we approach marketing in the digital era.

First, we've gone from addressable to guessable.   The concept of a persistent identifier that we could hang our programs on has evaporated in the digital age.   Gone is the physical address and phone number that formed the foundation of our knowing a consumer .  Now IP addresses and cookies form the 'bedrock' of digital knowledge where we often have to do a lot of inferring about an individual.  This means we need to think in terms of probabilities rather than absolutes.   It also may mean we should focus on the behavior itself and worry less about who is doing something. 

Second, we've gone from our media calendar to consumers' internal clocks.  The concept of a prescribed sequence of campaign steps within a defined calendar no longer applies when everything is interactive.  The notion of batches or drops needs to be augmented with support for pull and accessible content.  This means we need to think in terms of continuous exposure.  The implication of this is that we need to abandon the silos of channel and think about surrounding the consumer with interesting content to help her choose.

Marketing and the Digital Self - Part 2

How does marketing need to change?

So, let’s assume for the moment it is ok to use the digital self for marketing purposes. What’s different?

While technology clearly leaves a footprint that can be leveraged in the marketing process an immediate implication is the creation of infinite paths to purchase. Because every surface or touch-point can act as a jumping off point the traditional view of what tools are appropriate for awareness – consideration – purchase needs to be completely rethought.

For years marketing wanted to deliver the right message to the right person at the right time and we’ve come close.  However, the adoption of social technology has meant more likely that some message is delivered to someone at some time.  And even if we succeed in delivering our message you can guarantee it will be vetted.  The questions that really need answers – is this a good price, what do you think, will it look good on me? – are a simple click, tap or swipe away. 

The combination of mercurial paths and hazy messaging results in the abdication of control and a significant change in influence wielded.   As a result, marketing has to deal with a fluid fog because the brutal truth is we can’t know exactly what is going on at any moment in time.

So, what should we be doing?  There are new areas for marketers to consider as they create their plans. 

1. Understand how choices are made. What are the sources of influence?  Given the immediate access to numerous points of view those companies that think about improving the probability of choice will win out.
2. Understand how content works.  What types of content affect choice? What should we be producing? Creating a blend of emotional, promotional, informational and communal content will be required to satisfy the left and right brain aspects of deciding.
3. Rethink the use of 'brand advertising' and 'direct-to-consumer' in the marketing mix.  A strong brand acts as an emotional short cut to a decision.   And in the recommendation economy this is consumer-to-consumer rather than air cover.  The focus should be on surrounding the consumer with interesting content to help her choose.  

These steps all rely on understanding the role of a brand in the era of digital self-expression. 

Marketing and the Digital Self - Part 1

How do digital bread crumbs relate to marketing?

The benefit of a digital world where nearly everything we see or touch is interactive is that we can actuate our human desire to share.  The rise of social technology reinforces the notion that “self-expression is the new entertainment”  (Arianna Huffington).  The unintended consequence of all of this for media is disruption as both new players and new channels of distribution emerge.  And with all things digital, as we share content we’re creating a trail of breadcrumbs, or a ‘digital self.’

And, this leaves us with a fresh question:

Do the signals left by digital self-expression represent a foundation for selling stuff?

On the one hand, this is viewed as a dream come true: We know what a person does, what her interests are, and with whom she communicates.   One can’t imagine a better way to create a target rich environment for personalized communication and or laser-focused advertising.  

However, identifying individuals based on their behavior raises a thorny question:  Do we need permission to use the digital self as a platform for targeting?   Many of the industry approaches to this dilemma take a loosely aggregated or anonymous approach to avoid identification of specific individuals through the creation of segments, predictive models or even an IP Zone.  While this is often framed as a question of privacy, there are some basic marketing questions as well. 

1. Are the crumbs we see reflective of the need we satisfy?
2. When does self-expression relate to a commercial interest?
3. How do we distinguish transient self-expression from core intent?

An approach would be to try to glean insights from the combination of purchase history and the digital self.   Since these represent two very different personae this route might help reduce the amount of inference, and frankly guesswork, required to understand a consumer’s intentions.   Yet to do this requires some serious integration work starting with the acceptance that we live in a brave new world.    

A View from the Summit

What does marketing look like from the top?

Given Adobe's client base, there is no doubt what happens at the Adobe Digital Marketing Summit doesn't stay here is Salt Lake City.  This year's message about the Digital Self is likely to be repeated in the conference rooms of the biggest companies with the question: What do we do now?

But this is a tale of two conferences: It was a marketing conference; it was a technology conference.  Those who had been to previous Omniture Summits missed the deep diving 'how do I' sessions.  The new marketers in attendance wanted less code or product marketing and more solutions and proof.   Reflecting this duality many a company sent both their geeks and their business minds as hallway conversations involved small groups. 

This being Adobe's Summit, their site-centric view of the world yielded a common refrain: convert traffic into sales.  Even the session on media monetization focused on segmenting site content based on traffic patterns.  Their position on driving traffic still remains one for their partners as evidenced by the sponsors - multichannel campaign management and search companies dominated the exhibit hall.  So, it falls short of being a full-blown digital marketing conference.

They continue to extend the Genesis idea of data integration in a couple of new ways - separate environments for handling of multi-channel and real time data (e.g. Insights), integration of analytics into reporting (Navigator - from the lab), and numerous social plays - usually from an ad or campaign point of view.   The idea of personal, direct-to-consumer communication remains elusive. 

Personal highlights:

• "Self-expression is the new entertainment" - Arianna Huffington's view of the time spent on line.  Whether permission is granted to use that to sell stuff remains to be seen.
• A chance encounter to have lunch with Adobe founder John Warnock where we discussed 'where was this all going?'   His view:  simple, end-to-end understanding of how to market.  The opportunity to have a point of view on what content to create is still open and 'is the right question'.
• "Fail fast to succeed faster" - the conclusion of a session on conversion testing delivered by my son.  I'll admit to being proud and prefer this to Biz Stone's recommendation to hire people who have failed in the past.

Implications:

1. At some point the adjective 'digital' will be dropped and this will be just the Adobe Marketing conference as those in attendance assume leadership positions.
2. Off-site strategy remains a big opportunity while on-site optimization is becoming a professional service. 
3. The blending of CMO and CIO will continue as both disciplines will be required going forward. Saying you don't understand technology or marketing will be career-limiting.
4. Consumer identification remains the holy grail.   All working sessions talked about the year-long integration challenges to get to a place where the vision could be realized. 
5. Data and analytics will be the fabric of marketing planning; they will have a seat at the table from the beginning.

Measuring Human Values

What fuels growth?

Earlier I had suggested that marketing should own a human activity like search, shop or view.  Last night Jim Stengel, former CMO of P&G, argued that companies should take the high road and focus on a core human value.   His list of five from the book Grow included:

• Eliciting Joy
• Enabling Connection
• Inspiring Exploration
• Evoking Pride
• Impacting Society

Focusing on this level results in some truly remarkable companies - Coca Cola, Starbucks, Discovery, Mercedes-Benz, and IBM - from the list of 50 that had superior performance over the past decade. 

These are clearly ethereal ideals that make measurement just a bit tricky.  Jim relayed a story that focused on measuring engagement with the ideal with just two questions.  Are employees living it? Are customers experiencing it?    If we are making progress along those two dimensions, then financial results follow.  As an example, for Visa the measurement is around both rational and emotional brand attributes:  Trust, Secure, Reliable and Empowerment, Freedom and Control.   All of this to create better living with better money.  

But just what is a human value?  Here is a potential definition.

• Values are beliefs. But they are beliefs tied inextricably to emotion, not objective, cold ideas.
• Values are a motivational construct. They refer to the desirable goals people strive to attain.
• Values transcend specific actions and situations. They are abstract goals.
• Values guide the selection or evaluation of actions, policies, people, and events. That is, values serve as standards or criteria.
• Values are ordered by importance relative to one another. People’s values form an ordered system of value priorities that characterize them as individuals.

These criteria suggest there are several other potential values or ideals that could serve the basis for a business.

As a marketing agency I'm still stuck at the verb level with 'helping people choose' - need to elevate my thinking.

Understanding How Content Works

What should we be asking ourselves?

With the rise of sharing and interactive communications, there is an opportunity for someone to step in and make the claim: "We know how content works."

The questions that need answers include:

• How should the impact of content be assessed?
• Is there a categorization scheme that makes it easier to assess and report on content?
• What content attributes affect choices and decisions?
• How do content types map to engagement and transactions?
• How do people consume content - by device, location and intent?
• What types of content accelerate the path to purchase?
• What are the relevant content consumption segments?

Plus a couple on who is doing it well today:

• Is this a client side activity or vendor/agency/publisher opportunity?
• What firms are actively pursuing the answers?

Take a Normative View

Quick: Is $28m in sales a good thing?

Well, the answer depends on a whole host of other factors the two most important being 1) what did we expect the number to be and 2) how does that stack up against similar plans.   Both expectations suggest that relative or normative assessments are as critical as an absolute figure.

If expectations were for $20m in sales, then the $28m is likely a good thing.
If the marketing plan used to generate the estimate typically generates $40m in sales then there is a potential issue to resolve.

One of the distinct advantages a supplier/vendor can bring to the table is a pan-industry view to help provide context and set expectations for performance.  Once upon a time I worked for BASES and we combined marketing plans and survey results to predict first year sales.  Our clients could have replicated the math but chose not to for two reasons:  they could kill the messenger if the results weren't up to par and they really wanted to know how their ideas stacked up against industry norms.

So, ask your vendor to tell you how your programs, plans, products stack up.  They should be glad to help you understand.

Recognizing Patterns from Nasty Bugs

What can we learn from epidemiology?

Stopping the spread of serious health risks like SARS and H1V1 now relies on analysis and technology as much as it does treatment and containment.  A recent article discusses various ways the health community is banding together to share information in order to create better early warning systems.

While marketing challenges pale in comparison to the arresting the spread of disease, there are a few points we should take away:

1. Integrate the data.  Silos helps no one in the long run.
2. Make it as real time as possible.  Winning the race to rapport requires a sense and respond mentality that doesn't fit the budget cycle.
3. Let the data decide.   Opportunities are often too complex for mere mortals to recognize in time.  

Social media has been suggested as an effective epidemiology tool, now may be the time to reverse the process and adopt those techniques for marketing.  

The Evolution of Content Distribution

How do we get from print to pull?

While marketers should own a verb of human activity, their vendors should own one dealing with either planning or execution.   When it comes to content distribution there are four possibilities that also form a continuum for the industry's transformation.

Print - Publish - Push - Pull

• Print:  the production and distribution of large quantities of content - usually defined by someone else. Companies are rewarded for operational excellence. 
• Publish: adds the creation of the content itself to the production function.  Content producers, custom or mainstream, are rewarded for editorial excellence (and reach).    
• Push: adds targeting to the mix and requires a deep understanding of the recipient's needs, behaviors and preferences.  Because this is typically direct-to-consumer suppliers are rewarded for return on investment.
• Pull: flip the whole process on its head and let consumers find content they need or have in mind. Those who can solve a disaggregated technology problem will be rewarded. 

The concepts of push and pull require new thinking - who consumed what, where and how and then what did they do (buy)?  The keys to look for:

• Trackable - each chunk of content has to be identifiable as a standalone entity.  The unique identifiers used in content or digital management should be viewed from a consumption not just a management perspective.
• Identifiable - consumption has to be specifically defined as to individual, location, and context (possibly inferred from device).
• Linkable - to determine the impact of content, consumption habits must be tied to purchase behavior. 

All this will eventually lead to answering the question: "What content should we produce?"  

And the first marketing services company to answer that question for its clients can make the claim: "We know how content works."

As Dave Mason sang in a World in Changes

World in changes going thru
I've got a lot to learn about you
World in changes going thru
You've got a lot to learn about me too

Two Rights Make a Wrong

Why is average the most dangerous word in marketing?

A recent post by George Michie over on Search Engine Land talked about reasons management might not like the numbers your reporting.   While written about paid search campaigns the lessons are applicable to all marketing endeavors.   Here's just one of the examples from that post:

Since management likes to see just the top-line numbers, this report would be a recipe for disaster - a tongue lashing at best and a pink slip at worst.

So, why did all three key metrics fall?   (And the question as to whether these are the right metrics in the first place is addressed in the original post.)

We mix our tactics for a number of reasons: create reach, ensure coverage, and stimulate different responses to name just a few.   Like all marketing tactics, some things are worth more to us than others in terms of their ability to generate contribution margin.   So, in this example because some keywords are worth more than others the manager made two right decisions between weeks 1 and 2.  First, she identified the allowable marketing cost for two segments of key words:  "Quality" - $500 vs. "Discount" - $150.  Second, she allocated spend according to the performance of each tactic to grow the business profitably.

The results: 75% more high quality visits and 67% fewer discount visits.  A much different story than the averages told above. 

In this case, two right decisions got transformed into one that was perceived as bonehead because averages were reported.  In fact, because of the fundamental change in approach we can't really compare the overall CPL $250 to $304 although I'm not sure I'd try to make that argument directly to my boss.

Notes to self:

1. When changing the mix over time, don't report aggregates only.
2. Don't report diagnostic measures, report what actually matters.
3. When segmenting, there is no overall average.

3 Eras of People and Place

How are the times are a-changing?

There have been three distinct eras of communication.

First, direct-to-consumer communication focused on a physical location since that is all we knew.   Direct mail and phone calls were direct to a specific place in hopes of reaching the intended recipient.   While these methods could be targeted, they lacked deep personalization and really didn't change much over time based on needs.  My house still has the same geo-demographics as it did when I moved in 20 years ago. 

Second, as technology allowed for portability, we sent our communications to a specific person via email and cell phones.   We knew very little if anything about where and how the message was consumed.  This time period was characterized by the 'right message to the right person at the right time' - nothing said about right place or conforming the message to be place relevant.

Now, we are entering a third era where we are beginning to understand person@place.   Looking at the intersection of place and person allows us to be much more context aware and focus on behavioral intent. Some themes...

• Augmented reality overlays information that would be relevant to a person at a given place.  And the rumor of Google's 'heads up display' glasses suggest this area will heat up.
• Ambient social networking identifies those around us that we might want to interact with; it may be the scariest trend of the year.
• SoLoMo and hyperlocal are common topics in search marketing via Foursquare among others.

Is this my new phone?

Analytic and Artisanal Marketing

Where should we focus in order to grow?

In an excerpt from "Grow", former P&G marketing leader Jim Stengel talks about what makes a brand enduring over the long-term. They focus on an 'ideal' or more simply put: what a company does to improve the lives of its customers. 

To create those brands requires an artistic temperament that understands that success is not only measured in operational terms but also in terms of quality of life.  In an interview on co.create he states "Certainly we have to measure the volume and the sales and the margin and cash, etc. Those are table stakes."

Thus, it seems that marketing not only needs to be at least a co-owner of the vision and passion it also needs to report on the organization's progress of achieving it in the minds' of consumers.   

This will definitely a new set of metrics that are not found in transaction data.

Matching Analytic Skills to Marketing Plans

Why do we analyze?

There seem to be several key questions we should be asking and answering.

1. Why are results different than we expected?
2. What do we expect to happen in the short term?
3. Are there opportunities that would put us on a better trajectory?

Each question requires a different mindset as well as skills to answer.  

The first typically focuses on exploration, discovery and detective work since we have an outcome and need to understand why it occurred.   These analysts have a strong desire to solve a riddle posed by deviations from existing marketing plans.

The second usually means estimating the future by understanding how business changes as a result of past activities.  These analysts want to create order out of chaos and link causes to effects to create a better marketing plan next time.

And the third is the most improvisational of all since it requires imagining a new future and then working backwards.  These analysts focus on the art of the possible and conjure up completely new marketing plans.

A good team will have a mix of these skills.