How do we align sales and marketing?
A recent post by Kevin Hillstrom challenged the analytic and vendor community to help figure out the problem of multi-channel sales attribution. When using multiple delivery vehicles and providing multiple shopping outlets the attribution question gets a bit muddy to put it mildly.
Two points:
1. As noted the other day, attribution shouldn't necessarily be used to measure productivity.
2. The next challenge should include other on-line tactics like microsites and affiliate networks.
That said, it's still a worthy cause.
Thursday, November 30, 2006
Tuesday, November 28, 2006
Multichannel Attribution Headaches
Should we always account for all sales?
One thorny problem in multi-channel retail is the attribution of sales to one of several marketing programs. Imagine a retailer which uses infomercials and 800 numbers, an e-commerce page on the corporate site and an integrated online campaign. To keep the example simple, the productivity of each medium is the ratio of revenue to costs. The 800 number and on-line campaign are arguably easier to account for; but the e-commerce site has no good way of tracking back to spending -- just where was the impetus to purchase generated? Ultimately the business question is: Which medium is more efficient or productive in terms of generating sales?
A common approach is to assign the mystery sales to a channel based on some sort of allocation. Examples of the basis for such analysis include a) dollars spent, b) dollars received, or c) an arbitrary/benchmark/standard figure.
If the goal of the exercise is to understand the efficiency of different media, then the allocation should be excluded from the analysis. Any conclusion which includes such allocations is going to be distorted based on the assumptions about how to divide the unaccounted sales.
Two common analytic problems exist: 1) When using a productivity measure that is based on the ratio of revenue to costs then the 'law of large numbers' makes it harder to prove that the 800 pound guerilla is better than the 90-pound weakling because of the big denominator; 2) When allocating from the top down, the problem is one of algebra not of productivity. By definition there is one allocation figure at which the efficiency of the two media is equivalent making it possible to game the analysis and prove a desired conclusion.
In sum, if the question is about productivity and efficiency then avoid the allocation step altogether and keep it apples to apples.
One thorny problem in multi-channel retail is the attribution of sales to one of several marketing programs. Imagine a retailer which uses infomercials and 800 numbers, an e-commerce page on the corporate site and an integrated online campaign. To keep the example simple, the productivity of each medium is the ratio of revenue to costs. The 800 number and on-line campaign are arguably easier to account for; but the e-commerce site has no good way of tracking back to spending -- just where was the impetus to purchase generated? Ultimately the business question is: Which medium is more efficient or productive in terms of generating sales?
A common approach is to assign the mystery sales to a channel based on some sort of allocation. Examples of the basis for such analysis include a) dollars spent, b) dollars received, or c) an arbitrary/benchmark/standard figure.
If the goal of the exercise is to understand the efficiency of different media, then the allocation should be excluded from the analysis. Any conclusion which includes such allocations is going to be distorted based on the assumptions about how to divide the unaccounted sales.
Two common analytic problems exist: 1) When using a productivity measure that is based on the ratio of revenue to costs then the 'law of large numbers' makes it harder to prove that the 800 pound guerilla is better than the 90-pound weakling because of the big denominator; 2) When allocating from the top down, the problem is one of algebra not of productivity. By definition there is one allocation figure at which the efficiency of the two media is equivalent making it possible to game the analysis and prove a desired conclusion.
In sum, if the question is about productivity and efficiency then avoid the allocation step altogether and keep it apples to apples.
Monday, November 27, 2006
Paying for Spam
How much is my information worth?
It appears everything has a price, even unwanted email.
Ken Magill highlights an interesting twist in the spam world. A new company, Boxbe, allows users to set a financial threshold at which mail from unapproved senders will be allowed through a challenge-response gateway.
Might be interesting to see what $0.01, $0.10 and $1.00 allows through.
It appears everything has a price, even unwanted email.
Ken Magill highlights an interesting twist in the spam world. A new company, Boxbe, allows users to set a financial threshold at which mail from unapproved senders will be allowed through a challenge-response gateway.
Might be interesting to see what $0.01, $0.10 and $1.00 allows through.
Wednesday, November 22, 2006
Holiday By the Numbers
How much turkey will we eat?
The census bureau, the know of all things numeric, released its 2006 figures for Thanksgiving.
The highlights:
The census bureau, the know of all things numeric, released its 2006 figures for Thanksgiving.
The highlights:
- 265 million turkeys, almost one per man, woman and child
- 664 million pounds of cranberries
- 1.6 billion pounds of sweet potatoes
Happy eating!
Tuesday, November 21, 2006
Name and Address: Keys to the Kingdom
Should name and address always be combined?
In most situations analysts need the ability to track people over time. Whether it is to assess contact strategy, compute life time value, or simply report customer file growth a consistent means of identifying unique individuals is required.
Often name and address are used in tandem to identify specific individuals, but that is neither a foolproof strategy nor always appropriate. One example: In the cable business the concept of 'homes passed' or those with a piece of coax in the wall is critical because it relates to large capital expenditures. Thus, the network folks usually don't care who lives at a given address but only whether that address is wired. Conversely, marketers are more interested in people because they consume products and services. They particularly want to know when they've moved so they can select the appropriate services.
A common customer table which contains both names and addresses falls somewhat short of satisfying business needs. A better solution rests in separating names and addresses into separate master lists. The intersection of the two, i.e. people who have access, represents the target universe.
In most situations analysts need the ability to track people over time. Whether it is to assess contact strategy, compute life time value, or simply report customer file growth a consistent means of identifying unique individuals is required.
Often name and address are used in tandem to identify specific individuals, but that is neither a foolproof strategy nor always appropriate. One example: In the cable business the concept of 'homes passed' or those with a piece of coax in the wall is critical because it relates to large capital expenditures. Thus, the network folks usually don't care who lives at a given address but only whether that address is wired. Conversely, marketers are more interested in people because they consume products and services. They particularly want to know when they've moved so they can select the appropriate services.
A common customer table which contains both names and addresses falls somewhat short of satisfying business needs. A better solution rests in separating names and addresses into separate master lists. The intersection of the two, i.e. people who have access, represents the target universe.
Friday, November 17, 2006
Reasons Why we Model
What business problems need models?
The use of modeling generally falls into one of three categories.
1. Prediction -- the most common interpretation is about predicting the future, either in aggregate or at the individual customer level. Examples include sales forecasting, acquisition and propensity scores.
2. Classification -- modeling is also useful for categorizing people (or items) into buckets. The goal is to minimize the differences within a given group while maximizing the differences across groups. Good segmentation schemes rely on classification so that messages can be better targeted.
3. Estimation -- plugging holes, e.g. missing data, or calculating a new metric, e.g. Life Time Value, represents the third area of modeling. This differs from prediction in the sense that there is no time element involved.
Prediction and estimation often use a similar set of techniques, just different data and assumptions; classification problems are best solved with a different set. Thus, any good modeler will want to understand the business problem and objective first before reaching into his satchel to pull out a tool.
For another view, see Katie Cole's piece in MultiChannel Merchant.
The use of modeling generally falls into one of three categories.
1. Prediction -- the most common interpretation is about predicting the future, either in aggregate or at the individual customer level. Examples include sales forecasting, acquisition and propensity scores.
2. Classification -- modeling is also useful for categorizing people (or items) into buckets. The goal is to minimize the differences within a given group while maximizing the differences across groups. Good segmentation schemes rely on classification so that messages can be better targeted.
3. Estimation -- plugging holes, e.g. missing data, or calculating a new metric, e.g. Life Time Value, represents the third area of modeling. This differs from prediction in the sense that there is no time element involved.
Prediction and estimation often use a similar set of techniques, just different data and assumptions; classification problems are best solved with a different set. Thus, any good modeler will want to understand the business problem and objective first before reaching into his satchel to pull out a tool.
For another view, see Katie Cole's piece in MultiChannel Merchant.
Value Drives Hygiene
What is the cleanest data in an organization?
Arguably the cleanest customer information in any organization is Accounts Receivable, or who owes us money.
It appears that the willingness to expend effort to clean and maintain data is directly related to the value that information ultimately provides. If there is difficulty in justifying cleaning and managing customer information then the alignment to the metrics and objectives that really matter to an organization isn't quite right. It might be time to go back and revisit the fundamental objectives.
Arguably the cleanest customer information in any organization is Accounts Receivable, or who owes us money.
It appears that the willingness to expend effort to clean and maintain data is directly related to the value that information ultimately provides. If there is difficulty in justifying cleaning and managing customer information then the alignment to the metrics and objectives that really matter to an organization isn't quite right. It might be time to go back and revisit the fundamental objectives.
Thursday, November 16, 2006
Vendor Client Relations
What should we be asking for?
Charles Pearlman gives a good discussion on the difference of what clients want and what vendors want was recently published in DM Review. A key point is that people often ask for things in terms they THINK the other side wants to hear. Clients don't want surveys and models, although that is how vendor pricing is done, they want the recommendations and actions that come of such services.
The same point can be made about internal services as well. Brand managers often need help in figuring out what do to. As recommended in the article, walking a mile in their shoes is good for the soul, or is that sole?
Charles Pearlman gives a good discussion on the difference of what clients want and what vendors want was recently published in DM Review. A key point is that people often ask for things in terms they THINK the other side wants to hear. Clients don't want surveys and models, although that is how vendor pricing is done, they want the recommendations and actions that come of such services.
The same point can be made about internal services as well. Brand managers often need help in figuring out what do to. As recommended in the article, walking a mile in their shoes is good for the soul, or is that sole?
Wednesday, November 15, 2006
Whiz Bang Techniques
What's driving the future of analytics?
In an article on trends in analytics, Rick Watrall of MarketingNPV summarizes the state of the industry:
In an article on trends in analytics, Rick Watrall of MarketingNPV summarizes the state of the industry:
The critical commonality among the trends above is that none of them involve the next "whiz-bang technique" coming down the pike. The future of marketing analytics will be determined more by people-driven processes, technology, and creative applications of unique solutions to individual situations.Of the drivers, I think the people-process is the most important driver. Successfulul adoption of analytics should come from a fundamental desire to understand not from the application of some technique or tool.
Analysis: Throw Away the Math
What does successful analysis look like?
At the turn of the century, as in the last one, economist Alfred Marshall summarized his rules on analysis.
1. Math is nothing but a shorthand language
2. Translate the equations back into words or theory
3. Illustrate with examples or stories from real life that are important.
4. Burn the math.
5. If no words or stories emerge, then burn the theory – don’t hold on to the numbers.
So when preparing reports and particularly presentations pepper them with real lifanecdoteses and stories and leave the equations and formulae in the trash bin.
Note: The story was related in Brue's "The Evolution of Economic Thought, 5th ed.," pg. 294.
At the turn of the century, as in the last one, economist Alfred Marshall summarized his rules on analysis.
1. Math is nothing but a shorthand language
2. Translate the equations back into words or theory
3. Illustrate with examples or stories from real life that are important.
4. Burn the math.
5. If no words or stories emerge, then burn the theory – don’t hold on to the numbers.
So when preparing reports and particularly presentations pepper them with real lifanecdoteses and stories and leave the equations and formulae in the trash bin.
Note: The story was related in Brue's "The Evolution of Economic Thought, 5th ed.," pg. 294.
Tuesday, November 14, 2006
Genesis Question
What is the one thing you should always ask technology vendor?
The beginnings of every software company shape its product and thinking for generations. The choices made in that first release will dictate what the product is good for and by inference what it isn't. Therefore, when evaluating software vendors ask them to describe what the original problem they attempted to solve and who that market was. You'll often find that after several iterations, the apple hasn't fallen very far from that particular tree. If you're looking for peaches, then find another tree.
The beginnings of every software company shape its product and thinking for generations. The choices made in that first release will dictate what the product is good for and by inference what it isn't. Therefore, when evaluating software vendors ask them to describe what the original problem they attempted to solve and who that market was. You'll often find that after several iterations, the apple hasn't fallen very far from that particular tree. If you're looking for peaches, then find another tree.
Monday, November 13, 2006
Marketing FICO Scores
What can we learn from the underwriting business?
Credit scores are based on statistical analysis of account and payment data. Generally they focus on determining the likelihood of making a bad decision when underwriting a loan. Higher risks are associated with higher returns for the lender - or larger payments for us. This is all then translated into a score -- FICO or the new VantageScore. What is clearly not discussed is how the score is actually created nor the accuracy of the underlying model -- those are trade secrets.
A quick look at the real estate or automotive section of a newspaper shows that whatever statistical tools are used and their validity, the results have real economic impact.
A conference on the Validation of Consumer Credit Models made the point.
Credit scores are based on statistical analysis of account and payment data. Generally they focus on determining the likelihood of making a bad decision when underwriting a loan. Higher risks are associated with higher returns for the lender - or larger payments for us. This is all then translated into a score -- FICO or the new VantageScore. What is clearly not discussed is how the score is actually created nor the accuracy of the underlying model -- those are trade secrets.
A quick look at the real estate or automotive section of a newspaper shows that whatever statistical tools are used and their validity, the results have real economic impact.
A conference on the Validation of Consumer Credit Models made the point.
Rather than establishing some arbitrary statistical criteria for a model's performance, the central question for validation is whether the model is working as intended and producing results that are at least as good as alternative approaches.The above is good advice for propensity models - be they acquisition or churn. We shouldn't talk about r-square, K-S, or goodness of fit but is our customer base growing the way we'd like?
Profitable Experiences
Is the tail wagging the dog?
A recent article by Martha Rogers of 1to1 fame discussed the benefits of Continental Airlines' loyalty program. In it there is the following:
A recent article by Martha Rogers of 1to1 fame discussed the benefits of Continental Airlines' loyalty program. In it there is the following:
Each customer has an associated customer value metric, from one to 100, based on the profitability of that passenger. By treating different customers differently based on their value, Continental gives each group the experience its members are looking for.But here's the rub, the experience people look for may only loosely relate to their profitability. Certainly experiences have costs and aligning the rewards and perks with value is an appropriate goal. The trick is to figure out how to make the desired experience profitable and not the other way around.
Friday, November 10, 2006
Targeting a Small Car Niche
When are less sales more?
In today's Wall Street Journal, an article described the marketing plans for Toyota's Scion. It seems that they are limiting next year's production to below this year's sales and shifting budget to event and experiential avenues. They even find MySpace as being too mainstream - imagine that, a car manufacture eschewing social media as so 2005.
Typical sales analyses and forecasts usually end up in trying to justify increases in unit sales or share under the premise that more is always better. It takes judgment and executive decisions (and a bit of moxi) to reject the spreadsheet models in favor of what's important.
In today's Wall Street Journal, an article described the marketing plans for Toyota's Scion. It seems that they are limiting next year's production to below this year's sales and shifting budget to event and experiential avenues. They even find MySpace as being too mainstream - imagine that, a car manufacture eschewing social media as so 2005.
Typical sales analyses and forecasts usually end up in trying to justify increases in unit sales or share under the premise that more is always better. It takes judgment and executive decisions (and a bit of moxi) to reject the spreadsheet models in favor of what's important.
Restaurant Marketing
What can we learn from Chefs?
"Heat: An Amateur's Adventures as Kitchen Slave..." quotes Mario Batali as describing the business in its rawest form as buying food, fixing it up, and making a profit. I guess the rest must be marketing.
Anthony Bourdain in "Kitchen Confidential" makes a strong case why creativity and innovation are NOT desirable when hiring line cooks or sous chefs. Customers want the same thing they had last time, they are not looking for a creative variation.
It seems that a successful restaurant is about understanding expectations of a target market and then consistently meeting (or exceeding) them. Sounds familiar.
"Heat: An Amateur's Adventures as Kitchen Slave..." quotes Mario Batali as describing the business in its rawest form as buying food, fixing it up, and making a profit. I guess the rest must be marketing.
Anthony Bourdain in "Kitchen Confidential" makes a strong case why creativity and innovation are NOT desirable when hiring line cooks or sous chefs. Customers want the same thing they had last time, they are not looking for a creative variation.
It seems that a successful restaurant is about understanding expectations of a target market and then consistently meeting (or exceeding) them. Sounds familiar.
Thursday, November 09, 2006
Selecting Square Pegs for Round Holes
Why is marketing analytics a struggle?
One of the underlying reasons for the frustration rests on the fact that technology and software in particular is extremely biased. It is (was) developed with a particular objective or goal in mind. If the current need has a different goal then trouble will emerge, usually sooner rather than later.
Consider the case of the marketing database. Today most systems are built on some form of relational database and usually have a 'customer table.' When drawn on the blackboard customers appear as rows and their attributes appear as columns. There is no bias in favor of rows or columns at this point - I should be able to add a new attribute just as easily as adding a new customer. However, when this logic was first implemented the designers focused on transactional applications where there were lots of rows coming and going, each with a stable set of attributes - think credit card transactions, reservations, retail sales, phone logs, etc. These systems optimized manipulating a row at a time for both speed and audit reasons. So it is easy to insert, update or delete a single customer record. This legacy still exists in nearly all commercial products.
But, and here comes the effect of that bias, marketing analysis is about comparing and contrasting groups of customers. How do responders differ from non-responders? What is the appropriate contact strategy? Who should we target? etc. This style of work tends to focus on the attributes of the entire customer base rather than a single customer at a time. As a result, marketing analytics needs an entirely different set of database functions.
In the language of database developers the commands for manipulating rows are generally much more powerful and flexible. The ability to add, amend, or drop a column (particularly by a user) is nascent at best.
I'd like to be able to add columns (propensity scores, various segmentation schemes, current and future life-time value, etc.) with the same ease that I can add another customer record.
Disclosure: I used to work for a firm that works on solving this problem and know what is possible when the emphasis is turned 90-degrees. There is also a start-up focusing on part of this topic as well.
One of the underlying reasons for the frustration rests on the fact that technology and software in particular is extremely biased. It is (was) developed with a particular objective or goal in mind. If the current need has a different goal then trouble will emerge, usually sooner rather than later.
Consider the case of the marketing database. Today most systems are built on some form of relational database and usually have a 'customer table.' When drawn on the blackboard customers appear as rows and their attributes appear as columns. There is no bias in favor of rows or columns at this point - I should be able to add a new attribute just as easily as adding a new customer. However, when this logic was first implemented the designers focused on transactional applications where there were lots of rows coming and going, each with a stable set of attributes - think credit card transactions, reservations, retail sales, phone logs, etc. These systems optimized manipulating a row at a time for both speed and audit reasons. So it is easy to insert, update or delete a single customer record. This legacy still exists in nearly all commercial products.
But, and here comes the effect of that bias, marketing analysis is about comparing and contrasting groups of customers. How do responders differ from non-responders? What is the appropriate contact strategy? Who should we target? etc. This style of work tends to focus on the attributes of the entire customer base rather than a single customer at a time. As a result, marketing analytics needs an entirely different set of database functions.
In the language of database developers the commands for manipulating rows are generally much more powerful and flexible. The ability to add, amend, or drop a column (particularly by a user) is nascent at best.
I'd like to be able to add columns (propensity scores, various segmentation schemes, current and future life-time value, etc.) with the same ease that I can add another customer record.
Disclosure: I used to work for a firm that works on solving this problem and know what is possible when the emphasis is turned 90-degrees. There is also a start-up focusing on part of this topic as well.
Moths to Light Bulbs
Do new tactics risk losing one's direction?
Moths confuse light bulbs with the moon and hence lose their sense of direction.
In an article on new media Douglas Rozen makes a similar point about marketing when he asks 'when did "cool" become a marketing metric?' The point is that just because something is bright and shiny it doesn't necessarily make it relevant to marketers or consumers. He uses examples about consumer-generated content to make the point - sponsoring ads for the Super Bowl or MySpace/YouTube videos.
It's a difficult balancing act. Leveraging new channels, interaction mediums, cost structures, etc. will always be uncharted territory. Until some semblance of understanding can be reached let's thank those willing to experiment in a whole new laboratory. But there should always be a clear direction and objective.
Moths confuse light bulbs with the moon and hence lose their sense of direction.
In an article on new media Douglas Rozen makes a similar point about marketing when he asks 'when did "cool" become a marketing metric?' The point is that just because something is bright and shiny it doesn't necessarily make it relevant to marketers or consumers. He uses examples about consumer-generated content to make the point - sponsoring ads for the Super Bowl or MySpace/YouTube videos.
It's a difficult balancing act. Leveraging new channels, interaction mediums, cost structures, etc. will always be uncharted territory. Until some semblance of understanding can be reached let's thank those willing to experiment in a whole new laboratory. But there should always be a clear direction and objective.
Wednesday, November 08, 2006
Equations as a Brand
So what does THIS mean?
With all due respect to Fair Isaac and the work they put in to raise the bar on quantifying customer behavior, the following banner portrays the wrong message.
As an analyst, my first thought was to try to figure out the equations.
As a business person, my first thought was to pass it down the line.
Guess which option won?
With all due respect to Fair Isaac and the work they put in to raise the bar on quantifying customer behavior, the following banner portrays the wrong message.
As an analyst, my first thought was to try to figure out the equations.As a business person, my first thought was to pass it down the line.
Guess which option won?
New Political Divide: Up vs. Down
Is Left - Right on the way out?
A quote this morning in Hotsoup sums up the issues many have with politics and its attendant media coverage.
While the up/down question is the right one, the Huckabee blog is still very left-right oriented. Guess the message hasn't gotten down to the handlers yet; so politics as usual.
The analogy sounds eerily similar to many analytic discussions with business leaders. They want to know whether we're moving up or down and often hear assumptions, technical limitations, qualityquatlity problems.
A quote this morning in Hotsoup sums up the issues many have with politics and its attendant media coverage.
"The average person in politics is thinking horizontally and talking about things right and left," Gov. Mike Huckabee of Arkansas, a potential candidate for the Republican presidential nomination in 2008, told HOTSOUP. "The average American thinks vertically - what they're interested in is not party ideology, but, 'Are you going to lift me up or take me down?'"
While the up/down question is the right one, the Huckabee blog is still very left-right oriented. Guess the message hasn't gotten down to the handlers yet; so politics as usual.
The analogy sounds eerily similar to many analytic discussions with business leaders. They want to know whether we're moving up or down and often hear assumptions, technical limitations, qualityquatlity problems.
Tuesday, November 07, 2006
Fantasy League for Marketing
How do you measure a changing market?
In yet another good read, Michael Lewis of "Moneyball" and "Liar's Poker" fame has written about the evolution of football in The Blind Side. In general its about game-changing events as they relate to a single position: the left offensive tackle. In particular its about a single individual who plays that position.
It also hints at the point that metrics only appear when the game changes in a substantive way.
If you ask how the defenses of the 50's, 60's and 70's compared to those of the 80's, 90's and 00's in terms of sacking the quarterback --- you're stuck. The NFL didn't keep individual sack statistics until 1982, which coincides with the advent of Lawrence Taylor (who didn't actually lead in that category until 1986) as the most feared player in football.
Multi-channel, integrated, or media neutral programs have changed the face of the market. What we don't quite yet have a handle on is how to measure the value of those campaigns. What metric reflects the impact of a campaign? Once we answer that, we could start a fantasy league for marketing.
In yet another good read, Michael Lewis of "Moneyball" and "Liar's Poker" fame has written about the evolution of football in The Blind Side. In general its about game-changing events as they relate to a single position: the left offensive tackle. In particular its about a single individual who plays that position.
It also hints at the point that metrics only appear when the game changes in a substantive way.
If you ask how the defenses of the 50's, 60's and 70's compared to those of the 80's, 90's and 00's in terms of sacking the quarterback --- you're stuck. The NFL didn't keep individual sack statistics until 1982, which coincides with the advent of Lawrence Taylor (who didn't actually lead in that category until 1986) as the most feared player in football.
Multi-channel, integrated, or media neutral programs have changed the face of the market. What we don't quite yet have a handle on is how to measure the value of those campaigns. What metric reflects the impact of a campaign? Once we answer that, we could start a fantasy league for marketing.
Language Skills: Statistical Abstraction
Does how we talk impact what we say?
In a recent DM Review article, the authors discuss different ways of accounting for two types of error based on using a sample: how tight the estimates are around a given value (precision) and how far the average estimate is from the truth (bias). Got to give them credit for avoiding terms like "stochastic" in the discussion.
However, what's missing in this discussion is why this matters from a business point of view. Or in the words of another economist - what is the 'oomph' of the analysis? It would have been helpful to understand scenarios where precision is more important than bias and vice versa. For instance what class of problems should focus on consistency, even if biased, versus overall accuracy?
This article left me thinking that it's no wonder firms like Williams Sonoma, Capital One, and Cisco are currently looking for senior people to manage customer information and analytical teams while simultaneously talking with C-level executives.
In a recent DM Review article, the authors discuss different ways of accounting for two types of error based on using a sample: how tight the estimates are around a given value (precision) and how far the average estimate is from the truth (bias). Got to give them credit for avoiding terms like "stochastic" in the discussion.
However, what's missing in this discussion is why this matters from a business point of view. Or in the words of another economist - what is the 'oomph' of the analysis? It would have been helpful to understand scenarios where precision is more important than bias and vice versa. For instance what class of problems should focus on consistency, even if biased, versus overall accuracy?
This article left me thinking that it's no wonder firms like Williams Sonoma, Capital One, and Cisco are currently looking for senior people to manage customer information and analytical teams while simultaneously talking with C-level executives.
Monday, November 06, 2006
Limits of Transaction Data
What can transaction data tell you?
On the one hand historic data can tell you a lot about a customer's behavior and with a bit of pixie dust can probably hint at the future.
However, it fails on two important fronts:
First, it is exactly what it is -- a firm's own transactions. Not category activity, not context, not reasons, and not intent.
Second, it is based on a given set of business rules and assumptions about how the world works. It does not support answering 'what if we changed the ...' type questions.
Without a doubt, marketing databases should include transaction history. Analyses built on transaction data confirm whether marketing activities impact behavior, at least at the macro level. But like many puzzles, it is only a starting point so don't stop there.
Like John Mayer, I'm Waiting on the world to change.
On the one hand historic data can tell you a lot about a customer's behavior and with a bit of pixie dust can probably hint at the future.
However, it fails on two important fronts:
First, it is exactly what it is -- a firm's own transactions. Not category activity, not context, not reasons, and not intent.
Second, it is based on a given set of business rules and assumptions about how the world works. It does not support answering 'what if we changed the ...' type questions.
Without a doubt, marketing databases should include transaction history. Analyses built on transaction data confirm whether marketing activities impact behavior, at least at the macro level. But like many puzzles, it is only a starting point so don't stop there.
Like John Mayer, I'm Waiting on the world to change.
Merger of Like Minds
When does 1+1=3?
Last week two of my favorite reads, Marketing Experiments and MarketingSherpa, announced they were merging. The combination of real laboratory experiments and secondary research, including case studies, provides a significant resource for determining what really works. I wish Anne Holland and Flint McGlaughlin all the best with their new company.
Now, if only they'd add traditional media to the mix.
Last week two of my favorite reads, Marketing Experiments and MarketingSherpa, announced they were merging. The combination of real laboratory experiments and secondary research, including case studies, provides a significant resource for determining what really works. I wish Anne Holland and Flint McGlaughlin all the best with their new company.
Now, if only they'd add traditional media to the mix.
Friday, November 03, 2006
Ten Tips on Accountability
Tom Breur publishes a monthly newsletter on a variety of data-related topics. The October list covers Marketing Accountability.
Good thoughts with a few that run counter to prevailing buzzwords.
Good thoughts with a few that run counter to prevailing buzzwords.
Memo to Software Vendors
Can software tools analyze?
Just went through my notes from the DMA exhibit hall and the number of 'analytic' mentions is at an all time high. But at the end of the day analysis is about understanding and determining if those new findings matter. These are both human endeavors.
So, I'd argue tools facilitate the process but rarely actually do analysis. I'd also argue that most tools focus on the low hanging fruit of manipulating the data they can get their hands on rather than helping with understanding or oomph.
Just went through my notes from the DMA exhibit hall and the number of 'analytic' mentions is at an all time high. But at the end of the day analysis is about understanding and determining if those new findings matter. These are both human endeavors.
So, I'd argue tools facilitate the process but rarely actually do analysis. I'd also argue that most tools focus on the low hanging fruit of manipulating the data they can get their hands on rather than helping with understanding or oomph.
Thursday, November 02, 2006
The Challenge of Definitions
Just how do we define 'customer'?
One of the most difficult things in data-driven marketing is getting agreement on what some commonly used terms actually mean. It is somewhat unfortunate that databases and computers are totally ignorant -- they simply reflect what we tell them to hold and do, so we have to be explicit in our instructions.
The answer to the above question is the proverbial "It depends." Only somewhat tongue-in-cheek I've argued the following funnel:
Sales -- any warm body
Marketing -- anyone with a need
Legal -- anyone we're allowed to talk to
Finance -- anyone who can issue a purchase order
Accounting -- anyone who owes us money
Similar funnels can be created for terms like 'revenue' -- contracted vs. booked vs. collected.
Rightfully or wrongly data-driven marketing requires us to sharpen our definitions to the point that computers and databases give us the right answers.
One of the most difficult things in data-driven marketing is getting agreement on what some commonly used terms actually mean. It is somewhat unfortunate that databases and computers are totally ignorant -- they simply reflect what we tell them to hold and do, so we have to be explicit in our instructions.
The answer to the above question is the proverbial "It depends." Only somewhat tongue-in-cheek I've argued the following funnel:
Sales -- any warm body
Marketing -- anyone with a need
Legal -- anyone we're allowed to talk to
Finance -- anyone who can issue a purchase order
Accounting -- anyone who owes us money
Similar funnels can be created for terms like 'revenue' -- contracted vs. booked vs. collected.
Rightfully or wrongly data-driven marketing requires us to sharpen our definitions to the point that computers and databases give us the right answers.
Chart Spinning
Does the orientation of graphics make a difference?
In this morning's mail was the following humor.
Unfortunately, rotating 180 degrees still portrays the same trend. Only difference is that the speakers will be standing on their heads.
90 degrees in either direction might have worked though.
Why? Because most of us naturally read left-to-right so assume that time is meant to pass that way as well in graphs. Hence the use of the horizontal axis to represent time in trend charts. Fighting a 'natural law' like this makes interpretation and analysis extremely difficult if not misleading. But possibly a shorter meeting.
In this morning's mail was the following humor.
Unfortunately, rotating 180 degrees still portrays the same trend. Only difference is that the speakers will be standing on their heads.
90 degrees in either direction might have worked though.
Why? Because most of us naturally read left-to-right so assume that time is meant to pass that way as well in graphs. Hence the use of the horizontal axis to represent time in trend charts. Fighting a 'natural law' like this makes interpretation and analysis extremely difficult if not misleading. But possibly a shorter meeting.
Wednesday, November 01, 2006
Strange Bed Fellows: 6 Sigma and CRM
What do operations research and creative have in common?
Turns out the answer is quite a lot. Six-sigma, known for its statistical rigor focuses a lot on the customer because the fundamental tenet is this:
Customers deserve to pay the least amount possible while firms deserve to make the most possible.
As a result understanding precisely what customers need becomes a critical component. "Voice of the Customer" is a central theme with a clear focus on segmentation to better understand the nuances of need. So, maybe there are a few more things we can learn from the stopwatch gang.
Turns out the answer is quite a lot. Six-sigma, known for its statistical rigor focuses a lot on the customer because the fundamental tenet is this:
Customers deserve to pay the least amount possible while firms deserve to make the most possible.
As a result understanding precisely what customers need becomes a critical component. "Voice of the Customer" is a central theme with a clear focus on segmentation to better understand the nuances of need. So, maybe there are a few more things we can learn from the stopwatch gang.
Subscribe to:
Posts (Atom)
Posts
