Friday, September 29, 2006

Analytic Paradox

Why are simple questions so difficult to answer?

A few years ago I wrote an article about this one. The basic premise was that a lot of marketing questions like "how have our customers changed?" weren't easily defined in terms of predefined 'dimensions and facts'. The argument was made that the available tools didn't fit that kind of question; they were built for more structured ones like "what is the sales split by standard segment?" While important, it wasn't going to help find something new that could be leveraged.

At the time I argued that
The technology should be indifferent to what you ask and where you want to go. It should be just as adept at manipulating the non-buyer as it is the buyer, such as creating a segment of people who haven't bought in the past six months. There should be no proscribed paths for navigation. And all of this must perform at train-of-thought speeds so that the "Aha!" moment isn't lost while watching the hourglass, or worse, waiting for a new report or database to be built.

In short there should be no penalty for asking a new question. Still think the above is true, but rather than a tool targeted just to marketing there are general needs in the business for such flexibility. Maybe the next generation of tools will mutate enough to drive the marginal cost of a new question to zero.

Type Casting Errors

What's in a phrase?

A recent post in the Direct Newsline recapped a press release from Lyris saying that email "False Positives" were still a problem. While the overall point is there is too much junk mail, I wondered just what exactly does this mean?

Had to go dust off a few things to figure that out.

Consider the question: “Is this email spam?” If this statement is true, then I will block the email.

Reality is either True or False in that the email either is (true) or isn't (false) spam; however, the test could be wrong in one of two ways.

  • I let it pass when I should have blocked it. This is False Negative- a spam email I have to suffer with.
  • I block it when I should have let it through. This is False Positive- a good email lost.

Note that the question could be asked the other way: “Is this email safe to pass on?” In this case ‘truth’ is associated with passing it along.

  • I let it pass when I should have blocked it. False Positive- more junk email.
  • I block it when I should have let it through. False Negative- less requested email.


False Positive means saying yes when you should have said no.

False Negative means saying no when you should have said yes.

Confusion over a phrase like "False Postives" arises because it depends on whether the question is either "is this spam" or "is this from a requested sender." In the first case higher false positive rates relates to loosing good email in the second it means more spam.

The consequences of being wrong are often very different depending on which side of the question you land. The American judicial system knows this well -- the emphasis is to err on the side of innocence not guilt. Consider business email vs. personal email at home where the need for filtering may be different. At work, I'll take a little garbage to make sure I get all of what I need. At home, I'm more likely to eliminate all garbage even at the risk of losing something that might be of interest.

This leaves us with two specific questions with respect to managing the cases where we are wrong.

1. What specific question are we answering?

2. Which type of error is more critical to the business and customer relationship? Blocking too much or allowing too much?

Both questions are important and should be considered in deciding how to filter email as well as any other activity where decision rules are put in place.

Thursday, September 28, 2006

The Killer App is Killing Us

The spreadsheet has been heralded as the killer app that drove PC penetration into business (not to mention the Blue brand). The argument was that it reduced errors in manual computations and made repetitive calculations faster. Both probably true.

However, spreadsheets contain errors. For some of the more interesting ones see the web site that tracks them. Articles summarizing research shows an error range of 20-40%. Estimates of $10 billion in errors have been reported. From personal experience I know of two audits of 'production processes' that used spreadsheets where 100% (read "every d*** one") contained errors. We simply weren't consistent across projects or markets. In both cases, real money was at risk since they controlled pricing and product quality.

Bad spreadsheets aren't an IT issue; they are a management issue and go to the heart of data-driven marketing. Please have an outsider from your department look at the mechanics and logic of your spreadsheets.

Wednesday, September 27, 2006

Left Brain Marketing Requires Right Brain Technology

In a joint think tank session with former Forrester analyst Eric Schmitt, the author of Left Brain Marketing, I argued that even with the move toward more analytic rigor, i.e. left-brain dominance, marketing tools still needed to be right-brain driven. Why?

The nature of marketing is to change the status quo: New markets, new products, and new customers. Moving the needle on proverbial dashboard means trying, experimenting, and testing new things which in turn requires extremely flexible tools. There just aren't that many out there that support the style of analysis and questioning that marketing needs to do.

Marketing must be able to define its requirements at the time the question arises, not six months earlier. So when evaluating tools, do it live with real data and real, unplanned questions.

Not Very Insightful

What's wrong with this picture?

I recently had the opportunity to review a customer profile analysis. The two basic questions were:
1. How do my customers differ from the general audience?
2. Are there differences across our brands?

I suppose that the results would help provide input and insight into all types of issues -- branding, promotion, creative, etc.

For each demographic variable, there were five pages that looked like the above. Points:
1. There was no summary answer to the questions.
2. It took much longer to review than necessary.
3. It was boring, making it even more difficult to answer the questions.
4. No added value since I don't consider converting a table to a chart a valuable service.

Unfortunately the analyst (or was it just someone used to working with desktop tools) missed the objective of the exercise entirely, which is to help us understand and provide guidance. A stack of charts doesn't cut it.

When preparing a presentation, reorganize material to fit the needs of the audience, rather than replicate the data crunching process. I would have liked to have seen two pages:

a. Comparison of customers to the general audience; highlighting where they differ.
b. Comparison across products; again pointing out differences in profiles.

The human brain, and hence the mind, is capable of absorbing a lot more than we give it credit for. Check out discussions and sites on 'visualization' starting with Stephen Few's "Perceptual Edge."

Finally, when presenting results be sure to answer the questions that funded the work in the first place.

Tuesday, September 26, 2006

Statistics Defined

How do I explain it?

As "Reducing risk by quantifying uncertainty."

For a variety of reasons we often use a small group of people to learn about the larger group from which it is drawn. Think A/B testing for example. To ensure we can generalize the results from a sample we apply statistics to estimate how confident we are in what we find.

A more complete defintion can be found on wikipedia.

Predictive Modeling

Just what does this term mean?

Simply put, predictive modeling puts truth in the phrase "the only thing new is history we haven't learned yet." It is a series of tools and techniques for a) figuring out what we don't know and b) how marketing variables impact choice.

Articles and posts often talk about it in the future sense -- will this customer churn? will that customer buy? While perfectly valid, predictive modeling is just as valid for figuring out the past and present as well. Estimating income, identifying visitors, and depicting fraud are all examples of predicting in the present tense.

In a lot of cases the result of the exercise is expressed as the likelihood that a customer will behave in a chosen manner or is somebody we've seen before or has a certain level of a desired trait. This approach has the benefit of allowing us to rank order customers or prospects as well as understand the levers that drive behavior.

How good that prediction is will be left to a future note.

Scientist as Brand

Technology companies often take the work of an academic or scientist and turn it into a product. In some cases the person becomes central to a specific brand; in others they form the foundation for a set of products.

Vladimir Vapnik is noted for his work in data mining products from KXen.

Genechi Taguchi, a Japanese expert in process manufacturing, is now often associated with several companies offering multi-variate testing of web sites.


Acronym Confusion

What's in a name?

For some reason we like to use acronyms, particularly three-letter ones to define something. Not sure why since it often leads to unnecessary confusion or debate. Do we really care about the nuances between Corporate Performance Management (CPM) and Business Performance Management (BPM)?

Their use creates some strange online results. Search for SEM and you get:
  • Society for Experimental Mechanics
  • Scanning Electron Micrsocopy
  • Society for Ethnomusicology
  • SEM Products (something to do with automotive repair)
  • Security Engineered Machinery

Nothing to do with marketing or search. Given the state of context aware advertising, it is often possible to get ads for the above within a marketing site. Sorry about that pay-per-click, but had to follow some of those links.

TLAs are often an attempt to categorize things to make it easier to compare, but the boundaries are fuzzy. Certainly product and technology companies go this route. How often have you heard "We're not an SFA company, we're a CRM company." Maybe we should just say what we mean.


Monday, September 25, 2006

Your Opinion Doesn't Count

So whose does?

In a recent post, Anne Holland of MarketingSherpa discussed the benefits of personas to marketing. To add my spin, they represent the soul and embodiment of a target audience. A good persona represents a single individual; not a typical 'average', not a composite, but a living breathing person that we can relate to and empathize with. Since we're all different, any given scenario is likely to have 3-5 personas.

Technology firms often uses personas as part of the development process. For instance Teradata's Relationship Manager is even marketed as a persona-based solution. Starting in the 80s as a means of improving product design, Alan Cooper focused on improving development by explicitly defining the target user. He has written several books on the subject; one of which is appropriate for marketing and appropriately titled.

Written for a business audience "The Inmates are Running the Asylum" makes a simple point -- we should design products before we build them. And in order to design, we must understand for whom we are building. The discussion goes on as to why software engineers are a different breed, making this more difficult than it ought to be. The point here is that marketing is no different: We must thoroughly understand with whom we are communicating.

To paraphrase the openings to Chapter 3 and 5: "It's harder than you might think to squander millions of dollars, but a flawed [marketing] process is well suited to the job. .... The real benefit of offering a well-[targeted] product or service is the fierce loyalty it generates in your clientele."

Now that's a plea for marketing to get involved.

New Word: Chartistics

The use (or misuse) of charts to claim validity or substance to a point of view.

See for other examples.

Wrong Conclusion

Can graphics lie?

They can certainly mislead. This chart was found in a discussion of evening news viewership.

Is the NBC audience significantly older than ABC's?

Not really. A table would a) take up less space and b) make the main point that the age profile is very similar: 59.7, 59.9 and 60.1 years. If five months makes a difference then I stand corrected.

Given that the median US age is 35+ and the adult median is around 45, where are the rest of the people getting their news?

Saturday, September 23, 2006

Legal Cocaine

Is this the brand name of the decade?

A recent introduction in the energy drink has a brand name with near universal recognition. Will it work?

Katie Couric and the End of Civilization

Can a news reader reverse a trend?

Nightly news audience is down. While the networks battle for bragging rights and advertising fees, the fact remains -- substantially fewer people watch the evening news.

Katie Couric's move to CBS generated a spate of short-term interest, but it isn't likely to reverse a 25 year trend.

Jared Diamond's "Collapse" discusses five reasons why civilizations collapse. He covers both historic and current societies ranging from Easter Island to Greenland to current day China and Australia. The core factors work over a long period of time and play off one another. They are:
  • Environmental damage
  • Climate change
  • Hostile neighbors
  • Friendly trade partners
  • Response to environmental problems
I wonder if there is something we can learn about the news world. Certainly the market or climate for news has changed; both in terms of outlets and content. Fragmentation and the rise of cable news media would qualify as the appearance of hostile neighbors for no other reason than the battle for audience. Friendly trading partners, i.e. advertisers, are looking for more relevance and return. What ever the reason, 'trust' in the news sources is also down. While the perceived political tendencies of individual outlets - left, right or somewhere else - may work to satisfy a specific target market; the longer term impact may be an overall erosion in the credibility of any news source. Four of the five factors suggest a collapse.

According to Diamond the fifth factor, how the society responds, is the key to survival. Only time will tell whether nightly news will reinvent itself or go the way of the Anasazi.

Analytic Community

The folks at Juice Analytics make some good points on data analysis along with some very practical examples using Excel.

Friday, September 22, 2006

TV Mash-up

Is YouTube the new network?

Ok, here's a mash-up from this season's new shows I might watch: Ron Trott of Justice (Fox) vs. Sebastian Stark of Shark (CBS). They're high-powered lawyers - one for the defense, one for the prosecution - who never lose. Since both take the same militaristic approach it should be over in one episode.

If we can create a dream football match for a commercial and feature length CG films, then we should have this by the end of the week. Right?

Sudoku-driven Marketing

Does analysis mean numbers?

There seems to be an assumption going around that analysis equates with numbers. Let's not forget that analysis means understanding, not necessarily quantifying. Data-driven marketing is important, if not critical; but it still requires augmenting the data with thinking.

Recently Lane Michel of Quaero and I both wrote pieces that used Sudoku as a reference to the marketing analysis question. Lane used the analogy that the marketing database was like a Sudoku puzzle -- it starts with a lot of gaps in it that need to be filled. I took the approach that strategic analysis is as much about critical thinking and logic as it is about numbers.

Numbers can provide an excellent, and often defensible, guide to the decision making process - but they do not replace it.

Patenting Behavior

Can customer behavior be patented?

Using customer information for the betterment of the relationship makes all the sense in the world. Harrah's has joined the right pricing movement with a patented system that allows them to develop custom room prices based on total activity. It is said to allow them to '...more effectively use its hotel room inventory, helping maximize total revenue...'

Harrah's Awarded Customer Valuation Patent: Financial News - Yahoo! Finance

Two Long Tails

Do long tail customers buy long tail products?

Chris Anderson's work on the impact of removing distribution bottlenecks on product demand suggests that there are similar effects with respect to customers. Unlimited access is a two-way street - more products and more customers. But how do the two tails interact? Who are the customers in the tail? How do we change our marketing practices? Does targeting the most valuable individuals still out-weigh the value of the aggregate?


It's time I jotted down ideas that run through my head about using information and analytics to better understand what customers think, need and do. I'll try to keep items short with links to interesting things. Hope you enjoy them.