marginally subversive

Why...

... Is a far bigger question, far more difficult to answer, and far more complex than a simple what or where or when.

For analysts - and I would associate myself with analysts more than planners after a decade and a half in this business - the biggest question is "why are things working or not working?"

Data can immediately show differences, trends, and similarities. Across time. Across groups of people. Across brands.

Models and equations can dissect components of a certain result into pieces and their future effects predicted with some confidence through true- or pseudo-simulation in a wargame or in a PC.

All these however are simple answering "what, when, where".

"Why" deserves a bit more than just crosstabs, coding, and modelling.

"Why" tells us underlying stories behind correlations. It tells us why certain interventions work and some don't. It tells us why certain scenarios existed in the past - and what elements would foment the same in the future.

That's what separates analysts from data crunchers.

don't be afraid of simplicity - and of complexity...

here we go again…

Media planners have always wanted to measure the “optimum level of reach and frequency of different combinations of media channels”.

The question is – to a certain extent – valid: What is the best way to manage finite financial resources across TV, radio, press, and other forms of communications to achieve the best impact?

It is not entirely a bad thing:  Once we know the answers – or the principles related  – to this question, we can then move towards optimization.

Given the fragmented media landscape and increasingly declining budgets, all the more that we should be answering this question and deciding based on those answers.

But should the internet and social media be included in these process?

 

 

TV, radio, presses and outdoor are still measured in terms of the ‘number of people who have seen (or who claim to have seen) the medium’.  That’s what GRPs – even PeopleMeter measures – are based on.

But the web – in all its myriad forms from banner ads, rich ads, search, action-seeking ads (either recommend, forward, click, save or download) – is an entirely different thing.

Whilst PeopleMeter and diary ratings are operationally defined as “people who click on a button in front of the TV to identify themselves” or “people who record their media exposures on a predefined diary”, the web is entirely different. 

 

Let’s not even talk about mobile phones – which encompass SMS, location-based and on-demand apps, bluecasting, mobile search, location-based push-SMS/messages…

 

The point is: Reach and frequency – and other OTS (opportunity-to-see/hear) metrics – on these “new” media.  It is not actionable nor relevant to say that “TV generates 70% of unduplicated reach, whilst the internet delivers an additional 15% reach”.

 

Nor can we say: “Site A delivers 35% reach, and with the additional Site B buys, we will have an additional 12% reach – resulting to 47% unduplicated reach of the internet campaign".

 

 

I understand the need for a single currency – because how can we optimize if we don’t have a single currency that would define goals and constraints?

But the call for a single currency shouldn’t disrespect the medium and its innate characteristics, strength, and roles in the lives of consumers.

 

There must be another way of creating a single, standard currency that applies to all.

In my opinion, that’s consumer response to the ads carried in different media.

research reviews and results

Why do marketers and branding experts have tracking in place?

So they can have a look at what's happened in the past - dissect it, cross-check the results with their goals and their objectives...

But knowing whether objectives were met or not, dissecting any campaign for the sake of understanding what went wrong and what didn't isn't enough.

We have to learn from it.

And be able to make decisions today about where the future is headed.

Any research presentations that look at the past and explains the past - but lack any predictive power or any inputs as to how we can shape the future - is incomplete and borders on being useless.

The whole point of looking at the past - and understanding it - is to make sure that we can make decisions now about our future.

Otherwise, it's all moot and academic.

the beauty of statistics…

… lies in its ability to provide us with frameworks and structures with which to see the world as it may be.

 

I am constantly barraged by questions on ‘goodness-of-fit’ measures.  People question models based on their adjusted r-squared’s, their AIC, their-D-W indicators and everything else that you can think about.  There are discussions that sometimes lead to frustrations from either the analysts’ part, the clients’, or the strategists’.

Everyone wants to be sure that the model is accurate.

That when 100 dollars are invested, indeed, there will be a return of 50%.

That when 200 dollars are invested, indeed, there will be 65% returns.

The thing is, every single model – regardless of how they were derived – are based on part observations.

And as far as I know, no one has seen the future.

 

There’s only one thing in my mind whenever I approach modeling projects (not the kind which involve photographers, human models, etc. – although I wouldn’t mind that kind of modeling project, too):  interpretation.

Can I explain the model?

Can the variables explain the movements?

Does the explanation hold?

Could it be the same in the future?

How can the model be wrong?

Are there alternative interpretations?

  

 

Here’s a wonderful quote from Andreas Buja:

 

A lot of people think that analytics – done by statisticians and econometricians – are all about creating models with ‘accuracies approaching 100%’.

I think analytics is all about interpretation that empowers others (and me) to more deeply understand, appreciate, and explore an idea.

 

After all, analytics people are still humans – and they have not seen the future.

The principles of Behavioural Economics can make communications work harder says Janet Hull, consultant head of marketing at the IPA..

via mediatel.co.uk

I am quite surprised that it is only now that marketing practitioners are actually asking the question whether behavioral economics, communications, and marketing are linked.

Hmmm. Isn't that obvious?

OK. Let's say for the sake of argument that a marketing practitioner is not familiar with "behavioral economics" and what it is.

But surely, one has studied in university basic economics - and what is economics? The study of the allocation of finite resources to optimize utility - however "utility" is measured (e.g., returns, production, revenues, production, happiness... ok, perhaps not the last one...)

Isn't it obvious then that economics has a place in marketing and in communications since everyday, we deal with decisions on how to optimize finite budgets to get the best returns on marketing moneys?

Now put in behavioral: That suggests... people. Individuals. Consumers.

Put the two together: study of people and individuals as consumers or audiences who are faced with options and are motivated to maximize their finite resources - money, time, attention, emotions - to generate the best returns.

I thought discrete choice theory, conjoint analysis, and similar techniques - and Bayesian statistics that allow for hierarchical and individual level analytics - are beginning to be a part of the world of marketing and communications.

Red herring or red hot? C'mon.

================================
What is behavioral economics?

From Wikipedia:

Behavioral economics is that branch of one,which deals with the study and application of analysis with scientific approach on social,emotional factors for understanding the consumers,investors and the market,and the resources.

The field is primarily concerned with the bounds of rationality (selfishness, self-control) of economic agents. Behavioral models typically integrate insights from psychology with neo-classical economic theory. Behavioral Finance has become the theoretical basis for technical analysis.

Measuring Outdoor Channels' Effectiveness

ADGist™ measures the effectiveness of print and billboards in a single glance. It’s a more realistic way to test ad effectiveness since most print and billboards get only a single glance when a person is flipping through a magazine or driving on the road. It is often thought that ll the money spent on the majority of ads that people don’t pay much attention to is wasted, but is this really the case? ADGist™ assesses that.

via marketresearchbulletin.com

... by measuring the initial impact (or "impression") of the ad. The original research is from PROMPT RESEARCH INSIGHTS (http://promptresearchinsights.com).

My take: Interesting - and potentially powerful in a very attention-deficit age. But what would interest me more would be how the indicators that AdGist comes up with are related to actual behavioral and/or attitudinal shifts.

Outdoor - which is one of the fastest growing media in the APAC region due to changes in audience lifestyles [more mobility, less in-home activities] and technological opportunities [caveat: I consider bluecasting, infrared casting, location-based SMS/MMS - whilst on the mobile phone - outdoor] - needs to be more measurable and 'tradeable'.

I firmly believe that

1. buying posters based on 'anecdotal, location-based justifications' should end;

2. approximating "GRPs" and "deweighting" them based on self-reported attention to the medium in order to justify our media plans should stop; and

3. transforming print ads - images, layout, text, and all - into outdoor posters must also be trashed.

I hope that ADGist(TM) could help with that.

on social networks and complexities

via iq.harvard.edu

And now something related about work...

Whilst on holidays, I have been trying to read up on two work-related themes (in between my Paulo Coelho, Dan Brown [not a fan], and a few other 'esoteric' readings): Network Analysis and Complexity Science.

Two learnings:

1. Network analysis/science is not easy. It requires deep thought - and very unconventional approaches and algorithms. Most of the articles I have read have emphasized that their work is only the beginning - and that there is more to learn.

... Which leads me to the current belief in advertising that "so long as you have 'word-of-mouth' in your communications campaign, you'd be fine..." and "oh by the way, let's create a page on (insert your favorite social-networking site) and an app for iPhone".

I think this is simplifying the beauty of social networks and their impact on communications, marketing and ad effectiveness, and consumer behavior and emotions significantly.

Approaching social networks as if they were another medium that can be controlled is - in my belief - simply the wrong way. Not just because we don't have control over the social medium as we do on other media - but because social media affect people differently because they work differently.

2. Complexity is the nature of things - and whilst we may want to simplify our communications through visuals and bullet points, our message will remain to be complex.

We can create PowerPoint slides filled only with pictures and single-words or lines - because it is simple - and simple is beautiful, and is effective.  We can create simple ads - filled with white space or capped with a catchy tagline or a jingly.

At least that's what we are taught - cognitive science says "schemas work - so leave them with schemas that they can bring along with them and process".

But it does become really complex because of the last two words: "... and process".

No matter how complex or simple our message is, it will cross over to complexity as soon as human processing starts.

A 10slide presentation - that is simple - could have as much complexity as a 100slide presentation. And the length, or style, or manner doesn't really matter: it's how the audience processes them that matters - and that adds that significant layer of complexity.

The same is true with social media and social networks: A seemingly simple 5-friend group (5node network) is very complex - and its complexities could well be similar with a 10-friend, 20-friend, or 100-friend network.

Because?

The human element is there.

We are born social - we are innately social.

And some are more social than others with a selective group - regardless of whether they already belong within the same group or not.

Adquants and Creativity: Opposite sides?

(This post is inspired by an article on CNNMoney.Com on adquants.)

A lot of people think that people are either predominantly right- or left-brained.

If one is predominantly right-brained, he or she is more likely to be creative, more "feeling"-orientated, more open to new things, and more open to the possibility of the absence of structures.  Organized chaos, according to common wisdom, is their mantra.

The left-brain orientated ones are far more structure-driven, more about equations and proofs and evidences, more about organization - less about chaos.  For left-brain orientated individuals, they seek evidence and proof more than hunches and gut-feeling.This belief I think is at the root of the "adquants versus creatives" debate in advertising.

Creativity demands an openness to possibilities and connections that no one ever thought existed.  For that to happen, it is believed that one has to be as "right-brain"-oriented as possible.

But these challenging economic times demand that marketeers be prudent and judicious in their investments of marketing dollars.  The "left-brains" then come in - because they are the ones who have the numbers, the scenarios, the 'evidence' whether certain decisions are are logical and thoroughly-researched, and would increase the probability of success.

Is there - however - a middle way?  Can adquants and the creative types co-exist?

My answer is yes.

And it goes beyond more than just making the two sides be in one room, in one department, or in one meeting.

Both will have to learn from each other in their quest for an effective marketing campaign.

• Adquants have to look at the ability of the creative types to create stories out of numbers - to weave stories based on numbers, and understand - beyond t- and F-tests, beyond eigenvalues and eigenfactors, beyond regression lines and probability - the impact of their formulas and equations.
  
  Adquants need to adopt the ‘right-brain-ness’ of the creative types to go beyond stat tables, the software results, spreadsheets and simulations, and all the crunched numbers to answer questions that creatives have also been tasked to answer.
  
• Creative types have to look at the ability of adquants to dissect problems and to work on a "question-hypothesis-test-course-correct" basis, and be open to possibilities that may seem to be "uncreative".  They will have to learn the structures, limitations, and findings that adquants create out of their inquiry of the world (from a numeric perspective) – and move within it.
  
  When creative types create solutions within the parameters that adquants may have uncovered – and know the parameters deeply so they can be incorporated into a campaign (and broken, if necessary), that’s when they will have worked together effectively.

The greatest challenge in the adworld today is not merely to marry the two disciplines - it is to marry the two mindsets that define those disciplines.

This reminded me of a conversation I had with a friend of mine from college on phenomenology, something that I think both adquants and creatives are very much involved in:

In phenomenology, you are essentially trying to explain something possibly unexplainable as objectively as you can whilst admitting that your mind cannot be perfectly objective.  To do this, you break down the phenomenon into its pieces – you create abstractions – and you understand each piece individually and how each relates with other pieces.  Then after you’re done, you bring it all together again – knowing that even if you have done so to the utmost of your ability, there is still more of that phenomenon that you cannot – and perhaps, never will – understand.

That is what I believe is necessary for adquants and creatives to coexist, work together, and be better thinkers and solutions-providers: the ability to think within – and outside – the parameters their expertise demand.

Photo from Quasimondo on Flickr.Com

what is "REAL" econometrics-based analytics?

A number of my friends have told me that the sexiest job in the world soon would be that of a statistician. 

I am not a statistician - I am a researcher who happened to believe that statistics is a tool that could be used to extract information and uncover insights from phenomena in the world.  I also happen to be interested in all things "numeratic": linear algebra, calculus, ordinary differential equations, operations research, queuing theory....

I have used statistics since day1 in my first job as a media strategist at Basic Ad, working on accounts such as McDonald's which demanded that their media plans be as accurate as possible (i.e., gotta use linear regression to predict with certainty the ratings) and which encouraged us to assess the relationship between their sales and advertising promotions.

Since then, I have not looked back.

A few years ago, I got introduced to the theory of econometrics - how statistical principles are used to explain economic phenomena, either in the macro- or microeconomic setting.  Then there was the growing interest in the marketing field towards using econometrics to justify media budgets and laydowns, as well as channel selections and all that.

It took not that long enough to realize that well, anybody who has the software and the patience to go through some wonderful things with numbers using matrices and spreadsheets, can do any model.  Heck, I even fell into the same trap - creating models that didn't make any sense.

That is when I realized that econometrics is all about - and perhaps, what statistical modeling on which econometrics is based: 

Creating, formulating, and coming up with equations relating two or three or more variables can be done by almost anyone who is interested.

It is like anybody who has the capacity to understand the saying (questionable, it may be) that "the past could be a good indicator of the future" and has the mathematical interest and bandwidth to create models can.

Take this case for example:

I am fascinated by the movements of the stock markets.  (And I am not an expert in finance nor in equity trading.)  Particularly, I was interested in the movement of the stock market.

I hypothesized - urged on by the current global economic crisis - that stock markets must be related one way or another:  That the Straits Times Index (STI) of Singapore should be affected by movements in the NASDAQ, the S&500, FTSE, VIX, and others. 

My hypothesis was that investors in the STI do take an interest in the overnight performance of these indicators: NASDAQ and the S&P500 were indicators for the US, FTSE is for the Euro markets, and VIX is an indicator that somehow reflects the "risk-attitudes/risk-expectations" of investors.

I therefore hypothesized that when the NASDAQ goes up and the S&P500 go up, and so do the FTSE - the STI would also go up.  But since investors are risk-averse, I also thought that the higher the VIX (and therefore the higher the expectations of risks and volatility in the future), it should impact the STI negatively. 

With this very little amount of information and hypothesis, I went on and crafted my models.

(Hold you criticisms of multicollinearity and autocorrelations...  For now, I am loosening the assumptions for my simple experiment.)

So... after several attempts, I got the following first model:

The model used the abovementioned weekly variables - FTSE, NASDAQ, S&P500, VIX - as predictors and ST as the variable.  (I also created an index - a NASDAQ/S&P500 index - since NASDAQ was more tech-orientated and SIngapore is more closely related to the tech-orientated indutries.)

Using data from 2000, the model (in red) seemed to track the data fairly well.  It was had several underpredictions at first, which attributed to the post-tech-bubble and 2001-2002 uncertainties.  The goodness-of-fit indicators were acceptable.  Adjusted R-squared is 0.721 - not bad!  The residuals were also fairly good.

(Again, hold off discussions of autocorrelations and multicollinearity.)

I figured, "Hmm, there must be a better way - could there be a better way?"  Can I increase my R-squared and other GIFs?

So after several (perhaps 25-30) iterations, I achieved the following:

 

Not too shabby, I said to myself - indeed, the GIF's increased.  Amongst them, an R-squared of 0.976 - for the movement of the STI.

(To the technicaly savvy: This was arrived using the same variables, but this time using weighted least squares.)

Obviously, I was giddy - I may just have modeled the movements of the STI in the past 9 years with simple mathematics and good old Microsoft Excel (and an add-in called XLSTAT).  I have derived the respective weights and parameters of each of the variables - and established their significance at an alpha-level of 0.01! (In some cases, some variables are actually less than 0.01!)

Then the realization crept in:

But modeling - econometrics, statistical modeling - is beyond creating models and equations.  It is creating explanations of why things have moved and acted that way in the past in order to clearly see the future.

And that was when I got into trouble.

 

So... the above are the most significant positive predictors.  The "FTSE x SP500" taken together explains a significant amount of volatility - not my hypothesis realy, but close enough.  But surprisingly, VIX also had a significant positive effect.  Not only that: If we multipled NASDAQ with VIX, there is an almost equally strong effect on the STI.  Squaring the NASDAQ and dividing it by the S&P500 also indicated that, well, it explained the changes in the STI - at an alpha level of less than 0.0001!

It became interesting when I started looking at what I called the "STI dampeners":

S&P500 and FYSE on their own were dampeners - and could shave off values in the STI.  Multipliying VIX with S&P500 also indicated a significant decline.  The most perplexing one was the last variable: VIX (risk-perceptions/expectations) x NASDAQ cumulatively divided by the S&P500 was the most significant dampener!

And again, modeling - econometrics, statistical modeling - is beyond creating models and equations.  It is creating explanations of why things have moved and acted that way in the past in order to clearly see the future.

And that was when I got into trouble: why?

Why would these relationships exist?  Why would the FTSE and the S&P500 be dampening the STI?  Don't we get cues from these two markets?  NASDAQ - on its own - was not a significant predictor, with t-statistic that is way below the acceptable range.

One could argue that it is the complexity thinking of investors - but still, why would VIX not be a dampener?  And it is only when one multiplied VIX with the S&P500 that I could accept that VIX is a dampener for the STI!

Econometrics is about why's - and not just the numbers and formulae.

I am no finance expert - I have a very basic understanding of the equity markets.  I have some knowledge of finance theory.  But I just cannot explain these numbers - and the formulae that I have derived.

If you ask me what the numbers and the formulas meant, I wouldn't be able to explain.  I could only say for certain that the suggested model from the equations tracked the real data well.  I could recite - in a non-monotonous voice (because I try to be not too boring when I present data) - the effects of each in terms of numbers.  And perhaps, even come up with an interesting visualization of the impacts of each of the factors and their interactions.

But ask me why VIX is positive on its own but now when it is multiplied with another variable - and I'd be lost.

Because econometrics - and statistical - modeling is all about uncovering the stories behind the numbers and explaining them.

Sure, because I now have the formulae for predicting the future of the STI given varying levels of the FTSE, the NASDAQ, the S&P500, and the VIX indices (predicting them altogether is another story), a software can be created, risk-tables associated with varying levels of the predictor indices can be programmed, and decisions can be made.  One can even make 'self-correcting', dynamically-adapting models from these models.

But as to why these formulae hold, I have no clue.

And that is where I would fail against the tenets of econometrics and statistical modeling:  The explanatory power, the stories, the why's and the how's and the how-come's.

That is where people's brains - and dare I say, people like me - come in.  Because after all the formulas have been established and tested, after all the parameters have been derived, and their individual and aggregate/combined effects on the predicted variable have been uncovered, the question then becomes "Why?"

The bottom-line:

• Whether we like it or not, theory and hypotheses before even the modeling starts is critical.
  
• Whether we like it or not, computers - for now - cannot take the place of creating explanations and stories.  Maybe, there will be a time when stories can be crafted by computers.  But for now, brains still matter.
  
• We need both people with technical and theoretical statistical and econometric background - but in the field of finance, of consumer behavior, of marketing... we also need experts who can ground these formulae into the real business world.  Into answers for real business questions.

To paraphrase Joyce Kilmer:

Models are made by fools like you and me
But only real integrative minds can make it gounded on reality.

---------------------------------------------------------------

PS. I got the data from Yahoo! Finance (http://finance.yahoo.com) which I think is one of the more friendler consumer-orientated finance sites on the web for all the indices from January 2000 till August 2009. 

In the first model, I used OLS, for the second model I used Weighted Least Squares with the last observations weighted 10x more than the older data. 

The residuals, by the way, for the first model are slightly heteroscedastic.  The second one fared better.  Both models, however, fail the Durbin-Watson Tests, which means they are autocorrelated at order 1.  The models also fail the more general Breusch-Godfrey Test, which means there are higher AC-orders.

My suggestion: use time-series if you are really interested.  ARIMA would do you good - and GARCH modeling.