marginally subversive

Planning for Marketing? Think Finance - Think Lifetime Value

via www.youtube.com

Marketing has borrowed quite extensively from Finance in the last 10 years.

As marketeers, we've started talking about returns on investments - finding the "biggest bang for our buck".

Some (like my former team in a previous company) have started talking about risk management procedures - not just in managing contracts (and employing "options"-like buying techniques), but also in delivering gross impressions and shares.

All these are well and good.

Perhaps this is one way of the marketing department proving themselves to the finance department - an attempt to say that 'we are more than just an expense line on the balance sheet'.

But as marketeers and suport-teams of the marketing department, we have got to do more.

Employing evidence-based marketing, metrics-based objective-setting and results evaluation, real profitability analyses that take into account present and future value of marketing investments --- all these need to be more apparent in the marketing department's language and in the language of its partners in the advertising, media, promotions, PR, research, and digital space.

on #analytics: data availability is no longer the problem

Data availability is no longer the problem.  Accessibility is no longer the issue.  The cost of acquiring data has gone down dramatically.

Neither is the technology behind data collation the problem.  There are a myriad of options that one can collect, store, discover, collate, systematize, analyze, and use data.

The real problem is people's apathy and/or antipathy to data and data analyses.

It's all about changing the mindsets of business managers - specifically of marketing managers who think that their jobs encompass merely looking after advertising and media technologies.

David Campbell, as interviewed by Alex Howard (at Oreilly Radar) has this advice to clients about using their data:

From the data side, we've lived in what we'd call a world of scarcity. We thought that data was expensive to store, so we had to get rid of it as soon as possible. You don't want it unless you have a good use for it. Now we think about data from a perspective of abundance.

Part of the challenge, 10 or 15 years years ago, was where do I go get the data? Where do I tap in? But in today's world, everything is so interconnected. It's just a matter of teeing into it. The phrase I've used instead of big data is "ambient data." It's just out there and available.

The recommendation would be to stop and think about the latent value in all that data that's there to be collected and that's fairly easy to store now. That's the challenge and the opportunity for all of us.

(Emphases mine; from http://radar.oreilly.com/2011/12/david-campbell.html)

Here's the video of the interview:

 

 

NUMERATI is now on Facebook at www.facebook.com/Numerati.Solutions

NUMERATI is now live on Facebook.  Click through to www.facebook.com/Numerati.Solutions for more information.

NUMERATI empowers marketeers make sound, fact-based decisions using bespoke marketing research and analytics particularly in these challenging, continuously-changing times. 

NUMERATI does this through

1. The use of advanced statistics (attitudinal-behavioral-demographic-based segmentation, econometrics-/regression-based modelling)

2. Providing advice in optimizing returns on existing marketing research and learning programs

3. Designing marketing learning and research programs which include bespoke metrics development, dashboard development, qualitative/quantitative research results "deep-dives" and integration, and what-if simulation

4. Providing strategic planning advice and consultancy

5. Training and marketing-skills development

For more information, visit us on www.facebook.com/Numerati.Solutions

of chickens and eggs: is there a definitive answer?

I had a very interesting conversation with a good friend a couple of weeks ago about causation and inferring causal relationships haphazardly.

One of the questions that stuck in my head was this:

Are big companies big because they advertise a lot?  
Or do they advertise a lot because they are big?

Sounds like a chicken or the egg situation: which came first, the chicken or the egg?

(My friend thinks it's neither - it's the archaeopteryx.  To which I argued, "That was a bird." To which she said, "Nope, that's the ancestors of birds." ... I have weird friends, I know.)

But it is something worth pondering - because at the root of this question is the question of causality.

Causality is supposed to be simple:  

If A happens, B happens.  
Without A, B does not happen.
Therefore, A causes B.

Or is it?

What if there is a lurking variable C that is somehow related to both A and B, such that:

If A happens, C happens.
If C happens, B happens.
If A does not happen, C does not happen.
If C does not happen, then B does not happen. 

Can we still consider arrive the same conclusion that "A causes B"?

 

Let's make it even more complicated:

If A and C happen, B happens.
If A happens, B does not happen.
If C happens, B does not happen.
If A and C don't happen, B does not happen.
Therefore A and C happening cause B happening.

 But what if A and C are themselves related somehow?  Worse, what if A causes C?  Or what if there was another variable D that we have not considered that is making the presence-absence of A and of C?  Can we still arrive at the same conclusion?

I am no logician.  And I am pretty sure that there is an answer to these.  

I can already hear my psych professor going "Do a proper experiment in a laboratory setting eliminating all potential confounding variables, and measure properly - and do it several times - and get others to do it several times too".  But sometimes, experiments are not easy to do.

Even the most stringent of experiments still tend to have some flaw in it.

So what now?

All I can say, for now, is - 

And for now, I have no answer (because my brains are swimming in some melatonergic medicine).

 

 

 

 

Fitting equations...

I am fairly sure that if one tries hard enough, one can describe a set of seemingly random data points using an equation or a set of equations.

What I am not sure of is if one can explain WHY such an equation or systems of equations hold or should be true.

marketing analytics: some more thoughts...

it has almost been a year since I wrote about marketing analytics and my attempt at capturing the "principles" that could drive at making marketing decision-making far more evidence- and fact-based.

i am republishing these thoughts again, but with a stronger emphasis on what's really marketing analytics is all about: interpretation and transformative insights.

marketing analytics is far more than equations and cross-tabulations and significance-testing - although these techniques form part of that.  it is the interpretation that really matters - the "so-what?" and "what-now?"

i realized this after almost a year of not dealing with equations and wonderful world of econometrics, matrices, error-testing, and all that.  [well, *almost* not dealing with equations - i still had to fit curves against numbers and trends and all that beyond OLS and all that...)

equations, econometircs, statistical modelling, regression, arima, arma, time-series, curve-fitting, stat-tests, and cross-tabs are all means towards bigger end: creating an interpretation that makes a difference in the decisions that we will make in an uncertain world.

here are - once again - my thoughts on "marketing analytics principles" with some additional thoughts or comments in brackets

[i will keep the quotes until - well - I retire and/or leave the world of marketing analytics as i don't think i am qualified at this stage to set the principles as real principles.]

 

Marketing Communications Econometrics ‘Principles’
Draft by PTiongson • Wednesday, September 30, 2009

[comment: indeed, this was written on 30 September 2009; how time flies...]

1. Every investment in marketing has an effect on the company’s sales/revenues business.

The goal of econometrics in marketing communications is to quantify these effects and be able to compare-contrast the effects of investments in marketing channels in order to make decisions.

[we are not in the business of driving awareness.  we are not in the business of "being present on-air".  we are not in the business of being on-air where our buddies can see our products - thereby giving us the chance to gloat.  

we are in the business of business:  driving revenues, sales, profits.  

i will even go further and say that we are not in the business of driving impressions, ratings, share of 'voice', 'share of consumer experiences'... if we are not aligned on that, there is no point in investing in advertising.  and yes, investing - not spending.]

2. Revenues and marketing investments have don’t have a 1to1 relationship.

There is a point where additional investments in marketing will not have any significant effect on revenues or other marketing KPIs; beyond which, any additional investments are considered wasteful.

On the other hand, there is also a minimum threshold that needs to be met in marketing investments in order to ‘break through’, be competitive, and generate returns on marketing investments.

This is the economics behind marketing investments – economics is all about the optimal allocation of finite resources, and marketing econometrics is aimed at understanding the brand’s opportunities to optimally allocate finite resources that are not linearly related.

This is where the concept of S-curved and diminishing-return response rates comes in.

[i believe it is human for us to believe that if something is good, then more of that same thing should be better.  case in point?  money.  that same belief seems to hold in decisions that we make for marketing: if 200Mln USD of investments in advertising worked in the past - then we need 600Mln USD will triple our profits or revenues.  Not.

even in my non-regression/non-econometric work in the past have shown that more is not always better - and that when it comes to investing in marketing, bigger is not always better.  i would go on to say that the higher your investments in marketing, the higher the risks - but not necessarily the profits and the returns.] 

3. Investments in marketing communications have an immediate effect on a brand’s revenue.

Marketing communications spent in time period “t” has a measurable impact within the same time period “t”. Marketing econometrics is aimed at understanding the immediate effects of investments so decision-makers can course-correct accordingly.

[we are not in the business of driving awareness.  we are in the business of driving revenues, sales, profits.  if we are not aligned on that, there is no point in investing in advertising.  and yes, investing - not spending.] 

4. Investments in marketing communications have a ‘lingering’ effect on a brand’s revenue.

Marketing communications spent in time period “t” has a measurable impact after time period “t” – though such an impact may well be diminished and not so strong as the initial impact.

Marketing econometrics is aimed at quantifying the persistent effect of a campaign on a brand’s revenue through the use of concepts such as decay rates and ad-stock.

5. Marketing accountability – and therefore, econometrics – needs to take into consideration differences in execution and messages.

While econometrics is a quantitative technique and a science aimed at quantifying relationships between two or more variables, it should take into account the messages that are being conveyed by the brand.

This may be achieved by incorporating dummy variables that code different campaigns of the brand, the use of granular (by-version) data, qualitatively reviewing and comparing-contrasting messages and executions, and other similar techniques.

[the above paragraph is a bit too focused on 'coding' dummy variables - the only way one can have a true understanding of the dynamics of a market is to have a synergistic approach to marketing analytics: one that encompases quantitative, pseudo-quantitative, and qualitative techniques.  sometimes, equations cannot answer the question "why?" - and therein may lie the need for further analyses using non-quant or hybrid-quant analytic strategies.]

6. Marketing econometrics analytics should result to action points.

Econometrics should result to potential action and/or decision points for the clients. These may include (1) making decisions on how much to spend on specific brands, (2) optimizing revenues and other campaign-effects against constrained, finite budgets, (3) optimizing investment lay-down to optimize revenues and returns, (4) creating different scenarios with varying effects, and (5) deciding the optimal course of action given current knowledge.

[if it doesn't lead to action, then what was the whole project for?]

7. Decisions about the future carry risks; these should be taken into consideration.

Econometrics is an application of statistics and probability to test hypothesis on the economics of marketing. As such, there are risks involved. The most basic of econometric models –

Y_t = β₀ + β_iX_i + ε_i

suggest that there are ‘errors’ that cannot fully be accounted for. These ‘errors’ – or innovations – are the source of risks that no matter how robust a certain model is, there will always be unaccounted for variation – and there will always be errors and risks.

Risks should therefore be considered in making decisions based on marketing econometrics – and any other technique.

[nobody has seen the future.  even the most prolific of quant-traders from the field of finance failed to project the breadth and depth of the financial crisis.  and with all the changing variables and scenarios that are all around us, it is foolish to say that "because the equation has a probability of error of less than 1%, therefore it is correct..."  (which is in itself a contradiction).

risks abound - and therefore must be accounted for, and not necessarily measured or quantified - but prepared for.]

8. Risk management principles need to be integrated into marketing analytics econometrics.

Risk management aims at quantifying and mitigating the impact of random, stochastic variation (and shocks) in the future. Marketing econometrics provides an opportunity for marketing executives to ‘look into possible futures’ – and a basis to mitigate the effects of negative events.

In the practice of marketing econometrics, this entails not just a review of the distribution of “errors”/“residuals” and their corresponding impact on the brands’ business. This also entails the understanding of each variable/factor’s contributions and effects on the company’s revenues, and the risks they pose on the company’s revenues.

[equations are usually back-tested (tested on previous data) to ensure their accuracy.  but that is not enough.  the past is not always a good predictor of the future - and errors, innovations, shifts happen.  and even with the most robust of models cannot capture all these possibilities.  risk management principles - "management" used very loosely - need to be put in place.  risk cognizance, risk awareness, and preparedness need to be put in place.

and that will require far more than just models, statistical tests, inferential statistics, back-testing, and everything numeric.]

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...

causality is one complex business…

… that this blog post is NOT going to do the topic justice. but i will trudge on nonetheless.

just because two things are moving together doesn’t mean they are causing each other

 

 

you can’t conclude that something causes something unless you have (1) thoroughly reviewed previously written research, (2) looked at alternative reasons why the two may be – and may not be causal, and (3) checked with common-sense.  a high “r-coefficient” or a high “r-squared” doesn’t mean anything.

 

 

if you don’t have an explanation why would A cause B – and why not something else (say, another variable C), and why not the other way around – i.e., B causes A – chances are the “causal relationship” is spurious.

 

***

once again, there’d be remarks that this is too academic. 

but it isn’t.

it’s simple, common-sense.

foresight is never 20/20 - but you gotta move on

I have always been fascinated with statistical models that try to project the future.  Regression-based modeling, econometrics, scenario-planning, and simulations (including techniques from the finance field) are some of the techniques that I found to be helpful in coming up with projections. 

In my previous life, we even came up with a "vision-statement" that encapsulated this view: "Empowering Foresight".

However, at the back of my mind is "foresight is never 20/20" and that "past events are never a guarantee of future events".

A constant reminder that in spite of all the due diligence, the rigor, the studies, the equations, and the scenarios that have been done, there is no guarantee that things are going to be as predicted.

Foresight is never 20/20 - hindsight is.

Seth Godin writes about this in his entry, It's (Always) Too Soon To Know For Sure.  He notes:

If you're racing, you better figure out what to do about the times that you don't know for sure...because more and more of your inputs are going to be tenuous, speculative and possibly wrong. Day traders have always understood this--all they do is trade on uncertainty. But you, too, if you're racing, are going to have to make decisions on less than perfect information.

Indeed, perfect information is hard to come by in spite of the deluge of data and the now-very-cheap methods of gathering information. 

But these days, one has to make decisions on the fly.

So what does one do?

My view is simple:

Do the due diligence, go through the rigor with the current information that you have, make a decision - but don't cast it in stone. 

Be flexible enough to change and respond to changes. 

And adapt.

That sounds like a cliche because it is.

But it is not always practiced.

It is true the doing everything that you can is not going to be a guarantee that things are going to be right. But it certainly doesn't mean you shouldn't even bother doing what you can to control the things that you can.

It will not be perfect - it can NEVER be perfect.

But that should never be a reason to skip the due diligence and the rigor:  one has to uncover as many potential opportunities and threats as possible, and come up with scenarios to respond to them. 

One will never encapsulate everything whilst planning, say, a campaign or a marketing program - that's why it's called the future.

The only recourse that we have is to plan to adapt and to learn and to respond.