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 –
Yt = β0 + βiXi + εi
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.]