Limits To Influence – “Where The Bloody Hell Are You?” or “How Much Difference do you Really Think you can make?”
A while back there was a a bit of controversy about the “So Where The Bloody Hell Are You?” campaign from Tourism Australia. Wikipedia has, as usual, something to say about it.
There were suggestions that perhaps the campaign did not perform as well as hoped (surprise, surprise), but I doubt whether anyone really collected hard evidence on the matter.
I am not going to comment on those matters directly (although there is plenty of interest and harmless enjoyment to be had from the problems of optimum advertising allocation, intervention models and the like) nor on the research design which, reportedly, had a budget of $6 million (the advertising spend was reportedly $180 million) . Nor on the wisdom of targeting a pyschographically defined market segment, nor the ecological fallacy.
There is a more interesting problem. It’s the problem of what is reasonable to expect, of quantifying how much we may be able to influence (any) system and its outcomes.
How likely is it that we can influence…
the number of tourists to Australia by any substantive amount? How likely is it that we can achieve stated market share goals?
What are “reasonable expectations”?; how much influenceable slack is there in any market?; And how can we avoid the later rejoinder “Well, What the Bloody Hell Did You Expect?”
Let’s stick with international tourism as our exemplar. There are some specifics of the pertinent dataset that are unique to the tourism context, but similar principles apply in other fields.
Now, is it worth spending megabucks on research and advertising? Or is the problem simply intractable? Are we already performing at a near optimal level? Are behaviors influenceable? Is there a decent return on investment (ROI) possible, or likely?
Now, it is certainly possible that Australia is not performing at optimum levels in the international tourism market. Take this quote from espresso.over-blog.com
And for a sidenote, Austalia’s tourism figure is pathetic, 5.5 million tourists per year, a mere trinkle of the global tourism, understandably owing to its isolated position and that non-desirable long hauls just to get to our shore. There’s room improvement there: check this out, wikipedia says, Andorra, that tiny sovereignty in the Pyrenées, sandwiched between France and Spain, entertains anually a staggering nine million tourists. And that’s just heaps, given the local population of merely 69 000 heads
OK, we can chuckle over the grammar and the malapropisms and the flawed reasoning .. but such reasoning/ wishful thinking is bandied around and, at its core, there is a sensible argument.
Or at least the beginnings of one. It goes as follows. Instead of simplistic extrapolations ( the population of Australia is 20 something million, divided by the 69000 population of Andorra gives us a factor of around 290, times 9 million = 2600 million potential tourists to Australia instead of the measly 5.5) , we do some modelling of international tourism flows (probably confining ourselves to long-haul travel).
The data will be a bit messy to collect, but at least it is mostly desk-research.
For every pair of countries (A and B) we collect
- a) the basic travel flow information (visitors per annum, visitor nights .. whatever we can get) : yes, this includes Australia as a source as well as a destination, and as many other pairs as we can get. We do not apply any censorship or exclusion at this stage .. eg excluding China because it is “different”.. we can do that in the analysis
- b) for each pair we collect some similarity measures – eg ethnic commonality, common heritage, trade agreements etc
- c) for each source (originating) country we include in the term a “propensity to travel” term (eg Germans travel more widely than French)
- d) distance and costs
Leave aside the obvious holes and ifs-and-buts .. we collect some data along the above lines, quite cheaply and quite well.
Now we proceed to fit a model .. just linear regression will do for the moment.
The model is
Visitation (to country I from country J) = some function of .. distance, cost, similarity, source characteristics ..
And, potentially, for each destination country we include in the model an “attractiveness” term (this is not data, this is model derived).
Well, what does this tell us?
Essentially the answers we are seeking. If 90% of visitor flows are determined by the “basics” of distance, cost, similarity .. if we don’t need to put in (many) destination-specific attractiveness terms, then it is a suggestion that we do not have much room to move .. i.e. that there is not a lot of “slack” in the system, we cannot influence it all that much via emotional appeals or awareness raising.
The interesting bit is the residuals
Still more interesting are the anomalies, the ”residuals”, the differences between what the model predicts and what the data says. If Australia has positive residuals .. i.e. the data points are higher than the model predictions, then this is an indication of a closeness to optimum (and again, a constraint on how much more we can achieve).
Residuals for other countries may also be very instructive and, given a large residual – a large discrepancy between what is happening in the real world and what the model says, we should go on a hunt for explanations.
Of course, from this point, being quite well informed about the real world and what the leverage points might be, we should start thinking about doing a choice modelling study or somesuch. Models on top of models. Or some informed research.
Or at the very least, form some reasonable expectations.
