How many people directly influence the result of a national election? In Britain, for example, the size of the electorate is roughly 30 million people. But the number of people who decide our political future at a general election is actually much less.
The two main parties in Britain each command a “natural” vote of about 30%. In other words, in almost any election, and regardless of any particular policies or the prevailing economic climate, 30% of the electorate will almost always vote for each of these parties. In a similar way, the natural vote of the remaining parties, taken together, is about 20%.
This leaves the 20% of voters – only six million people - whose vote is volatile across elections and who therefore could conceivably influence the overall outcome. But this isn’t the answer to our question either.
Of that electorally volatile 20%, only one fifth live in marginal constituencies, where their variability can overcome the natural vote in those areas. Revising our figure to account for this leaves 4% of the overall electorate. And of that 4%, a quarter of them won’t actually bother to vote on Election Day, leaving us with the answer to our question: that only 3% of the electorate, or 900,000 people - from a total electorate of 30 million - directly influence who will govern the country for the next five years.
That’s less than one in thirty of the electorate. In a country as varied, in a society as diverse, in a culture as opinionated, this is a striking phenomenon, but there it is.
Now consider the activities of two political parties during a general election campaign, where one of the parties knows about this compression and the other doesn’t. It’s easy to see that the first party would very likely conduct a highly focussed, efficient campaign, targetting the issues and desires of swing voters in marginal constituencies. The other party, lost in the scale of the task, would spread its resources too widely, spending a lot of money and time on lobbying voters whose opinions it couldn’t change.
Of course, the political parties in real life know all this, and act accordingly. They drench marginal constituencies in senior cabinet members during the campaign, reserving cheaply printed, rainwater-soluble print brochures for the rest of us.
Even more impressively, the election special televsion pundits can often magically predict the outcome of the entire election, with high accuracy, after the first two or three marginal seats have declared.
Electoral mathematics displays the fortunate property of being controlled by a small subset of the measured population. It’s also fortunate for our parties and pundits that they know which subset this is.
Can the same effect be brought to bear in IT? Could we identify this effect within our application infrastructure and derive analogous benefits of better targetted investment and prediction of issues?
There can be up to 1000 individual measurements obtainable from a typical, large n-tier application. That’s over half a million measurements to trend and analyse across the estate as a whole, in a large organisation. Sumerian’s Service Delivery Analytics routinely monitors all of these for our customers, detecting emerging change.
But with one of our most innovative clients we’re now applying IT Analytics, and in particular, clustering techniques, to identify that small percentage of measurements that most strongly indicate application health and emerging issues. We hope to build deep predictive models around these key attributes, calibrating them with the knowledge of prior history.
By identifying these “swing voters”, and studying them deeply, we hope to advance further, the state of the art in pro-active outage and degradation prevention. We’ll keep you posted as to how we get on.
