In the spring of 1902 the Hanoi governor was worried: a bubonic plague outbreak threatened to spiral out of control. The governor wanted to exterminate the city's rats but he lacked the necessary staff. After puzzling over his conundrum he had a brainwave: a bounty for every rat's tail handed in.
At first, his idea worked wonderfully - tails poured in. Hundreds each day in March, thousands in May, peaking on 12 June at a staggering 20,114.
But there was a problem. Despite the growing tail collection, there was no decline in plague victims. Nor was the rat population shrinking. In fact, if anything, there were more rats - but they were tail-less.
The bounty had unleashed the entrepreneurial spirit of the populace. They were capturing rats, feeding rats, breeding rats and lopping the tails off rats. The one thing they weren't doing was killing rats. Rats were too valuable.
The target had encouraged the specific behaviour that was rewarded - handing in rats' tails - but it had a counter-productive effect on the underlying intention - a reduction in rats.
From a safe distance, this tale of unintended consequences seems farcical. But similar examples have been recorded with alarming regularity. It's so common that the German economist, Horst Siebert, coined the term 'the cobra effect', so called as the first recorded example was an attempt to cull snakes in Delhi that backfired.
Siebert discovered that whenever there is a sliver of difference between a metric and the objective then unintended consequences are unleashed.
Marketing is a victim of the cobra effect
Advertisers, just like the governor, are trying to resolve complex problems with simple targets. Of course, the details are different - brands are grappling with unsatisfactory returns on investment rather than vermin infestations. But the consequences are equally perverse.
In order to boost campaign performance, digital advertisers set a range of targets. However, these metrics are almost invariably short term - visits, views or sales. That's not speculation. Peter Field's recent analysis of the IPA Effectiveness Databank found that the proportion of entries with a short-term goal had risen from 7% in 2006 to 33% in 2014.
Short-term metrics are popular as they're straightforward to collect and interpret. This ease appeals to marketers as they must manage more data than ever. Since time-pressured marketers crave shortcuts, they put their scepticism on hold and will themselves to believe that messy, complex reality can be captured with a simple metric.
Once measurements are added to a campaign they take on a life of their own. It feels remiss, unprofessional even, to ignore data, so plans are optimised to the metric being captured. To paraphrase Marshall McLuhan, we shape our metrics, then they shape us.
Once short-term metrics have been collected it's easy to generate improvements: dropping the worst performers quickly boosts results. In a world of uncertainty, that progress is addictive. In Martin Weigel's memorable phrase, short-termism is 'marketing crack'.
But we're optimising to a sub-optimal position: what works best in the short term isn't ideal in the long term. Les Binet and Peter Field showed in The Long and Short of It, their earlier analysis of the IPA Effectiveness entries, that direct approaches outperform brand approaches by 50 per cent over one to two years. But brand approaches are six times more effective when judged on a three year-plus timescale.
The short-term focus creates an illusion of success. But illusions tend to end abruptly. Marketers are like the man in the old story who jumps from a skyscraper. As he's falling, as he passes each floor, he reassures himself: 'So far, so good,' he says, 'so far, so good.'
Unless marketers change, they're in for an unpleasant landing.
So what can we do?
First, insist on balanced metrics. If metrics are flawed because they give a partial view of the problem, then capturing a single type of data will be misleading. Instead, campaigns need to track a variety of genuinely different types of measurement.
Every digital campaign needs to complement short-term tracking with analysis of how longer-term brand metrics have been affected. The costs for exposed versus control brand tracking have dropped far enough now, that this aim is feasible.
Second, we need to recognise, as the sociologist William Bruce Cameron stated, that not everything that counts can be counted, and not everything that can be counted counts.
No metric, or range of metrics, can fully capture a complex problem. So we should avoid giving undue deference to metrics. They need to be given the importance they deserve - but no more. Even in a world of limitless data there is a crucial role for human intuition and judgement.
Unless we recognise this, we'll continue to deliver rats' tails, not rats.