Don’t Shoot the Messenger
Times are hard. The recession is biting and news is dominated by gloomy headlines forecasting the end of the economic system as we know it. If we are to believe certain less reputable sources we will all be living in a post-apocalyptic dystopia in a matter months. Scary. Economics and politics act as a looking-glass into the psyche of society – these headlines are shifting newspapers so we must take notice and treat this news with trepidation.
The economy provides many lessons that we can apply to our systems of analytics and insight. Most of our current economic theory was developed and fine-tuned during the ‘good times’. Boom, bonuses, and free-trade were the buzz words of success. Our industry also revelled in the pre-2007 bounty; it’s in this environment that the holy grail of attribution modelling was found. Attribution is the systematic examination of the value of specific media channels in the user journey; with a view to gaining quality insight into user behaviour, which can then in turn optimise the marketing mix and maximise returns.
Attribution is Dead
Thursday 6th May 2010 may seem like a distant memory but it does in fact hold massive significance. On this day the UK went to the polls and ultimately removed Gordon Brown from office. The leader who years earlier as chancellor had heralded the end of boom and bust. His preference was for a more hands off, ‘automated’ approach to managing the economy and a reliance on a complex system that would regulate and maintain itself. On that same pleasant day in New York, the financial markets experienced the largest daily point swing in Dow Jones history; now known by the suitably fun moniker of ‘The Flash Crash of 2010’. In little over half an hour, $200million was wiped off the value of Western civilisation. The cause was the over-reliance on a form of algorithmic trading.
Algorithmic trading or by its sexier name, Quant trading, relies on statistical modelling of risk in financial products to dictate trading decision. Sound familiar? The raw power of data at work on trillions of dollars’ worth of assets is astounding. Data is being hailed as the king and future of our industry, yet when it is used in a short-sighted manner without context to gain a ‘quick win’ the model can be dangerously incomplete.
The Quant system, designed to make trading more secure and profitable, backfired with spectacular results. The small trigger (Greek debt … we can’t shake it) that behaved differently to the trend caused a torrent of data that led to a tumbling cascade in the Dow Jones index. Attribution modelling also relies on a few too many assumptions and by trying to internalise and explain inherently volatile external factors it leaves itself susceptible to failure. I would like to point out that Attribution modelling is unlikely to be the cause of the next global banking crisis, but I do think that placing too much trust in a particular system can leave us painfully vulnerable.
‘Any intelligent fool can make things bigger and more complex…
It takes a touch of genius … and a lot of courage to move in the opposite direction’ – Einstein.
Attribution relies heavily on consistent, uniform, and de-duped data and doesn’t even begin to address the elephant-in-the-room; people are irrational. How is it possible to attribute value to channels that users do not ‘interact’ with in any consistent way? Attribution modelling must address these flaws before it can be truly complete as a strategic solution. Granular data in itself is no justification to its own collection. How can the data be relevant if we can’t see past the haze of a model to the actual behaviour of a user?
To this end, I suggest that Audience be crowned as the new buzz word of 2012. It seems far more logical to attribute value using a system that fundamentally builds upon the user as its foundation. This will help answer the challenging question, “How do I serve individuals the most relevant content, at all stages of the user journey, across all channels?” By using methods such as audience and contextual targeting we can better focus budget in attracting users that offer us, as marketers, a better ROI. Attribution modelling is trying to achieve the same goal – but is obscured by its own statistical and technical complexities.
Maybe my ideas of Audience’s ascension are heretical to the ears of analysts – but Audience does seem a friendlier proposal to pitch to clients. Long Live Audience.