There is a lot of buzz around Data Driven Attribution, known variously as, Marketing Attribution, Multi-Touch Attribution (MTA), Multi-Screen, Multi-Channel Attribution etc.
Marketing Attribution is a subject close to every marketer’s heart and remains a tough problem to solve. It’s all about measurability and accountability. It has been an advertiser’s Holy Grail since the late 19th century, when US department store merchant John Wanamaker famously quipped: ‘Half of the money I spend on advertising is wasted; the problem is I don’t know which half.’
Defining Marketing Attribution:
Attribution is the process of understanding and assigning credit to marketing channels that eventually leads to a customer purchase. The consumer journey is more complex today than ever before. With the proliferation of digital platforms and online and offline channels, the consumer’s’ path to purchase is no longer linear – it transcends channels and devices. Digital marketing spends have already surpassed all the other media formats. A snapshot of the ad-spent composition in the US for the period 2008-2016 is shown below:
(Source of data: https://www.emarketer.com)
Attributing right amount of credit to the channels that led to the final consumer purchase remains a daunting task and every marketer wants to know where to put her/his advertisement money.
The difficulty in solving the attribution problem was well brought out by Avinash Kaushik, who is a noted authority in marketing attribution, in his popular blog post a few years ago.
‘Now it is true that with sufficient analytical skills, time, patience and God’s direct blessing to you it might be possible to do complete multi-channel attribution analysis where the multichannel includes multiple online ad channels, behaviour of the person across devices and the impact online and offline. Sadly, that is incredible hard do as a whole. And when I say incredible hard, I mean almost impossible. And when I say almost impossible, I mean only attempt that after you know you’ve fixed all other problems with your advertising, online and offline existence and people. All three’.
He therefore recommends heuristics driven models such as ‘Last Click’, ‘First Click’, ‘Linear Interaction’, and ‘Time Decay’ etc. to determine attribution. (https://www.kaushik.net/avinash/multi-channel-attribution-definitions-models/). This of course stated in 2012, and since then a lot has changed and much more data has flown down the digital highway.
This may be a good time to take you through the heuristics that are widely used in digital attribution. Green shows the extent of attribution.
Last Click is both the most commonly used model of all the heuristics and also has been criticized for being one of the most inaccurate. The Last Click model assigns 100 per cent of revenue generated to the last customer touch point before a purchase. The gaps in this model could cause advertisers to invest their funds in wrong channels, leading to an adverse impact on the revenue and marketing inefficiencies.
It is understandable why many may want to classify the research for an attribution solution as a classic case of what John Tukey had said many years ago, ‘The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data’.
A Big Market for Data Driven Attribution:
While the above arguments hold good, but in this data-driven world there is no escaping hunt for a marketing attribution solution. Recently I read a fun book on Facebook’s approach of monetizing advertisements by Antonio Garcia Martinez’s – ‘Chaos Monkeys: Obscene Fortune and Random Failure in Silicon Valley’.
He appears skeptical about the workability of attribution but hints that just because there is no solution to this attribution problem does not mean money can’t be made by packaging a suitable solution around it. Very True. So long as enormous amount of money is being spent on digital media, there will be a demand for knowing what worked and what did not, and there will be people offering the latest algorithms claiming to solve the attribution problem.
Therefore, there is a big market for data-driven attribution solutions.
The data-driven attribution models may be weak today, but will get better tomorrow and may be perfect day after tomorrow. Behavioural and Cognitive Science research in the last decade informs that we do not really know why we act and behave the way we do. The consumer may not know what triggered the purchase and why did s/he buy? But just because we are not self-aware, does not mean there is no reason behind it.
As more transactions go digital, and with enough data, attribution will work. Big Data can tease it out.
In my subsequent posts, I will provide further details of popular data-driven models such K-Order Markov, Logistic Regression, Random Forests, Naïve Bayesian, Shapley Value Method etc.