Nowadays, advertising have more and more and advertising options, e.g. online and offline marketing, paid social, KOLs, event marketing, reputation marketing, knowledge marketing etc. Thanks to the fast developing new medial and digital technologies, companies have varieties of advertising options for better ROI.
The questions is, how can we measure the performance of each channel and source? While it still remains a big challenge to measure the correlation between online and offline advertising campaigns, fortunately multi-channel performance analysis is possible for online marketing, e.g. to measure how important content marketing is for a sale campaign.
The western world has been doing research on attribution models for thirty to forty years, and that provides a solid theoretical ground for analyzing multi touchpoints performance. For the moment, there are five main attribution model applied to online marketing field:
1) Last interaction/click attribution model
A model that gives credit to the last interaction or click. It ignores the impact of any other touchpoints in the whole customer journey and is the most unwise attribution model out of all.
2) Last non-direct click attribution model
A model that gives credit to the last interaction except direct. Obviously It ignores the efforts of direct traffic and undervalues the importance of brand campaign.
3) First Interaction/click attribution model
A model that gives all the credit to the first click. This is as unwise as the first attribution model, because as we observe, there are so many touchpoints happening before the conversion closes, sometimes even thousands.
4) Linear Attribution model
A model that gives even attribution to each channel. For 5 touchpoints are involved in the customer journey before the conversion, each channel get 20%. However, this model is a utopian model, as it assumes each channel plays the same important role in earning the final conversion.
5) Time Decay attribution model
A model that gives the most credit to the channel which is closest to the conversion. The percentage of the importance is based on algorithm. The good thing about this model is that we can customise the half-life of decay and insert our own predictive judgements and estimations into the attribution journey.
6) Position based attribution model
A model that gives 40% of the credit to the first and the last click or interaction and the remaining 20% goes to the rest of the channels involved. It tries to leverage the downsides of the last click/attribution model and first click attribution model, while recognising the weighted importance of both.
New: Data-driven attribution model
Those models are all rule-based models and they are very likely to fall into the trap that inappropriate attributions are given to the wrong channels. In paid ads channel, Google has launched the data driven attribution model (DDA) in May 2016, and received wide acceptance by advertisers.
DDA is not based on pre-defined rules, which are rather inflexible and stable. Instead, it uses algorithmic attribution, meaning that DDA calculates the probability of a conversion. It identifies the most influential ad interactions at a particular point.
In DDA modelling, the conversion channels do not consider other channels but only paid ads.
Customise your model!
Beyond paid advertising, companies usually have the options to customize their preferences for touchpoints and defines their own attribution solutions. It also often happens that one channel might get less conversions, and other ones get more credits. This looks unfair at the first glimpse, however, all the channel teams should pay attention to the bigger picture and look out for the overall performance – which is an ultimate benefit for the company.
Assisted conversions
One suggestion to check how important one touchpoint or channel in terms of contribution to conversions, we can check the assisted conversions report in Google Analytics – a great report to get an idea of how each channels is helping others.
However, there is no right or wrong about any sort of the above-mentioned attribution models. Companies should choose and more importantly adjust their attribution model based on their current and long-term business goals.
The Digital Teahouse 