There is a clear challenge occurring within the current media mix, which typically comes down to the large number of touch points a typical customer will interact with before making their final purchase.
Current standard attribution models are able to show the key points within the customer journey; for example, the first interaction which started the customer on the path to conversion or the last interaction which secured the purchase.
This information is of course important, but by no means enough. Advertisers should have a mind-set that no single channel is entirely responsible for the purchase, and that every channel contributes to the customer’s final purchase decision in some way.
Applying this mind-set ensures the data available can be fully utilised, meaning data naturally becomes more relevant, comprehensive and easily actionable.
Introducing data driven attribution
The data driven attribution model is an additional model now available. There is a key difference with the data driven attribution model and standard models. Rather than using static proportions to assign value to a touch point, an algorithm is used to determine the true value of each touch point. The model is constantly learning and is updated on a weekly basis, to ensure it is making the best representation of the data available.
Example
The current setup of a static attribution model could be assigning a high value to social campaigns, as it is assumed that they impact in lots of conversions. If in reality social campaigns are only playing a very small part in the overall journey, and it is actually PPC which is the largest influence, the data driven attribution algorithm will determine this and lower the value accordingly.
All Response Media Viewpoint
Attribution is all about being able to analyse past performance data in order to then influence key campaign decisions going forward.
Pros
✓ Justification of media spend: Not only will the model allow the identification of the best touch points, it will also identify poor performing areas to ensure budgets can be aligned correctly.
✓ Enhanced media mix: Since the model shows the true value of each touch point, it allows the overall media mix to be enhanced. For example, if a gap is identified within the customer journey, it then opens the opportunity to introduce a new form of media to plug this gap.
✓ Understand the multi-channel funnel: This understanding is crucial and ultimately allows campaigns to be planned better and more robustly.
✓ SMART DECISION MAKING: The underlying thing which data driven attribution opens up, is the use of actual live campaign data to them make smarter decisions moving forward.
Cons
χ Flexibility: The main issue with the use of data driven attribution models is flexibility, and mainly the lack of it.
- If a data driven attribution model experiences a change in tags or groupings, then the model can no longer be applied and needs to re-learn before being utilised again.
- Look back windows should be at the default of 30 days; this could be fine for some industries but not for others; e.g. long haul travel can have a long click to purchase time.
- Although the model being updated can be seen as a positive, it could also be looked at negatively as this could simply be not often enough or too often for others.
- Data Driven attribution data is only available from the date the model is configured, not retroactively.
χ Inaccuracies – A data-driven model will be less accurate if there is a time gap between training and attribution. The larger the gap, the larger the impact on accuracy.
Data-driven attribution is a progression for digital but still is not perfect. At ARM, we have trialled a number of different data driven attribution technologies to get to this view point. It is not just digital channels that we need to take in a silo, but the full media mix. With ARMalytics, our advanced tech and measurement suite, we see the clear impact of TV/radio on web visits and the data driven models are not taking this into account yet.
