AI: The Future of Digital Media, Now
|CATEGORY||ROI CAMPAIGN AWARD|
|CAMPAIGN TITLE||AI: The Future of Digital Media, Now|
|BRAND||Aviva motor insurance|
|TEAM MEMBERS||Chris Davies, Business Director, Zenith
Cameron Davies, Data & Insights Executive, Zenith
Toby Feakins, Planning Manager, Zenith
Shaazie Guthrie, Director, Newcast
Data Sciences Team @ Zenith
The automated implementation of modelled output into DoubleClick buy platforms has been a world first and delivered impressive results:
For a mature business area in a highly competitive sector (car insurance) this is a very significant result.
With Aviva investing millions of pounds in digital channels each year in the UK, we realised that we needed to produce a more advanced, always-on approach to drive continuous improvement for Aviva and at a more detailed level than ever been done before (at keyword or placement level).
Market-leading vendors fell short in some aspects of delivering multi-touch attribution models. Data validation was not rigorous and some models lacked sophistication and the ability to implement the results in a consistent and reliable way, so we knew we had to create our own solution for Aviva.
The key challenge was to improve ROI for a major client’s digital investment in a relatively mature sector on activity already heralded as delivering best practice – and for a client who was up for pitch.
We conducted an initial POC exercise to demonstrate capability before we moved into Phase 1 activation through 2016; disciplined project management and regular client communication ensured that the key stakeholders were engaged at every step of the journey. The agreement was to focus upon car insurance and activation within channels initially; with the majority of investment in search this was the priority for automation of the execution.
The Data Sciences team in the UK has developed and evolved our market-leading approach to machine-learned, algorithmic attribution modelling, which was perfect to meet the Aviva challenge: moving it away from ‘last-click’ and human determined decisions to data-driven and automated ones.
The ‘machine-learned’ approach means the model is engineered to continually consume data and the algorithm ‘learns’ (or updates) on an ongoing basis so that the output can be implemented as quickly as the data is fed in across different digital touchpoints.
We have worked closely with Doubleclick (Google), taking the lead in auditing the quality of their data feeds and crucially we have connected our modelled output automatically into DoubleClick platforms for both online display (through DoubleClick Bid Manager) and search (into DoubleClick Search). This is a world first for an agency/advertiser to inject modelled output into DoubleClick platforms, and influence Google’s own algorithms.
Through a very robust test set-up we were able to record a 6% improvement in CPQ (Cost per Quote) for search activity impacted by attribution output (the Test) versus a control group, and for display we’ve seen CPQ drop by 10%.
We have determined that these efficiency gains can generate Aviva over £700,000 more in sales revenue for the same investment level.
It was a major contributing factor to the Digital & Data Transformation piece that helped retain the Aviva business, whilst also showcasing the first demonstration of advanced machine-learning that is a key pillar of the new ROI+ positioning, which has inspired other client opportunities too.
“We’re delighted that as part of our commitment to digital and media transformation at Aviva we are breaking new ground with this pilot automation of our search and display. The benefits of attribution modelling will be realised in terms of ROI improvement as well as through operational efficiency.”
James Turner, Head of Marketing (Trading), Aviva