2020 Gold Best Use of Marketing Automation
16 Dec 2020
Campaign Name: Hyper-optimum Bidding Strategy: A Global First
Doing the client’s bidding
Hyperoptic wished to boost its share of the UK broadband market with a world-first use of algorithms.
Hyperoptic only has 2% of total UK broadband availability. Securing an order through display activity is a real challenge as geo-fencing isn't exact enough to reach users in a known Hyperoptic-served home.
Merkle’s brief was to use programmatic display to achieve multiple KPIs: boost broadband orders, leads, and site visits and engagements for education.
The strategy focused on reducing waste and improving order and micro-conversion efficiencies. Micro-conversions include leads of future customers who can’t currently access Hyperoptic; site visits; site engagements.
Using Google’s automated bidding strategies, Merkle had already made progress on manual bidding. However, these strategies only allowed optimisation towards orders.
The bidding strategy was advanced by writing custom algorithms in python using Google betas. The first used DV360 conversion data. The second was a global first and integrated with GA360 goals. These algorithms allocated weightings to all conversions (e.g. page lands, form completions, time on site, orders) depending on their value.
Building custom algorithms meant they obtained increased data resulting in more efficient learnings and optimisations. Additionally, the algorithm can assess data for impressions that land on site, search for postcode and leave, for example. Google’s own bid strategies would ignore these impressions while the algorithm learned, whereas the new custom algorithms used every impression.
For both the algorithms Merkle created an experiment in DV360 that allowed a split of the target audience evenly between the different bid strategies. Both algorithms were tested at different stages of the funnel; in both prospecting and remarketing, to ensure detailed results for audience groups were obtained.
Overall, the use of bid strategies has been a success. Primary and secondary KPIs in remarketing have improved, and opportunities identified for improvements in prospecting to yield better results.
Custom DV360 algorithm —
Order CPA 17% lower
Leads CPA 19% lower
Page Visits CPA 12% lower
Initial Prospecting Results:
Order CPA 71% lower
Leads CPA 74% lower
Prospecting after reweighting DV360 algorithm and including GA360 algorithm:
Order CPA 78% lower using DV360 and GA360 algorithms compared to MinCPA.
A reweighted version of the DV360 algorithm has now been deployed alongside the GA360 integrated test. These algorithms have outperformed Google’s MinCPA bid strategy.
Merkle - Amelia Currie, Programmatic Lead â€“ Solutions Specialist - James Fort, Senior Associate Programmatic - Autumn Chard, Junior Associate Programmatic