The NMPi team began by implementing our granular Google Shopping campaign structure; this improved our understanding of which search terms lead to sales. The campaign structure is composed of a tiered framework, whereby search queries are segmented based on their likelihood of converting. In doing so, we were able to push our impression share up for terms that were performing well and pull back in areas where spend was being wasted.
The campaigns were then built out to bid at a product-specific level. This meant that every single product in Dune’s feed had its own unique bid, which allowed us to ensure that the right products were being pushed at the right time – rather than the obscure products often chosen by Google’s algorithm.
The final piece of the puzzle was our machine learning technology, which was applied to both of the above strategies. Based on insights from the previous 30 days, the platform analysed both the search queries and products that were performing well and dynamically restructured our campaigns each day to ensure they were as efficient as possible at achieving Dune’s goals.