Illuminating the Complex Path to Purchase
In this piece, Peter Falcone, director of analytics EMEA, Flashtalking, tells RetailTechNews that shopping is no longer as simple as walking into a local store and selecting an item. Today’s consumers are influenced by a multitude of digital channels, from search engines and comparison sites, to marketing emails and display advertising, with a recent PWC survey revealing 37% of shoppers take purchase inspiration from social networks.
When the multiple channels that contribute to conversions are combined with the numerous devices on which these channels can be experienced – including desktop, mobile, digital out-of-home, and TV – understanding an individual’s convoluted path to purchase becomes challenging. The return on investment (ROI) of each marketing touchpoint is tricky to quantify, making efficient allocation of marketing budgets problematic.
And the ever-increasing proportion of users who are blocking or deleting cookies is exacerbating this issue – resulting in lost impressions, clicks, and conversions.
Retailers need to rethink their approach to marketing attribution to make the most of their marketing budgets. So, how is this technology evolving, and what should retailers be doing to accurately measure the contribution of each touchpoint, while reducing their reliance on the cookie?
Attribution models become multi-touchpoint
Most retail marketers already realise standard single touchpoint attribution modelling – where 100% credit for a conversion is assigned to a single touchpoint – is ineffective in today’s digital world. These models, including last interaction, first interaction, and last indirect click – which assigns credit to the channel prior to conversion that indirectly brings the user to the retailer’s website – all ignore the fact that purchase decisions are influenced by several touchpoints.
Retailers are moving away from these methods towards more sophisticated multi-touchpointattribution models that identify the conversion path and assign credit to multiple interactions. From the linear model, which allocates an equal proportion of the value of a conversion to each touchpoint, to the position-based model that apportions more credit to the first and last touchpoints and less to the channels in between – multi-touchpoint attribution recognises the numerous channels that influence purchases.
But, while these more complex models are an improvement on standard attribution, they still don’t represent the actual purchase decision process, as they are based on static pre-defined distributions, not on the real impact of each interaction. Moreover, they do not reflect how diverse channels work together to drive business outcomes at each stage of the purchase journey. Finally, and most importantly, these models don’t take into account non-converting paths to purchase, which are where the majority of the media spend goes, and where greater learnings are needed to reduce wasted budgets.
Algorithmic attribution reflects the real path to purchase
To truly understand the efficiency of different marketing channels and effectively allocate budgets, retail marketers need to take measurement to the next level through algorithmic attribution models. Rather than being based on static, pre-defined rules, algorithmic attribution uses machine learning to provide a realistic, real-time evaluation of the role each individual touchpoint plays in driving a conversion. It works across multiple digital channels and devices and allows for interactions and synergies between touchpoints. Retailers can run multiple algorithmic attribution models concurrently, with a system in place to select the most accurate model, delivering a picture of marketing performance that advertisers can trust.
Overcoming cookie rejection with probabilistic IDs
To make the most of algorithmic attribution, retailers need to reduce their reliance on cookies, which are frequently blocked or deleted by users, making it difficult to associate media touchpoints with a particular purchase. One solution is to create probabilistic IDs for devices or browsers that can connect user interactions with media, even when cookies are not present. Probabilistic IDs are generated from a multitude of data signals that are non-personal and, therefore, comply with privacy and data regulations. By accounting for cookie rejection, and producing a more complete data set, retail marketers can significantly increase the effectiveness of their algorithmic attribution strategies.
With the shopper’s path to purchase only set to become more complex, accurate measurement of marketing channels should be a high priority for retailers. When combined with probabilistic IDs, rather than just cookies, algorithmic attribution illuminates the individual purchase journey and delivers clear, actionable, real-time insight that retailer marketers can use to effectively allocate budgets and drive the best possible ROI.