While the social network’s ad revenue grew 63% year over year in the second quarter, it only added 8 million new users. Interim CEO ...
Retailers unhappy with last-click attribution need to think about customer value.
Attribution remains a contentious issue within digital marketing. It provokes a great deal of debate about the best solutions for attribution modelling and how it should be implemented. Attribution is a fashionable buzz word and is something numerous retailers are looking at a solution for.
There are various models retailers are considering. The most common form of attribution used by the majority of advertisers is single-click attribution. This is traditionally the last click to attribute the sale to the last referring channel. This could also be looked at on a first-click basis to reward the channel that instigated the sale and will vary depending on the retailer.
More retailers are looking to adopt a multi-attribution model. Multi-attribution allows for more than one affiliate or channel to be rewarded for their role in the path to conversion. For example, if the consumer visited three affiliate sites before making the purchase, the commission could be split amongst the three. This could be weighted equally or a retailer could decide to pay a higher share to one of the touch points, depending on who they perceive to have the greatest influence over the transaction.
Multi-attribution, especially where the affiliate channel is concerned, is born out of the common misconception that there often are a significant number of touch points involved within any one transaction. Data that we have investigated for a number of retailers show that this is not the case within the affiliate channel.
Looking at a large retailer it is evident that the vast majority of transactions only have one affiliate referrer (92.66%). There are some transactions that have up to 10 affiliate interactions but this is minimal rather than the norm.
Another common misconception is that incentivised traffic typically overwrites other affiliates. Incentivised traffic includes cashback and coupon sites that offer their visitors an incentive to transact with a retailer. Again, looking at the data this is minimal. Not only is there a relatively low proportion of sales that has more than one affiliate interaction, but when there is more than one affiliate involved, it tends to be the same promotional type being overwritten. For example, it is likely to be a consumer comparing voucher code or cashback offerings across more than one site.
Looking at the data for the retailer mentioned above, voucher code sites overwrite other voucher code sites 42% of the time and loyalty/reward sites a further 26.5% of the time.
Extending this beyond just the affiliate channel to look at multichannel customer journeys it is again evident that incentivised sites are not just overwriting other publishers/channels and have high levels of single interaction sales. This goes a long way to prove their value in driving incremental sales—that is, the conversion would not have taken place had it not been for the incentivised site.
The following chart looks at the incentivised affiliates across a fashion retailer, and there are a number of transactions where they are the only touch point within the customer journey. There are also a number of sales these sites have been involved in but were not the final interaction—reiterating the point they are also influencing customers earlier on in the customer journey.
There are a number of issues that arise when discussing attribution modelling. Firstly, it is important an attribution strategy is considered across all digital channels. This in itself throws up complications as all channels are paid for in different ways, some just on outcome (affiliate channel) some regardless of outcomes (PPC, Display). Thus, to compare all channels in the same way ignores the various commercial models in place.
It will always be necessary to assess influence and contribution, even if a channel is not credited with a sale on a last interaction basis. Clicks are an arbitrary measurement: what happens beyond the click is equally if not more important.
Additionally, within the affiliate channel, attribution can have its own complexities. Attribution models that involve multi-attribution can be difficult to create and require manual implementation. It is also problematic to introduce this for some promotional types. For example cashback, loyalty and reward sites need to know a commission amount they will earn for a sale in order to pass this on to their members. If there is a multi-attribution model in place, it is impossible to be transparent with their members. Cashback sites pay the commission they earn back to their members. In a mult-attribution model where there is split commission, a cashback site will be unable to display the correct rates to their members as they will not know how much commission they will earn.
While advertisers and agencies have placed a lot of emphasis on developing attribution models to understand the customer journey up to the point of conversion, one significant piece missing from this is what happens after the conversion? How valuable are the customers being driven by each online channel? An advertiser’s own internal KPIs [key performance metrics] will determine the metrics that constitute value. For example this can be the generation of new customers or increased lifetime values.
Value attribution is a method that looks at this post-conversion data—the process of understanding what has happened beyond the purchase and the channels that have been instrumental in driving valuable customers. By looking at the value driven beyond the last click, it is also possible to look at additional/alternative payment models to reward publishers for the customers they are driving.
Additionally, advertisers could be looking at the publishers that are adding value earlier in the customer journey. For example, content or comparison sites could be involved in a significantly higher proportion of sales than they are credited for on a last-click basis.