Why measuring performance in E-Retail Media is a must and how to move beyond ROI.
The e-retail media market has been growing almost continuously for years.
Over the last few years, E-Retail Media has experienced a growth that continues to attract a multitude of players and, with them, an increasing number of solutions, as well as a large portfolio of campaigns. This market growth is reflected in investments, since according to the Observatoire de la Pub (and included in the mapping carried out by the IAB), investments in E-Retail Media by brands have increased by +42% vs. 2020*. The objective here is to make a brand stand out from the crowd via increasingly innovative communication and visibility devices and, above all, seamlessly integrated within the user experience of a given website.
Like the rest of the media and branded activities, E-Retail Media is suffering the full force of long negotiations and general uncertainty in FMCG. Nevertheless, it remains one of the few media sectors that continues to innovate and in a context where digital transforms everything in its path, it is unavoidable.
The marketing mix must adapt to cope with the multiplication of solutions and products proposed to brands.
With the explosion of digital solutions, the traditional marketing mix known until now has to evolve in order to adapt to new ways of consuming and thus respond to new shoppers’ expectations. The new shopper experience needs to be as fluid as possible, on whichever device or channel used by the shopper, who wants to have the same experience, no matter the way he/she accesses it.
Having said that, we quickly realise that the in-store experience is definitely different from the digital one. How many times have we ordered online a product and were surprised to discover that the product received was different from the one we had in mind? Neither the colour, the size nor the texture we imagined, etc. It is clear that the in-store experience addresses a different need than the digital one.
So, we have to admit that the marketing mix we have been taught until now has to radically adapt and transform itself into an “e-marketing mix” in order to address the evolution and needs of the market, which are evolving at a frantic pace.
With this adaptation of digital offer, the number of products and services launched and tested is numerous. Digital opens up unprecedented possibilities, making it possible to test, stop, start again, and above all to innovate, in short, to develop the offer and the experiences available. The field of possibilities is vast, and above all, the speed of implementation makes it possible to be extremely flexible and reactive.
In this context, the marketing mix needs to evolve in order to adapt to the specificities of the digital environment, and gradually move towards an “e-marketing mix”. The 4Ps (Product; Price; Place; Promotion) must evolve because they alone cannot address the entire digital strategy. Some even go as far as talking about an “E” model, including the notion of Experience or Engagement, which are all relevant elements in the age of digitalisation.
On the other hand, this digital offer leads to complexity in measuring performance and comparing players.
The richness and speed of development of these offers also imply different performance measurements. Each service provider thus measures performance according to the strengths and advantages of the solution implemented. This diversity is very rich and allows brands to communicate at different levels and via numerous levers. The downside, if we can call it that way, is that it muddies the waters when it comes to comparing solutions. After all, what better way to make decisions and justify the level of investments than by measuring ROI, i.e. the return on an investment made?
Behind a KPI which seems unique, there is in fact a complexity directly linked to the diversity of calculation methods that can be associated with it. Indeed, media agencies will rely more on the ROAS (Return On Ad Spend) method base, while other service providers will use the difference in the level of turnover generated during a campaign (vs. another period of comparison) in relation to the total investment of the campaign… The calculation methods are diverse and different because the access to a wider range of data has now been made possible. It is now possible to have access to pure purchase data, for example, and not only to the shopping basket. This enables a much more granular and powerful operational use, but also finer and quicker analysis of what works and what needs to be improved.
Beyond financial performance, the importance of the Data quality at the service of Machine Learning.
The only common point between all these measures lies in the quality of the Data used for these calculations. Although we now have access to a large amount of data, this does not mean that all data is equal, whether in terms of granularity, quality, reliability, history etc. Moreover, it should be noted that processing such a large amount of data, analysing it and/or using it operationally represents a challenge in itself, which not all players are able to address.
Access to shoppers’ purchase data (receipts) enables us to understand their behaviour, preferences and habits more precisely, over long periods on a large number of purchasing acts. It is therefore easy to see that the quality of the available data is critical, as it is on this basis that learning models are developed and, above all, can be improved and predictive models pushed (via Machine Learning). This Machine Learning is all the more important as it ultimately allows the development of increasingly effective and, above all, increasingly relevant products to be pushed to a site’s shoppers. With this vision, we go far beyond “pure” investment, i.e. as we understand it from a purely financial point of view.
Companies which use First Party Data are certainly shopper centric, but in order to be so, they must first and foremost be Data Centric. As we have seen, data and its quality is critical in the development of products and in particular the statistical models used to run all operations. Machine Learning, and above all, to make its learning and development reliable, depends solely on the quality of the data used, its reliability, its regular updating and its granularity.
In addition to the level of turnover generated during a campaign, the purchase history available is also important, particularly for identifying the real impact of campaigns via the notion of recruitment. It is thus possible to draw up a typology of shoppers to address as closely as possible what can be defined as the real impact of a campaign. This typology divides shoppers into three categories: new, reactivated and loyal. The new shoppers correspond to shoppers who did not buy the references 6 months before the campaign (this period can be updated and adapted according to the products animated), and who started to do so during the activation. Reactivated shoppers have the same definition as new shoppers, with the difference that they did not buy the products 3 months before the campaign (and their duration can also be adapted according to the one defined for new shoppers). Finally, loyal shoppers are the ones who were already buying the products animated 3 months before the campaign and who naturally continued to do so during the campaign. To measure the impact of the campaign and isolate the windfall effect, the A/B testing is used to measure the incremental effect of the campaign on their purchases (a more detailed article has been devoted to the A/B testing method here). This typology, combined with the A/B testing method, is the one favoured by Lucky cart in order to determine the performance measurement of its campaigns.
This typology plays an important role in the way the ROI is calculated because we will only consider the “impact” part of the campaign, i.e. the sum of the turnover of the new, the reactivated shoppers and the incremental generated by the loyal ones. This makes it possible to go beyond the pure financial investment made for a given campaign, and to measure the real impact of the activation on the actual purchases of the shoppers who were exposed to it.
But this is only the numerator part of the ROI calculation. When we look at the denominator, we can see that the level of ROI is interdependent on the level of initial investment. The latter is not the same depending on the activation, and different positions can be taken on this point. Indeed, a media agency will consider the pure media investment, while other providers may propose a package which includes all the actual costs. One can also only consider the cost of the promotional investment, the cost per click, the level of exposure etc.
ROI, as an essential indicator, but needs to be analysed along with the real business impact generated.
In conclusion, ROI is an indispensable and unavoidable indicator, as it enables decisions to be made, particularly for steering a brand’s marketing investments. On the other hand, it also seems crucial, in parallel to this measurement, to focus on measuring the business impact of the actions that are carried out. These actions enable us to address customers and prospects more directly, in an efficient, profitable and relevant manner over the long term by pushing offers that are increasingly adapted to the end shoppers.
ROI analysis therefore makes it possible to analyse a brand’s competitiveness and attractiveness, but the key to success also lies in the ability to carry out coherent global marketing actions for the continuous improvement of this indicator and to ensure the good performance of future activations for the brand.
* Alliance Digitale – 2022