pixel facebook


Mazeberry: Attribution model, what use for companies?

Published on by Fabrice Donnadieu - updated on

The attribution model used by retailers is of paramount importance. This conference explores the possibilities offered by this notion.


Attribution is a keyword that is heard everywhere, invoked by many providers and about which brands have many questions. The idea of ​​this conference is this notion and what brands can do on a daily basis.

What is Mazeberry

The group Ysance Mazeberry was born in 2005 in the world of data, Big data and artificial intelligence. It has 160 people who support 250 clients in the awarding market. The company works a lot in the field of physical and digital retail, its area of ​​expertise being the omni-channel attribution model. It is one of the most mature companies in terms of digital attribution.

While 90% of retailers' turnover is in point of sale, what is the impact of the digital budget on what happens in stores? The role of Mazberry is to provide answers to its customers.

What is the point of attribution?

Whatever the business (retail, banking, insurance online, etc.), there is competition in all sectors. Either because there are giants in the industry or because customers are hyperconnected and break silos, making them difficult to capture.

It is therefore important to react through media investment, but acquisition costs tend to increase, with a cost per share (CPA) or a cost per click up. Customer insight makes it possible to better target, as well as the product knowledge and attribution model that allows the right content or channel to be associated with the right customer profiles. A good understanding of the courses allows to apply to the course of the prospects what already works on the course of the customers.

However, various complications make this understanding difficult, such as data (cross device, omni commerce), the tool (more and more tools are piling up, but do not communicate with each other), the lack of time, expertise or resource.

5 tips to optimize the attribution model

When a conversion takes place, different points of contact are captured and it is important to spread the turnover (turnover) on these different touch points.

The attribution is imperfect

Retailers should not put too much hope behind the award. This is not a useless job, but there is no perfect attribution model. It is possible to make assumptions, but it remains difficult to identify the good points of contact. There are three major schools of attribution:

  • Data driven: definition of a tree with different points of contact to see what would have happened by removing one of them.
  • Algorithmic: the theory of the game For example, if three people play and earn 100 euros, they will want to distribute the sum between them. But if one of them has contributed less to the victory than the other two, then they will try to distribute in proportion to the contribution of each.
  • Position based: a fair division of the contribution of the different points of contact.

But each attribution model does not identify the waste, that is to say, which costs less, but which is really useless. It is also possible to save on what is cheap, and sometimes cheaper to clean up the mess than to optimize financial bleeds.

Banking on the coalition

Last year, in order to boost e-commerce conversions, Mazeberry decided to set up an operation dedicated to "abandoners".

This operation was intended to send e-mails to people who have abandoned their shopping cart. By setting up the company's multitouch attribution model, the company was able to see that retargeting email is present in 4% of the business, 1,100 conversions were attributed according to the Mazeberry model, an equivalent turnover was realized, as well as an acquisition cost of 8.56.

The email retargeting is important and has weight in the business of the company. Among the best coalitions analyzed, the "best friend" of retargeting is the newsletter. But the latter cannibalize half of the retargeting audience and in 82% of cases in which the levers work together, the newsletter precedes the email retargeting. We must not stop the email retargeting, since it brings an interesting acquisition cost.

The interest of frequent scenarios

The analysis of the levers must be done in relation to their presence in the conversion. A vertical table is created to look horizontally at customer experiences and detect scenarios within the Mazeberry platform.

By creating tables with powerful scenarios in terms of business, a finding gradually emerges, namely that the chance of seeing a particular scenario in conversions is low. It is important to ask frequent scenarios and stop retargeting to save money.

The impact of deduplication and the notion of contribution

Thanks to its affiliate platform, Mazeberry has noticed important double commissions. It was therefore necessary to deduplicate the last affiliates. For example, in the case of a conversion in which the customer was in order, by display, search, then affiliation, both affiliations were previously paid. [SM1]

But after the implementation of deduplication, the budget went from 10,000 to 5,000 euros. It was a good decision to save money. But the company realized that in this scheme, the first affiliation was an affinity blog and the second was the cash back part. The deduplication system cut the remuneration of the blog affiliation in favor of the cash backer because it was awarded by last click.

This is why it is important to set up the "smart" affiliation, because if it is necessary to deduplicate, the last click model remains dangerous. In this case, the allocation lever is not the last click in fact costs while it may contribute more to the conversion. The notion of contribution makes it possible to prove that an affiliate has an interest in the value chain.

Master the data

With every interaction with the customer, you have to give yourself the opportunity to leave a trace that you can reuse tomorrow. CRMs treat loyal customers, while the acquisition looks for new customers.

Attribution model budgets are effective when you control data. We now have tools that capture the data at the level of digital devices, so as to create the "customer journey" where we find both the conversion and the various elements in post click. This allows for a complete value chain.

For retailers, it is possible to create an omnichannel identifier and reconcile CRM and digital data in order to work on the notion of a complete attribution model. The goal is to achieve an offline and online value chain, regardless of the sales support (e-commerce, apps, mobile and physical store).

Speaker:Romain BAERT

Similar article 

Discover Altima and Ubisoft's workshop on data tech and the customer experience

Back to the article list

01 / 03 oct. 2024 Biarritz

#1to1ExpClient #1to1Biarritz