TECHNOLOGY

Wizaly’s technology choices

technologie wizaly

Wizaly’s attribution technology aims to measure precisely the contribution of each of your channels to your conversions.

Let’s assume a conversion path with N touchpoints of various natures :

  • clicks or impressions
  • offline (TV, radio, printed flyers) or online (banners, videos, search, social, affiliates, CRM…)
  • for a fee (SEM, TV, …) or free (SEO, CRM, visit to a store or call to a call center)
  • media (TV, SEM ….) or non-media (CRM, visit to a store, ….)
point de touche conversion finale

… what is the contribution of each touchpoint in the final conversion ?

Wizaly’s algorithms are based on five strong technology choices :

  1. A purely data-driven algorithmic model
  2. A cooperative model based on game theory, known as Shapley Value
  3. A unique algorithmic integration of the influence of Wall Gardens
  4. Taking into account 6 important parameters : channel, medium, source, timing, type of interaction, engagement
  5. Optimization through machine learning and artificial intelligence

Why we made those choices ?

1. A data-driven algorithmic model

Attribution models can be classified into two categories : predefined or data-driven.

Predefined models use a function defined by the user to give a weight to each touchpoint.

In this category, one can find the traditional « last-click » model, but also more complex models that take into account a large number of parameters, such as the number of pages viewed, the timing, clicks or impressions, the nature of the touchpoint etc… All those models share the same weakness, namely, the introduction of subjective human bias into the model. The result will then reflect the initial bias : a model based on the first touchpoint will give a strong contribution to social, videos and banners, whereas a model based on last touchpoint will emphasize search and retargeting.

Only those models that are purely based on data remove the subjective human bias : that’s the case for Wizaly’ model, as Wizaly’s model is purely data-driven !

modèle algorithmique data-driven
Modèle coopératif Shapley

2. A cooperative model

Models can be either correlated or uncorrelated.

Correlated models consider that conversion paths should be analyzed as a whole, since previous touchpoints influence subsequent touchpoints. Uncorrelated models consider that each interaction is uncorrelated from previous interactions, similar to rolling the ice, where previous results have no influence on what comes out the next time you roll the dice.

A cooperative model is much closer to reality. As an example, it is well-known that exposure to TV advertisement increases the click rate on a banner, or that you should cap the number of advertising banners shown a person since their marginal impact decreases beyond a certain number of exposures.

A cooperative model based on Shapley Value of modified Shapley Value algorithm is closest to the attribution issue. Shapley’s model is a model that comes from game theory and it earned Shapley a Nobel Prize in 1951. It calculates the marginal contribution of a player in a game. This model is the one chosen by Wizaly.

3. Measuring Walled Gardens’ influence

Ever heard Walled Gardens ? They are closed ecosystems where all operations are controlled by the operator of the Walled Garden.

Walled Gardens limit their sharing to statistical data. They won’t let you measure the influence of each impression, also known as post-view tracking, which makes it harder to measure their contribution.
Wizaly’s « Walled Garden » algorithm correlated multidimensional time series to calculate the statistical impact of the impressions of those « closely guarded » channels. Wizaly lets you measure the statistical influence of an exposure to an advertisement within a Walled Garden.

Algorithme Walled Garden Wizaly

4. A multidimensional algorithm

We take into account 6 important parameters : channel, medium, source, timing, type of interaction, engagement.

Taking into account 6 parameters make the model more complex, but it also makes it closer to reality.

5. Optimization using machine learning

Optimizing a problem with a limited number of conversions but an almost infinite universe of possibles is a complicated problem : typically, conversions number only in the tens of thousands and user paths in millions.

We won’t get into the algorithmics to get to the best solution. Let’s just say that machine-learning algorithms that originate from artificial intelligence are best at handling this family of optimization problems.

algorithme data driven

We let you into our secrets : conceretely, what does Wizaly do ?

Wizaly’s attribution algorithm determines the contribution of each marketing channel to conversion using only data.

The algorithm compares various paths that differ only by one touchpoint in order to assess the impact of this touchpoint to the conversion rate. Si an additional touchpoint increases the conversion rate, then this touchpoint will be allocated an increasing weight.

In this way, the model adapts to changes in the behavior of consumers as well as to the evolution of your media mix to give each touchpoint its appropriate contribution in users’ conversion paths.

schéma algorithme wizaly

Going the extra mile with data-driven marketing

Wizaly’s new « Walled Garden » attributin algorithm lets advertisers measure the real impact of social networks or YouTube in their conversions.

The algorithm calculates the statistical impact of advertising views within « Walled Gardens », the closed ecosystems that won’t let third-party solutions individually track impressions. Wizaly’s attribution algorithm uses artificial intelligence to analyze time series between Walled Garden impressions (by channel, source, country, device) and the traffic or other marketing channels to measure precisely the real impact of Walled Garden impressions. Wizaly can thus measure precisely and objectively the real impact of each channel in your marketing mix to reach your goals.

This same algorithm could also be used to measure performance in a cookieless world where third-party cookies have disappeared and only first-party cookies are allowed.

An objective vision of performance thanks to qualitative data collection.

Wizaly’s platform integrates three major data sources to provide rigorous performance analysis of your marketing mix at the desired level of granularity.

Post-click tracking : Thanks to a tag integration on the advertiser’s site, Wizaly captures all the events that take place on the site. Specifically, the origin of traffic as well as conversions, but also all the actions carried out by the Internet user once on the advertiser’s site and which are considered relevant in the analysis of consumer paths and behaviors.

Post view tracking : The various technologies offered by Wizaly, whether it is a print pixel or a javascript tag, can be adapted to all the market’s advertising agencies to track impressions. These are then reintegrated into the conversion paths of Internet users in order to be appreciated at their fair value via our attribution algorithm.

Integration of media spending : Thanks to its connector library, Wizaly is able to interface with the majority of digital players in order to simply capture the expenses related to your media campaigns.