The renewal of the model of Mediaplanning: how to improve prediction in a digital world
Mediametrie
Awards: Advertising Solutions (Highly Commended)
Project summary
In a world where advertisers seek for more transparency, measuring campaign performances with a 3rd party, independent & audited body is key.
With the success of post campaign analysis solutions such as Nielsen Digital Ad Ratings, advertisers can now precisely track daily campaigns reach, frequency and impressions by demo.
Predicting campaign performances is a more complex science. Historically Médiamétrie has been proposing in France a Mediaplanning solution.
Mediaplanning is currently based on Internet Global audience (IAB 2018 Research award winner) which reports top 6000 Brands and top 1000 applications monthly. Individual data are shared with Centre Servers which distribute calculated probabilities through their tools to agencies, saleshouses to evaluate and optimize reach & frequency for a selected mediaplan.
However, mediaplanning relies on theorical audiences which shows some limits; most of the time, we observed over estimated reach, especially when selecting multi publishers advertising packs or programmatic campaigns. Médiametrie developed a scientific approach to reconcile pre and post campaign evaluation.
1. Objective(s)
The objective of Médiamétrie project was to offer more precise campaign planning recommendations by reconciling planning and post campaign performances.
Firstly, the aim of the R&D was to set the right balance between the reach and the frequency globally and per support by defining calculations to be applied by the Centre servers.
Secondly, by developing “EvalWeb”, its own Mediaplanning API, Médiamétrie aims at proposing its own engine to connect mediaplanning data to the whole digital eco system.
2. Methodology
Methodology consisted in comparing DAR results based on 47 campaigns & 192 supports, and the equivalent prediction in Mediaplanning.
Methodology took into consideration features related to digital mediaplanning such as share of voice, capping and target page purchasing. A key approach was to challenge the models and particularly the assumption of the independence in the distribution process of ads on pages viewed on a site by an individual which tend to affect the levels of coverage.
Médiamétrie proposed a Markovian Binomial model defining the level of correlation to be applied by type of support.
3. Key results
Médiamétrie model showed positive results.
For all the 192 supports, the deviations of coverage with our new model went on average from 1.6pt to 0.3pt and the variation of the deviations around this average was attenuated (the standard deviation going from 0.0186 to 0.0108).
The model we developed also improves the forecasts by type of support. We can say that the distributions by support are more or less centred (average around 0) and that the standard deviation reduces.
More details are available in the document downloaded “IAB Europe Research Awards 2019 Mediametrie Eval Web – Summary & Results” (p11-12).
4. Impact and Application
Centres servers all accepted to implement those new calculations to reinforce Digital mediaplanning role.
Agencies strengthened their role on the advertising market by setting up mediaplan based on independent and audited data.
EvalWeb, the new web service (single calculation engine transported on a distant server) to benefit from those calculations with a simplified implementation of the solution and optimized calculation times for the mediaplanning service providers is in beta test.
5. Significance
Trust reinforced for media agencies and their clients when establishing their strategic recommendations as predictions are stronger and coherent between the various mediaplanning service providers
Other fields of application for other medias or in addition with Internet: this approach is valid for the Press mediaplanning (print + digital contacts)