Products & Services

One brand or product, similar period, advertise both on TV and online. Question is how to optimize the media mix. What would be rational budget allocation? How to maximize the total reach?

For that task we need to know:

  • What % of target group has seen the ad?
  • What was the duplicated reach, it means how many people saw the ad both on TV and on Internet?
  • What was the incremental reach of internet, it means how many people saw the ad online but not on TV?
  • What was the combined reach, it means how many people saw the ad either on TV or Internet?

How to solve it?

There are different approaches, all them are some sort of variation on following scenarios:

  • Random duplication
    • Great possibility as benchmark. Problem is that it does not reflect the specifics of single campaign, target group etc.
  • Single source measurement
    • Without exaggeration it can be called as “Utopia” as there are so many obstacles in sample size, data quality, costs and so on
  • Data fusion
    • Good and quite used solution, but quite demanding as it needs to work with TV and online respondent level data
  • “Imprinting”
    • Statistically advanced solution during which aggregated online campaign results are statistically “imprinted” onto (extended) TAM panel

Data sources

Solution of Nielsen Admosphere – SimCross – belongs to the last group – “imprinting”. There are 3 different sources used:

  • Results of any online campaign measurement
    • Reach and its sociodemographic profile
    • Impressions per websites
  • Nielsen Admosphere TAM panel
    • TV measurement
    • Respondents’ demography
    • Internet usage frequency / home connection
  • Continual Survey
    • Time spent on Internet
    • Concrete TV channels and websites

Overall scheme

Main characteristics / advantages of SimCross™

  • Combines single source data and statistical modeling
  • Independent on specific online measurement – it is possible to use anyone, only condition is to provide needed information
  • TAM panel is used as a „playground“ for respondent level calculations
  • Possible to report results for any target group
  • Does not introduce additional statistical error