Understanding Multi Data Planning

Multi Data Explained

The ultimate goal of any multi-media planning system is to be able to combine different media sources and best estimate reach and frequency for the total campaign. The biggest obstacle is to intelligently combine the siloed media data while accounting for consumption duplication so that we can estimate deduplicated reach.

However, whatever method is used, it should best reflect the currency R&F per media type. That is, the TV reach distribution must be similar as the one for the schedule in the actual planning system. With Multibasing or fusion of the data beforehand, it is impossible for the results of current integration techniques to give you identical results to its original survey. This is simply due to there being two different respondent sets along with their weights. This method seeks to primarily attend to both mix media R&F and maintain original distributions, but also incorporate other TelmarHelixa goals.

Advantages over current Multibasing and Fusion include:

Plan Version 2.5

1. Maintains the currency R&F. Users don’t see different results of the same media schedules reach when combined. Inherently in MultiBasing and Fusion, the newly created dataset can maintain some metrics of the original data but when a media plan is applied to the data, R&F results are always different to the original survey as we are doing R&F calculations on a different set of respondents. Here, individual surveys are maintained and we combine their campaign total survey results AFTER doing the R&F and not before by mixing the individual vehicles and then calculating R&F.
2. Allows for Brand (BMW drivers) targets on the HUB survey to be used for R&F estimates. This is similar to what is done for other TelmarHelixa tools 'Transmit' and 'AudioPlanner' to give Brand R&F estimates on currency data.
3. It is additive and respects media cumes. I.e. Men + Women = All adults R&F - unlike our previous MultiBasing.
4. There is no declining reach problem inherent in Formula based R&F.
5. Respondent level mixing of R&F distributions make for much easier and faster calculations. We do not have a set of results per target. One set of data to which multiple targets may be applied. No calculated duplications per pair of vehicles which cannot be scaled (made fast)!
6. Easier data setup. No need for common set of variables across all imported Data types. When combining 3-4 surveys we do not require common demographics across all datasets. Merely that each one may be combined with the Target survey.
7. Does not regress to the mean by using single set of cells across surveys.

AI / ML Extensions

1. Allows for AI and ML algorithms to be applied to the probabilities at a respondent level therefore making it much more powerful and using the complete target survey’s capabilities.
2. Allows for the Creative adjustment of the probabilities before rollup.
3. Allows for segmentation models and adjustments to be media specific.
4. Solus and additional reach per media can be calculated.

Method

There are 3 main steps in calculating the R&F for a MultiSurvey run :

1. Perform R&F as normal for all media vehicles within each survey.
2. Export all R&F distribution results by cell, per survey and import into the Target survey.
3. Calculate mixed R&F on the Hub/Main/Target survey and apply the chosen Brand or Demo targets to the results.

Survey R&F

For every survey in the campaign we calculate R&F totals for each vehicle, group total and campaign total. That is, given your spots/GRP’s we do the exact same calculations as you would do if you had the campaign in single survey mode. Therefore R&F will use the exact formula the single survey mandated and there is no question of the original data gives different estimates of reach.

This R&F is done for each cell or demo combination that we have set up in data. Therefore a schedule target of Men would match exactly since Gender is one of the common demos used. Same for other common cells we decide to use like Age groups, Social Class. 

The important part to note here is that our final mixed R&F will match identically for any demo used in the cell makeup. For all other Targets we will copy the method used in Transmit/Audioplanner and estimate a R&F. It is only possible to cover the main demos for size of calculations as is the case with any MultiBasing, Fusion or even general survey weighting for data suppliers.

Key Benefits

• Far superior method to MultiBasing or any Fusion, as not only will GRP’s match the original data, but so to will the actual reach curve by preserving total currency duplication across vehicles and using the original data and models.
  • We ensure that R&F at individual vehicles match.
  • Combination of vehicles in the same survey match like TV total.
  • Combinations of Media types in the same survey match at campaign total level.
  • Anything can be compared by a client and should match when using a common demo.