As mentioned in my previous post “Audience Onboarding” there are very few data sources that cannot be onboarded through one method or another.
The main thing to look out for when choosing a 3rd or 2nd party audience, is the accuracy of that audience.
There are many providers of audience segments within the digital advertising environment. Some are offered by the main players (Facebook and Google) whereas others are compiled from various more transparent sources.
By transparent I mean that you can research the collection method and make a judgement over the accuracy and suitability to your campaign.
Points of Interest – Audiences compiled from their proximity to a place of interest or event. Sports Stadiums, Concerts, Car Dealerships etc. This is a common way to make assumptions around interests and is often used for real time OOH or Mobile advertising.
It has also been used to compile standard audience segments however. Be wary that interests haven’t lapsed such as targeting visitors to a car dealership.
Transactions – Audiences compiled from purchase history data. Much like point of interest, this audience source should be relatively accurate to the assumptions it makes.
Behavioural – this tends to relate to actions taken such as liking or sharing certain content in social media. Whereas this would suggest a direct correlation with an interest it is usually less suggestive of an intent to act. For instance, a like or share could be down to appreciation of the content (finding it funny or aspirational) rather than an intent to buy that product.
Declared – these audiences are usually sourced from surveys or forms. Information such as date of birth, home address, charity support or financial services used, can be collected through lifestyle surveys for specific use in marketing activity. This data is accurate but usually less suggestive of a specific immediate intent to act.
Derived/Modelled – These audiences are ones that I would be the most wary of. Models and profiles are often used to expand audience scale however it is rare that a provider will share information about how the model has been created or how large the sample data is.
Some models can be accurate, such as models created from finding look-a-likes to your own customer data. In most cases however, the accuracy of the model will be directly proportionate to the accuracy of the seed data being modelled.
Even with 3rd and 2nd party branded audiences however, things aren’t always what they seem. You may see things like Credit Card brands or Government Bodies and start to wonder how they have been able to make data available for advertising. The answer is that they are aggregating and matching at a very inaccurate level such as postal area.
The key to understanding accuracy is understanding source and matching process.
If you have any questions around audience sources and matching please feel free to contact us for a chat.