Why introduce segments with different affinities?

Advertising is in essence, reaching an audience with a pervasive message in order to achieve certain goals. A traditional planning process carefully defines a target audience, analyse their online behaviour and identifies the full universe of online media properties and placements that are capable of reaching the target group. Another often more efficient option is to leverage data to help you to reach your target audience.

AudienceData enables you to spend your marketing budget on the right consumers directly, reducing the amount of impressions that are wasted. In order to pick the right strategy for any campaign you need to be able to measure and compare the effectiveness of different strategies.

Measuring the effectiveness of online media properties and placements have for years been defined by well-known metrics like affinity and reach. Measuring the effectiveness of data have however been more ambiguous. One problem that is increasingly present in many of the data audiences being offered is that behavioural data is used for classifying visitors into different demographic, interest or intent categories using increasingly complex machine learning algorithms.

The challenge with machine learning is that data scientists sometimes decide to rely on complex unsupervised learning models and end up releasing new data audiences where the algorithm that created the output is more or less impossible to understand for humans. A black-box approach. When that approach is used, the data output will often fall under Clarke’s third law:

Any sufficiently advanced technology is indistinguishable from magic”.

And it is hard to sell ‘magic’ in an increasingly data-driven world where we rely on validation, performance-KPI’s and hard facts. Given the fact that all data are not created equal, AudienceProject has decided to add a declaration of content to our available demographic data segments. Moving forward our demographic audience segments will be rated by affinity.

Why rate probabilistic data segments using affinity?

Affinity is the definition of a data segments performance against a particular target audience versus the performance if you target the average population. Affinity is the metric that allows you to compare the performance of programmatic data driven strategies versus traditional media placement planning. A data driven strategy should only be pursued when it delivers more value than the traditional approach.

Affinity is also the metric that quantifies the reduction in wasted impressions on any given campaign. It puts a very tangible monetary value on the value that a skilled planner can add to an online campaign.

How did we do?

How to document the accuracy?

​Using targeting with the right conditioning