It’s a simple truth that the quality of data can directly affect the efficient deployment of marketing budgets and campaign success. So, it’s all the more baffling that many online advertisers don’t question the origin, accuracy or scale of the audiences they buy programmatically.
Right next to fraud and viewability, measurement and audience validation should be at the foundation of marketing success metrics for 2017.
Defining Data
Firstly, it’s important to understand some basic truths about data in order to understand it’s role in programmatic advertising.
-
Offline data
This tends to be derived from PII (personally identifiable information) -
Online data
Anonymous data even if it was derived from PII originally -
1st Party Data
Data which marketers collect and own e.g. site registrations, surveys. This data is often deterministic meaning it has one-to-one correspondence with the actual attributes and behaviours of individuals and their devices. This data has greater accuracy but can often lack scale. -
3rd Party Data
Data owned by another source who then sells it to marketers. This data can be deterministic or probabilistic (based on inferred or modelled attributes rather than one-to-one correspondence). When probabilistic, this data is easy to scale but can be less accurate
What are the data dilemmas?
Accuracy Drops
Many companies strive to move their offline customer personas and models online. However, this process typically causes data segments to retain only 20-50% of their accuracy. To compensate, companies then over-model, leading to poor online targeting performance.
Time Sensitivity
On-boarded data (and most 3rd party) is time sensitive. The time needed to integrate data into a DMP/DSP usually causes additional data decay. Using this stale data causes results to suffer.
Weak Models
The more attributes you have on consumers the better you can predict their behaviour. However, if using 1st party data – which can suffer from a lack of scale – modelling platforms will not have enough data to provide accurate models
Applying the Wrong Measurement Methodology
Data measurement is not a one size fits all process. Programmatic systems use either panel or predictive measurement. Panel measurement relies on small deterministic data to infer coverage while predictive measurement scores the entire internet with a probability score for greater granularity. While panel measurement works well for targeting large samples, accuracy drops when measurement requires more granularity.
The data you need
Online advertisers need to constantly optimise their campaigns to drive the success of their campaigns. This means not only using quality data but also quality measurement methodologies. Advertisers should seek to leverage data that is:
- Fresh – data that was recently collected
- Massive – smaller data can make modelling difficult and inaccurate
- Accurate at scale – to target large samples, it should not loose accuracy when scaled
This means that as a marketer you should:
- Never assume your programmatic partner’s data meets your standards
- Ask lots of questions
- Know the methodology used and the source of all the data that powers your targeting and measurement
Contact Anthony Ord from Acquire Online to see how to use data most effectively to maximum ROI. Ph 027 649 9198
Source: Huffingtonpost