Key considerations when comparing DMPs and CDPs to InfoSum’s UDP - Part Two
In part one of our series, we looked at how we define each of these platforms, and the infrastructure and data types utilised by each. Now, in part two, we will examine how each solution enables data collaboration and handles identity and data cleansing.
Data collaboration is the ability for multiple parties to benefit from the insights unlocked through the combining of their various data sets. A prime example of this would be data consortiums such as The Ozone Project. Unfortunately, data collaboration in the DMP/CDP world poses challenges, due to their centralised infrastructure. This centralisation of data means that each party would have to pool their data in a central repository, significant security and privacy risk and one which brings with it the challenges of arduous InfoSec/compliance reviews.
Due to the federated architecture that underpins InfoSum’s UDP, data collaboration becomes much more seamless, secure and privacy-safe. Each party remains in complete control of their data. Through our fine-grained permission controls, they can enable other parties to analyse their data, without any of the data sharing risks. The use of various privacy controls (such as differential privacy techniques), means consumer privacy is never risked.
The ability to tie identities to behavioural and demographic data is a vital process to powering programmatic advertising and people-based marketing as we know it today. For DMPs, this is achieved through a cookie syncing process which involves cookie data being widely and frequently shared and mapped between parties and platforms. However, this is only possible for third-party cookies, which are already blocked entirely on Safari and Firefox, and Google committed to retiring third-party cookies by 2022.
The alternative method, available via both CDPs and InfoSum’s UDP, is to utilise an identity graph. However, the approach that InfoSum and CDP providers take to utilizing identity graphs is very different.
CDPs generally go down one of two routes; they either build their own identity graph, or they will license a sense of identity from a third-party, which requires the third-party to append a master ID to their data. Utilising their own identity graph to resolve identities is a step in the right direction for CDPs, this ensures that data is not being moved around to onboard offline data.
However, there are a couple of pitfalls. Firstly, it is likely that multiple parties identity data is being combined to build the identity graph. And secondly, in their early days, these identity graphs will lack the scale that established players have.
InfoSum’s approach to identity resolution continues our theme of non-movement of data. First-party data and an identity graph can be virtually overlayed for data onboarding and activation, without requiring the first-party data to be shared. This means that not only does everyone retain full control of their data, they also protect their own identity as it’s not being used to power the identity graph.
The InfoSum approach additionally unlocks the option of multiple identity graphs being layered together (for example, desktop, mobile and TV), rather than being tied to a single one. This powers a holistic approach to media planning.
Subscribe to the InfoSum Newsletter
Data comes in all different shapes and sizes. Therefore making these disparate datasets ingestible by a single data platform often requires a process commonly known as ‘extract, transform, load’, or simply ETL.
For DMPs, the majority of the data they store is unstructured ‘log-level’ data focused on web activity and tied to the DMP’s central ID, therefore there is no data cleansing required.
However, where a company needs to onboard their offline data into either a DMP or a CDP, there is significant data manipulation required before it can be read and mapped to a users profile. This process can be long, painful and expensive as it requires a great deal of data manipulation.
InfoSum’s UDP flips ETL to ELT (extract, load, transform) by requiring no change to the original data before uploading. Instead, our normalisation process standardises and maps the data to our global schema. This process is almost entirely automatic, with human intervention only required occasionally when a data field does not map to an existing data category. This unique approach to ETL removes the need for expensive data migration and manipulation processes and allows data to be uploaded and available for analysis in minutes.