10 Step Guide to Using DMPs for Insights

Data Management Platforms (DMPs) are an increasingly important part of the advertising technology ecosystem. Their role is to combine data together – often 1st and 3rd party – and use that data to create audiences that clients can deliver media to programmatically.

And these data warehouses are HUGE, often containing data on almost every adult in a population. The number of data points on each individual is also increasing rapidly.

So with great scale and the ability to connect data sources together, DMPs could also be fueling the next generation of market research. But how would an agency or advertiser go about making this happen?

10-steps

1. Start with a specific use case: Just because you can connect data, doesn’t mean you should. Leveraging technology is a process and so you should start with the most valuable use cases first, to make sure you see return on your investment quickly.

2. Audit your own data: The data within DMPs is available to anyone who wants to pay, so the real value comes through combining this data with your unique 1st party data. This could be customer data, audience data, website traffic or primary research. To connect the data, you need to have personally identifiable information (PII) in your data set such as name, email address, postal address, device ID etc and the appropriate consent. You can also tag your digital assets and campaigns to make sure this data is able to flow through as well.

3. Identify the data you need: If your use case is about understanding share of wallet, then you will need sales data. If you want to segment a population based on life stage and lifestyle, then you’ll need detailed demographics.

4. Work directly with data owners: Much 3rd party data is commoditised, and quality can be hard to verify. When you work directly with other data owners, it becomes far easier to know you are getting the real thing.

5. Make sure you can measure success: Many people both agency and client side will be skeptical of efforts in this space, for good reason! You must have a way of measuring the success of any endeavor, and set milestones along the way that must be met to unlock additional investment.

6. Interrogate the data: In this world there is sadly a lot of terrible data. If you know someone who can point out the good from the bad, then you are in luck. If not, then you can ask the following questions that can help you to judge the quality:

• What is the source of the data?
• Has this data been verified at the individual level?
• How much of the data has been modelled?

7. Decide where the data will be stored: You can partner with DMPs and ‘onboard’ your data into their systems where it is automatically combined with the other data in there. This is easy to do but sometimes you may not want to export your data, or you may have systems and capabilities internally you want the data to flow into. Some DMPs will allow you to bring their data into your systems. The primary downside here is cost as you may incur additional license fees for the data.

8. Get data science involved: Wherever you decide the store the data, I can guarantee one thing… It will not be analyst ready. Work will be needed to transform the data into something more usable. This is not a job for excel.

9. Get marketing analytics involved: With bigger data sets, comes bigger possibilities. Get analytics support in thinking about best to analyse the data to deliver against your use case.

10. Make sure you have an ongoing business case: If you haven’t realised this already, leveraging DMPs for insights is no small undertaking. You will likely need a business case upfront to justify the investment and an ongoing business case as you look to build further.

Despite some of these challenges, we do believe the next generation of market research will increasingly utilise DMPs for insight. The benefits of connecting data at scale will allow us to finally close the gaps between insights, marketing and customer data.

References:

What is a DPM?

Why you need to interrogate the data.