Monday, November 4, 2019


DMP, CDP, Data Lakes, Oh My...
A Data Primer for Marketers


As someone who started my professional career in Marketing, I along with many of you have a love/hate relationship with data. Love in the sense that having the right data is all-knowing and can bring you wealth, power and fame. Hate in the sense that not having all of it, or having it and it’s not complete, not accurate and/or not relevant. 

We all know customers fully expect a consistent, cohesive conversation, from the moment they first meet a brand to their last interaction, regardless of channel or device. To pull this off, marketers require…well, you know, data.

The lines between marketing expertise and data expertise continue to blur and may in fact already be almost indistinguishable. Marketers are now expected to be data scientists, or at least be able to have intelligent conversations with those who live and breathe data every day. (BTW that's marketers).

If you find yourself in this position, (and want to impress your colleagues with your data chops) below are definitions and use cases for common data platforms designed to get us marketers closer to customer experience expectations.

CDP (Customer Data Platform)

The biggest difference between a CDP and other types of data platforms is that a CDP is used to manage customer data. It is a data resource used by and typically controlled by marketers. While DMPs tend to focus on top-of-funnel initiatives, CDPs focus on mid to lower-funnel activities.

Think of a CDP as a central repository of siloed pockets of customer data; things like personal identifiable information (PII), transactional data, behavioral data, explicit data, offline data, call center data, etc. that marketers need in order to create specific customer segments or to drive 1:1 conversations. Additionally CDPs can also clean and dedupe this information. The real magic is that it matches all these touchpoints into a single customer record, holding and consolidating a customer’s interactions across channels/devices. Lastly it can feed this information to other platforms, like BI tools, DMPs and marketing automation platforms. It can be a dream come true for personalization, as well as customer insights, especially for those marketers currently frustrated by data silos.

If you are shopping for a CDP, one important point of comparison is the breadth of off-the-shelf integrations each solution has. Make sure the solution can easily integrate to your other data systems. Another point of comparison is the extent with which these customer profiles are updated in real time. These single customer records are useless if they are even a few hours old.

There are a host of CDP point solutions on the market as well as CDP functionality built into some existing enterprise marketing automation platforms. If you are interested in learning more, below is a list of CDP vendors to get you started. It is not an exhaustive list nor a recommendation, but a good place to start additional research:


DMP (Data Management Platform)

First of all, who came up with all these names? Data Management Platform, Customer Data Platform - good grief. I suggest a new acronym: CONFUSED (Customer Or Non-customer Frameworks for Unstructured or Structured Enterprise Data). Not bad eh? It only took me 5 days to come up with that 😊.

Anyway, a DMP is typically used for top of funnel initiatives as mentioned above. While it can ingest 1st party data, it focuses primarily on 2nd and 3rd party anonymous data. It’s great for web display ads, retargeting, look-alike modeling, and other lead gen efforts at scale. 

Publishers also use DMPs to monetize their data. Unlike CDPs that keep an ongoing historical record of activities, DMP data is often kept for shorter periods of time. Basically you can model the attributes of your best customers and have a DMP go out and find more just like them. One important point of comparison is match rate. How effective are they at matching these attributes and stitching together the identity dots across cookies, device IDs and IP addresses? 

If you’d like a more in-depth DMP primer, check out my DMP Primer video I recorded back in 2016. It’s still relevant today.

Below is a list of DMP vendors to get you started. It is not an exhaustive list nor a recommendation, but a good place to start additional research:

Data Lake vs. Data Warehouse

Data Lakes (as well as Data Warehouses) are both used to store big data. However a Data Lake holds raw unstructured data, the purpose for which is as yet undefined. Consider it a dumping ground for any data you think you might want to use sometime in the future. A Data Warehouse typically holds structured data that has already been filtered or processed for more specific purposes. A Data Lake is used by Data Scientists, while Data Warehouses are often used by the broader business. Companies often need both, as each requires different amounts of storage capacity and specialized tools to manage.

 BONUS: AI vs Machine Learning 

These two buzz words are thrown around everywhere, even in real estate broker ads. So what’s the difference?

According to the Artificial Intelligence Marketing Association, “Artificial Intelligence (AI) is the emulation of human intelligence process by machines be it a computer or an IOT device. The process involves learning (extracting meaningful insights and patterns), predicting (use the gathered information to help devices make plausible future decisions) and self-correcting (the art of getting more efficient every day). This is made possible by the three major wings of Artificial Intelligence — Machine Learning (ML), Natural Language Processing (NLP) and Deep Learning (Neural Networks). AI has established itself in every business sphere around the world -encompassing everything from Rule-based machine learning to visual classification. Its applications range from object recognition to preventing high-end cybersecurity threats.”

Artificial Intelligence is what Tesla uses to drive cars autonomously. It’s what manages intelligent stock trading systems. Here are some very interesting stats about AI, published by CMO.com

Machine Learning is what the iPhone uses to blur backgrounds in Portrait mode. It’s what virtual assistants use in their speech algorithms, (using NLP Natural Language Processing). Waze, Google Maps, any of the driving direction apps use lots of machine learning to find you the quickest way there. 

All in all, it’s a great time to be alive and enjoy all this cool technology! Hopefully you now have a better understanding of these terms. Now go impress your favorite data architect, or better yet, your boss.

Steve Kellogg
Crowds2Crowds

1 comment: