Using data to bring focus to who your customers actually are
I recently had the chance to present at The Media Kitchen’s 12th Annual Venture Conference. The theme of the event was “The End of Serendipitous Discovery” and focused on how customer no longer stumbles upon brands, but rather brands have to focus on how to get in front of the right people and make them customers.
My presentation made the case for data being a core component in any strategy that is trying to find the right consumers in an effort to make them customers. There isn’t anything particularly interesting about that point of view, but I find that too often people will acknowledge the potential of data in helping them understand their customers, but their fundamental understanding of what that means is surface level.
I suspect that the primary reason for a less than nuanced understanding of data’s power can be attributed to the fact that “data” is a nebulous term. It is used as shorthand for hundreds of potential assets, strategies, and underlying representations of the world around us. With that comes an over-simplified, if not muddled, view of what data is and what data can be.
You can find the entire presentation here. Below I’ll walk through some of the key points I touched on at the conference.
There isn’t much controversial here.
Again, this should be fairly self-explanatory to people, but it is an important distinction to make. I often hear people assert that “first-party data” is high quality and “third-party data” is crap. It’s an opinion that is rooted in some historical truth, but there is nothing inherently crappy about third-party data. Instead, it relates to how third-party data has been packaged for consumption.
Ideally, we could eliminate any preconceived notions we had about data so we can better grasp its strengths and weaknesses. That being hard to do, and data being nebulous we’re going to spend the rest of the presentation talking in the context of images – which seems fitting for an Instagram obsessed culture.
Here we’re representing first-party data as pixels that make up an image. As you can see, each piece of information is represented by a different color.
When assembling all of our first party data we start to get an image of who our target customer is. In a perfect world, a brand would have enough first-party data to build a high-resolution version of this picture, but that is almost never the case. Instead what the brand gets looks more akin to an 8-bit character from an 80’s video game.
Now we’re introducing third-party data as it has been thought about traditionally. It shares some similarities with first-party data insofar that it encodes information about the user. The similarities stop there. In our rendering of third-party data, you should notice that the pixels are much larger than the pixels in the first party data slide. That is meant to represent that third-party data is not nearly as precise as first-party data can be. You’ll also notice that some third-party data disagrees with itself, which is related to the precision, but also a side effect of the sources of the third party data having different incentives (sell as much as possible) then the buyers of the third party data (make sure it’s specific and useful).
The result is that when you mix third-party data with first-party data, you don’t create a sharper picture. Instead, you create an image that an impressionist painter might be proud of, but that does nothing in helping you understand your customer.
The resulting image is a function of mixing precise first-party data with imprecise and often incorrect third party data. One might argue that we’re done here. Case closed. Let’s just stop using third party data. In fact we’ve seen this argued before. I would suggest an alternative. What if instead of having third-party data look nothing like first-party data we could bridge the gap and have them take the same form.
Yeah, I’m saying it can happen.
Keeping our image metaphor going, let’s take a close look at what I’m suggesting. The thing to keep in mind on the following slides is that typically third-party data is non-specific. It is a binary representation of a classification someone has made about a person. Are they a business traveler? Yes or no? Do they plan on buying a car? Yes or no? Are they passionate about bird watching? Yes or no? The challenge lies in that each of those questions is somewhat subjective. How often does someone have to travel for business to be a business traveler? When do they plan on buying the car? How does one define passion?
As you can see here, we are no longer trying to use third-party data for the purpose of creating a broad classification, but rather we are using it to define very specific actions at a snapshot in time. Simply put we’re making the third-party data represent something more precise. Something more akin to the first-party data we represented earlier in the presentation.
One of the nice things about encoding data this way is you can have it represent completely disparate behaviors. In the previous example, the data represented a location. Here it represents a behavior on their mobile phone.
And here a browser-based behavior.
Now that we’ve created more precise pixels, when assembled we create a higher-resolution picture.
Here I am, in all of my glory.
Now it should be noted that in the real world brands aren’t actually creating a high-resolution image of any of their individual customers, but rather they are creating a composite image of specific customer cohorts. In fact, in the same way, that you likely don’t care what color any given pixel is on your high definition TV is brands need not care about the actual information encoded in the examples above.
In fact, there are ways to basically remove all of the underlying attributes from any given pixel and have it represent information encoded to something analogous to a color. This allows folks to create these pictures, but to do so in a way that is mindful to consumer privacy and compliant with various jurisdictional restrictions.
In an odd way, it is somewhat related to “This Person Does Not Exist“ which is a site that shows images of what looks like humans, but in reality, are all AI generated. That technology works by feeding that AI a lot of pictures of actual humans and letting it create a net new person algorithmically.
In the end, data is going to be an important component of any brand’s strategy going forward. It’s our hope that they start to look at more precise data sets so they can encode a higher resolution composite view of their customers.