Kevin asks, “What’s the difference between unrefined and refined data?”
As part of a larger discussion about metrics in marketing, Kevin’s question came up in reference to the importance of data. Unrefined data is as useful to business as crude oil is to a car – which is to say, it isn’t useful at all. In fact, like crude oil put in a car’s gas tank, unrefined data can be just as harmful to a business. Only when data has passed through the data lifecycle will it transform into business impact.
Watch the video to understand the complete data lifecycle.
Can’t see anything? Watch it on YouTube here.
Listen to the audio here:
- Got a question for You Ask, I’ll Answer? Submit it here!
- Subscribe to my weekly newsletter for more useful marketing tips.
- Find older episodes of You Ask, I Answer on my YouTube channel.
- Need help with your company’s data and analytics? Let me know!
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.
In today’s you ask I answer. Kevin asks, What’s the difference between refined and unrefined data? This is in response to a discussion thread we were having with the the PR student chat on Friday about the usefulness of data. One of the things I like to say is that data is the new oil, which is a great analogy. This is an expression I heard first in 2006. And when you think about it, it makes total sense because data by itself like crude oil, not super useful, right? It’s kind of messy, sticky, gunky the crude oil doesn’t burn very well unrefined you know uncleaned data doesn’t analyze very well doesn’t certainly doesn’t give you insights and things. So what does it mean when we talk about refined versus on refined data data has to go through a process the same as crude oil in terms of how we turn into a usable product. And that’s the data life cycle. So let’s go ahead and bring the
this is the data lifecycle from red to green around the circle, all the different steps that you need to, to be able to take to process your data and turn it into something useful. So each step can take a tremendous amount of time, depending on the size of the data, how clean it is, where you’re getting it from, whether has to be merged with any other forms of data. So all that stuff has to be taken into account as you start working with your data. So let’s dig into this. What exactly is is in the box as it were, when we’re talking about the data life cycle, you first have to ingest your data, which means you take it in from whatever sources you’ve got new data from, that could be databases, it could be unstructured data, like social media data or news articles. If we’re talking about public relations, it could be machinery data, or aircraft engine data, depending on on what kind of data we’re working with. We’ve got ingest it which means we’ve got to take it in and get it into some kind of format that we can then start to work on the next thing to do is
You do your analysis, you you look at the data and see what condition is it in? This is the first step in what’s typically called exploratory data analysis. And this is what’s data is missing what data looks like, there are anomalies are there, formatting problems, things like that, once you’ve done your analysis, you repair it, how do you fix the data, make it compatible with the systems you’re going to be working with. You fill in missing values, if you need to do amputations, stuff like that. The next step is to clean the data, which is to remove incorrect data, again, with depending on what you’re working with. This could be system anomalies. This could be interference. If you’re working in public relations, and you’re trying to get an understanding of a media space, you have to really feel the last year so you’ve had to include filters for certain politicians by name because they seem to soak up and inject themselves into every single news story that it hasn’t even things have nothing to do with your client. So that’s sort of this cleaning process.
Once you’ve done the clean, you prepare the data for analysis. And that means typically
do restructuring it as needed of reformatting it. So for those who are database geeks, and this is going from either normalization or do normalization, making data work with the software that you’re going to be working with. The next step is augmentation, which is when you take data and you add additional data to it. This is especially important in machine learning where you’ll need to classify or quantify or provide other insights to your data. So that may mean for example, turning qualitative variables into into semi quantitative by transforming it into dummy variables, you may need to add additional data from the outside or emerging additional data sets once you’re ready to start processing the data beginning by more exploration, what are the connections what are the correlations and and what are the unusual things you can find that the data you compare? It depends especially
Doing machine learning with other models. If you if you are doing
validation, you’ll have a test, a training data center test data set. But you’re going to compare your data to other known good data sets to make sure that you’re getting valid conclusions or potential conclusions. And then you move on to really what’s the predictive portion of the data, which is, in a lot of ways, like a hypothesis in the scientific method. I mean, it is hypothesis and the scientific method, you predict what the date is, it tells you and then you prescribe where you come up with a solution for what to do based on that data. And depending on the model you’re building that may be something that a human does, he made hand off that that may be a part of the process. We’re handoff some analysis to another part of business. It may also be a machine that does the prescriptive work that says, Okay, I’m going to take it all new data and use it and match it up with the existing data. This is modeling This is the modeling portion where you take all those predictions and progressive
and turn them into machine rules. ways that you can deploy your data in a in a scalable way. And then you validate the model. you test it, make sure that it works that I works as intended. And you in a lot in many ways you you check to see is your hypothesis correct or incorrect is a true or false as with all things in the scientific method, a hypothesis is a true or false statement. So you want to make sure that your data gives you your model gives you that answer. Once you validated it, you move on to refining and how can you tune it up and improve it without overfitting it to make it as accurate as possible, as refined as possible and then you deploy your model across your business that can help all the business users with their data. Then you observe what happened when you rolled out this model. Did the the end result that you were looking for get better or did you create a result that you didn’t have before. Now that is a lot to cover in just about
Five minutes here of the data life cycle. But all of these steps are connected, some of them are automated, some of them are not. Some of them use human judgment. Some of them use machine judgment. But all of these are parts of the state of life cycle that you need to go through in order to to really get the most out of your data to turn it into that refined product that
that the business can use that your users and your and your business stakeholders can make practical use of when you think about measurement and analytics in whatever discipline you’re in. from public relations to trance oceanic shipping,
the one thing is missing from this is sort of the overall strategy. What’s the goal of the data and that that happens outside of the life cycle that happens before you even touch data is what’s the goal and then what are and how do you know you’re going to get to a goal, what tactics what choices will you make, and then how will you execute the data life cycle so that’s the one thing I would say.
is not here that it is presumed that you have done in advance before you start working with data in order to make refined data, refined data can take a very long time to put together. Refined data can be very difficult, very expensive, good. And that’s why data scientists are in such high demand right now. And so be prepared for that. When you begin your data journey, be prepared that it’s going to take a while and that it’s the answers will not always be obvious. And that that it will take a lot of effort to turn it into a truly usable product. But once you do, your business will scale faster than you can possibly imagine. Because you’ve got the data and other people don’t or the people’s data may not be as good if they haven’t followed the process as well. And that’s how you turn your data into a competitive advantage. you execute the data lifecycle better and faster than your competitors. That’s why artificial intelligence and machine learning are so critical now to data because the value
You have AI is acceleration and accuracy, better data, faster data. So you go from data insights to deployed strategies so much faster when you have a on your side. So great question, Kevin great discussion about how we can be using data. And as you can see, no matter what profession you’re in, this is going to have a major impact on every line of business. The faster you get to embrace machine learning and artificial intelligence, the faster you’ll take advantage of the data you have and turn it into business impact. Thanks for the question. As always, please subscribe to the YouTube channel and the newsletter and I’ll talk to you soon. Take care
if you want help with your company’s data and analytics. Visit Trust Insights calm today and let us know how we can help you
Want to read more like this from Christopher Penn? Get updates here:
Get your copy of AI For Marketers (2019 Edition)