Speaker 1: Welcome to the Data Strategist channel. My name is Stuart and today's question is what is a data strategy? Well within this video I'm going to give you a brief high-level overview of what a data strategy is and hopefully it's going to give you a good starting point to begin that data journey that your company needs. So what is a data strategy? Where can I get one? How much is going to cost? What technology stack do I need to invest in? Who do I need to hire? These are all the things that you're going to ask yourself or be asked when you're trying to define a strategy or come up with something. But in essence a data strategy is a defined approach on what we can do with the data that we collect or generate within our product or business. And from this we need to define a mission statement or a really defined strategy from that. And a mission statement can look like this. As a business we need to become a data-driven company to be able to make intelligent decisions based on customer experience. There's a really key phrase in there and that's a data-driven company. And I want to take two seconds just to quickly analyze what a data-driven business is. There's a lot of online opinion pieces to white papers but for me it boils down to two factors. The first one is decision-making. So making decisions with data and analytics over assumptions. Now this allows you to keep up to date with your ever-evolving customer base and their actual needs. But more importantly for me is the company culture. And this is allowing your employees to read, work and analyze data sets. Now let's take an example of that. An ops manager can look at the server logs and data to look for redundant servers. Now this can save companies a lot of money and not just in new customers or revenue streams or executive decisions but within your organization. Now I'm going to be sharing a video around how to build a data-driven company within your business so please keep an eye out for that. But back to our data strategy. We need to work out what is our end goal and how we're going to interpret that data or information that helps with our mission statement we just came up with around trying to make a more intelligent decision based on customer experience. So is that going to be something like reporting where I can look at the data and I can make those decisions based on the trends that I see in front of me? Or is it going to be something like a recommendation model where I can upsell to my customers based on shared characteristics that they have? Or do I need something where I can look into the future and trying to determine the market before it's happened? The truth is a data strategy will cover all of these. It's not just one. You'll probably have a data science function, a reporting function, a forecasting function. These are all part of one data strategy. But to try and keep it simple let's break down a reporting strategy. So on the right hand side I have got my reports and on the left hand side I've got multiple data sources like JSON files, a customer relationship manager, Excel spreadsheets scattered around my organization and system data. And what I need to do is get them from A to B. So the first thing I need to do is gather all that information, transform it and load. Which is called an ETL process. Now this ETL process can cover taking data in, it then looks at standardizing your data sets or cleansing your data sets, removing any personal information or sensitive information and then loading it into a system or a database or some sort of storage area that you'd have. And then finally on top of that you'd have your analytics section. So you would actually pull the data from left to right and be able to store it and look at trends over time. But how do I know the data that I'm going to collect is needed for my end goal? How do I know the ETL process can handle all the information I need before my business does or my business scales? Well my advice is to flip the arrow. So now what I'm doing is looking at the metrics that I need to fulfill my goal. So where are they stored and how often do I need to pull them through? Looking from a metric down approach is really in my experience a better way for data engineers, DBAs and analysts to come up with how they can model the data. So do they need to do aggregation? Do they really need a relational model or is a known non-relational fine? The list goes on. But the main problem that people stumble upon is not knowing what they want to see before they actually see it. So with this approach your teams can have a better understanding of the end state. So I urge you to run sort of mock up sessions of the dashboards that you're looking to see or do whiteboard sessions with the teams. And this kind of brings us to our next point which is probably the two key features of any good data strategy and that is data governance and master data management. So what are they? Let's bring in data governance. Now data governance is the exercise of authority, control and shared decision-making over the management of data assets. And this covers things like security, architecture, operations, deployment and data quality. These are key for having a strong data strategy and it's something that you need to embed into your company and organization early or otherwise things start to get very messy. And on the other side of that, it's not so much messy but very complicated. But the other side of it is master data management. So what is it? Master data management is to define and manage the critical data of an organization to assure a common understanding, consistency, accuracy and control in the ongoing maintenance and application use of the information. That's quite a mouthful but it comes down to reference data and metadata. Note that these two are very different things. They are not the same thing. Most people think that master data management is data governance and vice versa. They are not. Data governance is something that you, is a culture that you create within your company and it's process management and flows. Master data management on the other hand is more around defining what a customer is in data terms. So in sales you may have someone that says a customer is a postcode but somebody in operations say a customer is an IP address. So it's really understanding and getting that consistency around what your data actually means. Another example would be time zones for example. Who has the correct time zone? So if a people in the US or let's say in New York they say the UTC plus four and then in the UK they're UTC. Who's correct? So it's around managing that data and reference data or metadata that goes along with it. These are two key aspects that you're going to need to have a very strong understanding of and to be able to build out. I do have other videos on these so please check them out so they can just have a bit more of an in-depth conversation around them. Now throughout all of this there's one thing I've not mentioned at all and it's a key rule to remember. I've not mentioned technology. Now the reason I've not mentioned technology in this video is because technology is only your enabler. It will only help you get to your end goal but it's not a key factor to it. The key factor that you have are people because the technology is not going to build itself and within 18 months it's going to change in some way or another. You may need to pivot away from technology for whatever reason or pivot to a new technology for commercial benefits. And then when we talk about people within data there's three brackets or three roles or three areas of expertise that I'm going to break this down into. And the first one is engineering. Now your engineering team have the mindset and the responsibility for scalability of your data sets, reliability of your platform and the data quality that you're pushing out. And I include your database analysts, your DBAs for short, in this function as well. Now this is a really good start or a good foundation to any data strategy because it helps with the standardization of data and it helps bring that version of the truth, the single version of the truth out. Then in our next bracket we have our analysts. Now here we're going to be looking for trends, insights and some analysis that's not necessarily business driven but data driven. So looking for those anomalies or things that we can't generally see with the human eye or that we believe that are happening but actually give us some real numbers behind that. And finally we have our data science area. Now this focuses on three P's which is preparation, prediction and perfection. Now preparation is in the form of feature engineering. So data sets for modeling, not features like a red button or a blue button on an application. Prediction is their goal to know what is coming up and why or how did it happen. And then in the end phase is perfection because in the earliest parts of a model it's probably going to be far from perfect but with more data, more features, more business intelligence and more understanding of what's going on the models can become more intelligent over time. So those are the three key areas of expertise that I want you to take away and have a little think about. And I've broken them down in other videos as well. So the key points I want you to take away from this video is the first thing is to understand your end goal or your mission statement early. And to do that you need to break down that end goal very quickly. Now the best way to break down that goal is understand your metrics first. All the features you need for data science because that metric top-down view really helps build the consistency, scalability and reliability that you're going to need for a strong data strategy. One of the key things here is instilling a data governance or a master data management process early because that is going to help you not have a data swamp but a data lake. And there are videos that I've got that are coming up on that as well. Now again this is one of the key points that I made earlier which is remember that technology is only an enabler. Technology will change. People and communication with your team is going to drive success in your data strategy. So thank you for watching. My name is Stuart and I am the data strategist. Please take away what you can from this. Leave your comments in the bottom and also hit the like and subscribe button just so you get up-to-date content from my videos. Thanks very much and I'll speak to you again soon.