Data Ops: A Complete Solution to Everything [Data-Driven]
Thank you to ObservePoint for putting on this wonderful Analytics Summit. I’ve been looking through the sessions, attended one earlier today, and it’s really good information and I’m honored to be presenting at the Analytics Summit.
We’re going to be talking about data operations and the complete solution to everything data-driven. Data operations is a buzz word in the industry these days. You hear it from, not only the digital channels, but you’ll hear it when IBM talks about their digital products or their data products. You’ll hear from every organization and I kind of wanted to put a little bit of a spin on it, about what I see as data operations and what we at MaassMedia see operations as, and why it is important to all organizations.
So before we get started. Just to give you a little background about myself. Kara did give me an introduction and I have a couple other things I just want to discuss with you guys as you get to stare at my beautiful mug on the on the screen there. So I am the V.P. of Analytics and Partner for MaassMedia which is a boutique analytics consulting firm based here in Philadelphia. We work with large. varieties of clients from small nonprofits to Fortune 50 companies and a variety of industries. And it’s it’s been a great time for me to be with MaassMedia and to be in this space. I have 12 years of experience in data collection integration analysis. One of the things we’ll be talking about today is data operations, and specifically, my experience stretches over data strategy data governance process design, multichannel analytics, platform configuration, and data integration architecture.
On a personal note. I’m a huge Philly sports fanatic. I am extremely happy about my Philadelphia Eagles. I am also the father of three wonderful girls ages 9, 6, and 5 they keep me extremely busy when I’m not working in analytics. So let’s go ahead get into the presentation.
So my first question to everybody listening is: how do you achieve clear and actionable analytics insights? Obviously it all starts with the data. Without collecting the correct data that is meaningful for your business, you cannot achieve the insights you need to push your organization forward. You need to have your data agreed to, what they are going to capture agreed upon, you have to ensure the quality, you have to ensure you can integrate with other data sources, and you can make it accessible throughout the organization. These were the themes you’re gonna hear throughout this entire presentation.
However sometimes we focus on getting right to the insights, right to the visualization, and start doing action immediately on the data. This is going to sometimes trivialize the importance of what the data quality is, which is the backbone for doing any of the other exciting things, such as insights of visualization, that we want to do.
Right now we want to achieve everything so fast. So the risk is that we could move beyond the hard part, which is the data quality and go right into the insights, visualization, the action. That’s the sexy part. That’s what is going to make us do well and push our organization forward.But we have to make sure we don’t trivialize the importance of data quality, which is one of the things we’re going to touch on today.
Reports and metrics used for assessing digital initiatives, such as marketing or commerce, sometimes can’t be trusted because of incomplete and inaccurate data. So if we don’t have a data operations strategy in place then making these decisions, given incomplete or inaccurate data, can be dangerous and actually cause harm for your organization. And we’ll talk about why that is coming up.
From a digital perspective, when attempting to integrate digital data, can pollute larger data sets when it’s integrated into other sources like your CRM data, your sales data, your call center records. All the data needs to be as accurate as possible and there should be a process and plan for dealing with these data quality issues. Because what do you do when bad data is collected? You need to have a plan, a process. These are themes and items that you’re going to hear me say over and over again throughout this presentation.
So what leads to poor data quality? Well there are many examples of what can lead to poor data. I worked with small non-profits, all the way up to Fortune 50 companies. All of these organizations have issues in different departments, in different areas. I know that all of you are going to have heard of what leads to this poor data quality: relevant data not being collected or not being collected correctly. That leads support data quality.
Maybe data is being collected but there’s errors in how it’s being fed into your analytics platform independent of which analytics platform you have or use. Maybe you don’t have any governance plan or poor maintenance around the collection of the data, which is going to result in the degradation of your data over time. A lot of companies I talked to only QA the data one time when data is collected they don’t do a continuous program of ensuring that the data is being collected correctly, continually over time. And I can’t stress how important that is especially when we begin integration of data into other data sources.
Perhaps you have a lack of definitions or conventions causing confusion on proper usage of the data. One department is calling this point of data one thing and another is calling it something else. This can cause issues throughout the organization. This is where data operation comes into play.
So how are you doing? Stop and think about that for a second. Let’s think if you’re doing everything correctly. I’ve got some questions I want you to ask yourself as you go through this presentation. 1. Do you have an overarching data strategy throughout the entire organization? 2. Do you have consistent or is there inconsistent data collection practices? Is every department collecting data the same way throughout your entire organization? Are there different groups in your organization perhaps collecting the same data differently? I’ve seen examples of organizations where the product team uses one tool, such as Mixpanel, to collect data on their product, whereas your analytics team is using Adobe Analytics as their source of record for their data. We have to make sure that these different groups are on the same page, and data operations will get us there.
Is there mistrust the accuracy of the data in your organization? How many times has somebody come up to you and you give them a report, they mention, “I don’t believe this data is correct.” Where does this data come from? How are you getting this data? Where did you pull this data from? Because they didn’t believe in what was actually being presented to them. Maybe you’re having difficulty combining data sources, so you get a holistic picture of how your company is doing and how your customers are interacting with your brand. Data operations is going to assist you with that.
And then, are you making this data accessible to all the stakeholders in your organization so they can make day of informed decisions? Or do you have to manually, or sometimes in an automated fashion, send the data out to every single person, all the time? Are you creating a data driven culture where they can access the data on a self-service model when they need to get access to it, so they can get the insights to take the action to run the business and make it more data driven?
So this is where data operations is going to come into play. The operations applies process, governance, and standards to data quality. Beyond that, it gives you a blueprint to integrate your data sources and make all the relevant data accessible to everyone that needs it in order to drive your business forward. There’s an overarching strategy that is going to touch all five pieces of the data operations landscape, from governance, to collection, to quality, to integration, and then eventually accessibility. We’re going to start talking about the data strategy.
This strategy needs to encompass all five aspects of day operations. The first thing that needs to be done is you have to work closely with your stakeholders to identify what their business objectives are. What are their key performance indicators? What’s going to make them succeed in your organization? What data do they currently use? What data do they want to use? That could be completely different. Maybe there is data that they want to get that they’re not currently getting. This is something that you need to work out in the very beginning. Figure out what data they need to run their business. What data are they currently getting? Do they believe that data is of utmost quality? Or do they have issues with the quality of the data? So you want to talk to each individual that is going to be using the data that you can provide.
Then you need to set requirements on how it’s going to be stored, visualized, shared, and analyzed. This strategy is going to go into every single aspect of this data operations. So you need to think of, beforehand, how you’re going to store it how are we going to visualize it. From there you can now determine the gaps. What are the gaps in the data collection? Once you know the gaps, what kind of resources are you going to need in order to collect that data that isn’t being collected? What’s the level of effort? What is your timing going to look like? What tools and platforms are you going to use?
Finally, let’s prioritize all this information so we have a plan, a strategy, going forward on how we’re going to handle our data operations. So this operational plan is going to be based on our five pillars. Data governance, collection quality assurance, data integration, and access. We’re to go into each one of these and talk about them specifically.
First of all we’re going to talk about data governance. It’s the law. A lot of times I see people talk about, “We want to collect the data before govern it.” I disagree with that method. We should have a plan of action on how the data should be formatted, from the availability, the usability, the integrity of the security of the data, before we even go about collecting it. Now a lot of people are going to say to me, “But I’m already collecting the data.” That is correct, but this is going to be a reset. This data operations is meant as a reset for your organization. We’ll talk about how to implement it in a few minutes, but I want you to know the different aspects of it.
You going to need a data standard and dictionary. What does that consist of sort of? Plain language definition of every data point that’s going to be collected, the format of how that data should be collected and stored, and then when and where should it be collected. I can hear people groaning already, “This is a lot of work.” You’re correct, it is a lot of work. And as we talk about implementation of data operations in a future slide, just remember that we’re going to have to start small, but think big. That’s my number one takeaway for you: start small think big.
So in one department we need to figure out what data needs to be collected, formatted, stored, and where and when it should be collected. And this is why it’s very important to start your strategy with understanding what data you need to collect. Because not every piece of data that is that is available to you needs to be collected or is relevant, so you make sure that we’re collecting the relevant data.
You need to create a process for naming, ingesting, and QAing any of the new data sources that are to come. This governance plan should cover every aspect of the rest of the data operations, so governance is going to is going to encompass collection, quality, integration, accessibility. Document everything. All the processes, all you test scripts for doing your data quality, all the standards. Everything needs to be documented so that everybody is on the same page, it’s easily shareable and everybody knows exactly what we’re going to do going forward. Finally you really need to provide training to convey the standards to all those who are involved with both defining the data, collecting the data, and the quality of the data.
So now that we had a data governance plan in place, we’re going to move onto the next part, data collection, where you’re going to configure the data collection based on your business requirements. That doesn’t mean collect every piece of data available to the organization. This means collecting the data that is relevant to your organization and it extends beyond just digital.
Sure there’s going to be digital touch points that need to be collected, but there needs to be a process around all data collection.
When I say all data collection, here’s some examples of sources that you should be collecting data from. Obviously your web site, your mobile app if you have one, your CRM software, your financial software, your marketing automation software, social networks. All of these data sources are going to be collected in order to get that 360 degree view of your customer and of your company. They all need to be collected. They all need to be have a governance plan around them, and then talk about quality integration accessibility.
When we’re talking about digital, I want to make sure that we understand we can leverage systems such as tag management systems and our data layers. Data layers probably the most important aspect of making sure that we keep the governance plan in place. The data layer allows us to have flexibility of what data is going to be collected, when its collective, and gives us a way to ensure that its collected the same time across the board. So make sure that you guys are using tag management systems and data layers, so ensure your consistency, completeness of your data collection, and your digital touch points.
Next let’s talk about data quality. This is going to bring trust to the data. This is one of the things that not many people want to do on an ongoing basis, but it needs to be done. Why? Well according to Forbes, 70 percent of marketers believe they have poor quality or inconsistent customer data. 70 percent. if we have any consistent or poor quality, how can we make decisions based on that data?
You need to develop quality assurance plans and test scripts for all the data is collected and stored throughout the organization. You don’t need to do this from scratch or start by building it in-house. There are tools out there. ObservePoint, a wonderful audit tool that will allow you to ensure proper data collection, run through key use cases, test scripts to validate expected data output throughout all of your digital data. So utilize the tools that are available to you. This will allow you to do ongoing quality assurance, and I can’t stress that enough. This is not a one-and-done thing. You have to continually, QA your data to make sure that it’s being collected properly or you’ll lose trust.
How else can we assure? We can create monthly or quarterly health check reports, so that we can create awareness for the stakeholders on the data quality. These will map directly to use cases to the different areas of data collection that you have set up for digital touch points. And we create a health check saying this is the score we had for this month, this is the score we had for this quarter, going forward, so that organizations all understand the data quality is of the most accurate possible amount. Now that we’ve decided how we’re going to collect the data, format through our government’s plan, we’ve collected our data, we’re ensuring the data quality, the next step is going to be data integration.
And here, we’re going to need to take a step by step approach. According to Forrester, firms only use between 27 and 40 percent of their business intelligence data. So why is that? Well, one, we have data that’s residing in silos, and it’s not integrated throughout the organization. And two, the data being collected isn’t either relevant or accessible. So if we followed our governance plan, we collect the data correctly, we’ve gone through data quality, we’ve already knocked out the second point. We know it’s relevant, now we need to make it accessible and integrated.
So how do we do that? We’re going to need to create a central repository for all of the data. That’s either going to be through a data warehouse that maybe has a built in in-house using your data architects, or you can use a tool such as DOMO, which is a data warehouse in the cloud where you can store all of your data and integrated with different data sources. When we do integrate new data sources, we have to make sure we’re methodically testing these new data sources. That means the integrity of the data repository. The last thing we want to do is have raw data entered the data warehouse that has not gone through the process of governance, collection, and quality.
When we start putting messy data in with all of our other data, we get garbage in, garbage out. In order to get to that to the next step of getting insights and driving your business forward with data, we have to make sure that we are collecting the correct data at the correct quality and integrating that correct data together.
So our last step is data accessibility. This is where we are going to give our organization insight.
It’s only valuable data that is used to inform business decisions. So you want to create a self-service environment that will empower users, the analysts, your executives, to be able to get the insight they need on an on-demand basis, which hopefully will be customized and personalized. Again, this is not the time to reinvent the wheel. Leverage interactive visualization platforms. It doesn’t matter which one, but you want to pick one and stick with it. You have Tableau, DOMO also does a visualization on top of the data warehouse in the cloud, companies like SweetSpot, or Microsoft BI. These are all great interactive visualization platforms that allow you to democratize your data throughout your organization and have the people that need the data to make decisions have it at their fingertips.
This also allows you to have your stakeholders gain trust in that data. The easier it is for them to access the data, to see it on an ongoing basis, and know that what’s going through this data operations strategy to the governance and collection and quality integration, will give them the ability to trust the data easier. So great, I gave you a great blue blueprint of how data operations works, what it’s consisted of, but who’s in charge of it? Simply enough, you’re going to have to have executive buy-in to execute.
The ones that I’ve seen succeed, are the ones that have a Chief Data Officer. This is a new term out in the market. I’ve seen probably a dozen or so organizations have Chief Data Officer positions. And these companies are really pushing the envelope on data operations, really getting it implemented throughout their entire organization. If that person doesn’t exist in your organization, then it’s going to be your highest ranking analytics person, perhaps a VP of analytics or director of analytics, but they’re going to need executive buy-in, and so the highest level you can go, for people that want to understand how important it is to have this data available throughout your organization and to make decisions on it.
Also know your company. I can’t stress this enough, but start small. Get one department or two departments involved. Explain to them the process what needs to be done and then show the results. When you get these small wins you can then explain it to different departments within the organization. However, data operations does touch all of your business, so eventually you’re going to want to have the whole organization on board to get the full impact. But if you start small, get some key small wins, you can push from there.
Small steps, big benefits. Every time you add a piece to the pie of data operations you’re going to create the value. So if you start with the marketing department, you’re only using marketing data, and then you can get the finance department on board with the data operations, and you can combine those two and be more powerful. Then we can combine our CRM, more powerful. We can combine our sales data, our backend sales data, our sales force data our, call center data. Every piece that you add to the data operations pie is going to make it stronger. When all departments are on board, your efficiency is going to increase and you’re going to you’re going to be rocking in the data world.
So this all sounds great in theory, but has anybody actually done this? Yes, of course, and I have a use case for you. So we’ve worked with a global B2B telecom company. When we were brought in, all the data that they had resided in silos. All the data that they needed to run their business correctly, their digital analytics data, their CRM data, their MPS data, and their marketing campaign data. It all came from different areas and was combined in Excel, which then an analyst was given the job of putting out daily, weekly, and monthly reports from this data.
As you can imagine, having to combine this different data from all these different areas, make sure it gets in the correct format, and then gets back out to the people that need to see it was not an easy feat. Some of it was automated, a lot of it was manual, and it took a lot of time and effort. So we came in and recommended to create a data operations strategy. First we needed to create a governance model for the data format and the collection. Then we mapped all the systems where the data was being collected, audited that data to make sure that it was complete, and that we were capturing all the data needs to be captured based on the requirements from the stakeholders. We fed all that data into DOMO in order to integrate that data and visualise it and make it accessible on-demand.
So what were the results of that? The time to produce the reports decrease from 20 hours a week to two hours a week. The stakeholders need a greater awareness of their sales variance, so now they have the data they need to look at their sales trajectory over time and make decisions on the fly to increase sales. And the biggest one is that the analyst can now focus on actual analysis as opposed to just data dumping. What I like to say is, it was a success. You too can own this face of pure accomplishment. One of my favorite memes all time.
So in conclusion, some takeaway points that I want you to focus on from this presentation. Data operations includes everything you need to make data-informed decisions to drive your organization forward. If you have a plan, if you have a strategy, an overarching data operations strategy that includes the governance, the collection, and the data quality, then you can move forward with your integration accessibility, you’re going to win.
However, every piece of this data operations strategy needs to be implemented in order to perform at your best. They all work in conjunction with one another. You can’t collect the data without understanding what needs to be collected from your governance plan. You can’t make the data accessible if it doesn’t have high quality. You can’t integrate the data when it’s dirty, and you can’t integrate the data unless you’re collecting it. So every one of these things needs to work in conjunction with each other for you to do your best.
And then when you’re implementing your data operations strategy, I want you to keep these items in mind. The social political climate of your organization, you need executive buy in to make this work. Don’t reinvent the wheel, use the tools available to you—ObservePoint for auditing and data quality, DOMO for data integration. Do it yourself or data warehouse, but don’t try to reinvent everything. Use the tools that are out there that can help you through this process. Start small. Pick one department, but think big. And as you’re creating your data repository, only put in the data that is clean that you have the utmost respect for.
That’s the end of it. I thank you guys very much for listening to my presentation on data operations. And again, I want to thank you very much for giving me the opportunity to speak today.