Data and analytics should be a core competence for every Chief Information Officer and Information Technology organization. Given the business, technical, and cultural challenges of using data to make informed business decisions, it’s no surprise that data science and analytics are hard.

So, how do you make your enterprise data and analytics program a success? Bruno Aziza, head of data and analytics for Google Cloud, explains his approach with valuable lessons based on practical experience.

The conversation includes these topics:

Bruno Aziza is head of data and analytics for Google Cloud. He specializes in scaling businesses & turning them into global leaders. He helped launch Alpine Data Labs (bought by Tibco), AppStream (bought by Symantec), SiSense (bought Periscope Data) & AtScale. He was at Business Objects when they went IPO (after acquiring Acta & Crystal Reports, & before SAP bought them for $7B). He was at Microsoft when they turned the Data & Analytics business into a $1B giant. Bruno has written 2 books on Data Analytics and Enterprise Performance Management. His allegiance is to the Analytics Community worldwide.

Transcript

Bruno Aziza: This is about looking across all your workload, so if you're using transactional or analytical workloads, how do you bring all of that together, reduce the cost, reduce the maintenance headache, and create an environment where now you can scale technically and also humanly?

Michael Krigsman: Bruno Aziza is the head of data and analytics for Google Cloud.

Bruno Aziza: When I started, not many people cared about databases and back-end issues when it came to data. They felt that it was a necessary evil. Now, organizations are realizing that if they're able to harness data faster than their competition, they really are able to do amazing things.

Great examples that you might be familiar with are things like anomaly detection, fraud analytics, or product recommendations. Think about, in your daily life, you go on a particular website and you're kind of shocked sometimes that a website knows you more than you might know yourself. They're able to recommend amazing content and products that you might not have thought about.

All these systems, all the work that goes behind serving you content that's highly personalized, highly relevant, and makes a great experience for you typically is powered by our technology.

How to choose a meaningful data problem?

Michael Krigsman: One of the significant challenges is choosing the right problem and the right set of data. How do you think about that? How do you go about making those selections?

Bruno Aziza: Think about the problems that are the most related to the business value that is driven for your organization. I think, if you look at the average tenure of the chief data officer, it's about a thousand days, probably a little bit less than a thousand days.

I think the reason for that is often the opportunity for data is so big that you tend to want to do everything. You end up just focusing on the business value, the business metrics. What is driving the bottom line with data? There's a huge opportunity there.

The two areas that we see people not fail but kind of lose their way is when you look at what I call the "why nots." Why not would we look at this use case? That sounds interesting but actually might not lead to a specific value.

Then there are the other use cases that sound interesting because they are highly innovative, but they're really not connected to some of the core issues that your organization is trying to fix.

What I would advise every chief data officer to prioritize their use cases with is not fix the existing or find net new ones. It's double-down on the use cases that your chief financial officer, your CEO, your COO, your CPO (chief product officer) is going to tell you, "I can rally behind that because it's going to drive bottom-line value to my organization." I think this is an important consideration.

Michael Krigsman: This seems easier said than done, and many data analytics folks, data science folks have a real tough time getting into the heart of the right problem. How do you winnow down, and what advice do you have for folks who are struggling with that?

Bruno Aziza: I think about what I call the five S's of the data opportunity. If you break them down, you realize that there are modern issues.

The first S is solve for speed. If you look at the opportunity with data, you're going to live in a world that is real-time. The ability for you to empower your people, drive value for your organization, is about delivering the right data at the right time and using speed as an advantage. That's your first S that you really should focus on.

This idea of discarding real-time infrastructure is really not a good idea. Real-time is now becoming just a must-have for every organization. That's the first S.

The second one is scale. Some organizations that we start working with tell us, "I'm not in the big data space." The news that I have for you is that everybody is.

In fact, we just call it data because if you look at what we're doing here, there's so much data that we are creating. Two-thirds of the data that any organization creates actually never gets analyzed. In fact, the economists, I think, came up with research showing that less than 5% of the data actually is a potential get in front of people that can make decisions on. This idea of building for scale is extremely, extremely important.

The third S is around security. We talked about speed. We talked about scale. Security is really important to build in the very first stages of your data strategy.

More data means more responsibility. We know sometimes people don't think about governance until they've started to kind of build their systems. You need to build with governance from day one.

Now, you don't need to just be focused on governance only. You don't want to be the guardian of data and people fear using data. But you do want to think about centralized governance to enable your people.

The fourth S is around simplicity. I know we'll talk a lot about this today, but Google was certainly late with this issue and this opportunity to deliver simple interfaces into your user community and still get the sophistication of the issue.

Customers that we work with are challenged with, "My environment and my problems are highly sophisticated, but I need an interface that's simple, so I can get adoption." If you're a data scientist, think about that.

Then the only way to get from sophistication to simplicity is my last S. It's what I call the smarts. Artificial intelligence is going to be the secret sauce on how you enable this.

The world of data is growing. We're not going to talk about how the volume of data is just out of control. But you have more data, more people that want it, more use cases, so you need to use artificial intelligence, automation to augment the capabilities you have today so you can deliver data with more simplicity. That's the framework that I think about when I work with organizations.

Michael Krigsman: Can you give us an example of a data problem that meets all of these criteria and tell us why?

Bruno Aziza: We talked about the issue of product recommendation, for instance. It does meet all of that.

For instance, you'd be a customer of, let's call it, a retailer. You're going to have a multi-channel relationship with this retailer.

You're going to go to their store. You're going to visit their site. You're going to visit their partner sites. This ability of thinking at a high level of scale is really important because the data points I'm going to get about Michael are going to be coming from multiple places.

Also, the ability that I have to project an option for you is related to this speed dimension that I talked about. It's not useful, Michael, if I suggest you a pair of shoes after you just went through the cart experience and you've already bought that pair of shoes. It's not really interesting to do that for you.

Also, building a platform that controls the way the data is being used. It's your data. It's not data that should be accessible by anybody else. Having a strong governance backbone to make sure that even though as a retailer I have a relationship with you, I've got to make sure that I've built an infrastructure so your data doesn't get away.

Think about not just your clickstream experience, but also your credit card information and so forth. You've got to think about that dimension.

Then artificial intelligence, like I just said, I can look at your profile and maybe what your clickstream information is, your cohort information, people that are just like you that I've made the right choices. Maybe they've taken advantage of discounts or group products in particular formulas where I can create an interesting experience for you.

Retailers, financial institutions, telecommunications organizations have this huge opportunity to use data and artificial intelligence to create compelling experiences for their customers and that return better value for them in the end.

Hopefully, I touched on all the dimensions that we just talked about here with this one example of product recommendation.

How important is data infrastructure?

Michael Krigsman: How much do you care about and focus on the infrastructure? The issue of prioritizing the business problems is at a pretty high level. Let's go all the way down to the other end, the infrastructure. How important is that and where does that come into play for you?

Bruno Aziza: Extremely important because we're now in a phase where organizations are moving to this modern data stack. You might have heard terms like the data mesh, for instance, an important term that has been popularized over the last few years.

The way to think about it is to step back and look at, what are the issues that your organization is trying to solve? We see, typically, three phases that organizations go through. I'll try to break them down for you.

The first one is what we call the data ocean. Actually, it's not me that calls that. It's customers like Vodafone, for instance, who popularized the term a couple of years ago.

The data ocean idea is that you want to broaden your perspective on where your data is as much as you can. This is capabilities around multi-cloud. Catalog technologies are important here.

This is about looking across all your workloads. If you're using transactional or analytical workloads, how do you bring all of that together, reduce your costs, reduce the maintenance headache, and create an environment where now you can scale technically and also humanly? As you have the opportunity of looking at so much more data, you're not going to hire a lot more people to match that scale, so you want to solve for the data ocean.

The technical infrastructure and the data ocean is different from the other two phases. What are the other two phases? The next one, as we just talked about, is this concept of the data mesh.

The problem you're trying to solve here in the data mesh is often people create data lakes. You hear this terminology saying they've become data swamps. The reason for that is because the data is stored, the data is observed, the data is cataloged, but it's not really acted upon.

What is a data mesh?

The concept of the data mesh is how do I create federated environments so now I can activate my business communities with data analytics environments that are relevant to them while I'm governing data century and that I can make sure that the data that people start with is the data that makes the most sense for the organization. Data mesh is about going from passive to active data. Data fabric technology is important here. And this idea of federation, access to analytics is important.

Finally, the third phase is where you are starting to think about you're evolving to now bringing new personas. In the first two phases, the data architects, the data scientists are important. In this third phase of what we call the data factory (popularized again by McKinsey) is this phase where now you're building data products.

Your chief product officers might come into play here. You might use universal semantic layer technology. It creates data-driven applications so you can repeatedly create products like the ones that we talked about, like product recommendation, anomaly detection, fraud analytics – all data products – all the way to creating your customer data platform so you can really have a 360-view of your customers.

I think the infrastructure is extremely important. What's also important is understanding your level of maturity and how you can accept this technology not just from the technological standpoint but also from an employee maturity standpoint. You're going to have to do a lot of training, a lot of communication to make sure people know what is it that they're trying to achieve.

Michael Krigsman: Again, you're always linking everything, all the technology aspects back to the business problem that you're trying to solve.

Bruno Aziza: It's the business problem and it's also the organizational footprint. The reason for why I really like this idea of the data mesh is because it describes what you're trying to achieve, how you should line up your infrastructure, but also how you should think about your organization.

We have lots of organizations that are asking, "Hey, should my data people report centrally into one organization? Should they be distributed?"

When you have these goals of the data ocean, the data mesh, and the data factory, you start now thinking, "How do I align my organizational footprint to best serve my business goals?"

I think those are important considerations.

As a vendor, I know you're probably expecting me to talk to you about products all day long, but I really think that the success of the data analytics and the data platform strategy is highly bound by your ability to galvanize your people, train them, and get them to work with you on achieving your business goals.

Michael Krigsman: Let's take a few questions from Twitter. First, we have Chris Peterson who asks, "In terms of data security and privacy, how does Google navigate the maze of different regulations internally? How do you as a company manage this stuff?"

Bruno Aziza: We internally probably have one of the most secure platforms for just managing data. You can look up this technology called Access Transparency and so forth. We also have specific industry teams that normally are working closely with customers on these issues, but also very versed with the issues for every specific industry.

I would say we build the security by design. Just like I was advising any company to think about it for their own data platform, that's what we've been doing on our own platform.

In fact, a lot of the issues that we're helping customers with today are issues that we have solved for ourselves. I think that's probably one of our competitive advantages (if I could talk about that). We relate very well with what is it like to create compelling experiences for the future of the analytics consumer.

If you think about your own organization, you're going to want people to consume information the same way that they go to Google.com today: A simple interface. It doesn't require any training. Provides high-level sophistication, high-level personalization, but still through an experience that's highly simple.

How can small organizations take advantage of data and analytics?

Michael Krigsman: Another question that's just come up on Twitter from Arsalan Khan relates to this complexity, to another aspect of this complexity. Arslan says, "Collecting this kind of data at this scale has real cost." He's wondering what smaller organizations can do to take advantage of data despite the costs.

Bruno Aziza: I forgot to mention that one of the important considerations is this domain of financial operations, particularly in the data mesh world. In the data mesh world, if you can imagine, now you have centralized data governance. You're creating your data hubs and your data neighborhoods (as some of the customers that we work with call them).

People start consuming and driving some compute costs, so how do you manage that? It's important that you think about choosing your platform that has flexible financial ops options.

What that means is we can create reservations. You could say, "Well, we're going to spending up to this cap. We don't spend more than that." Or you can allocate particular compute capacity to specific workloads that you can either predict or that you can kind of give a range to.

What I would say are two things. One is, there's the management of the costs, but there's also, when costs increase, it's not always bad news. It's also that your people are actually engaged, and they are actually using the platform to drive value.

I would not just look at costs by itself. I would look at price and performance relationships, and I would look at costs and value relationships because, if you look at our industry today – we've been working on it for the last 30 years – the adoption rates of technologies are very, very low. In the business intelligence space, we're talking about 30% adoption. In the AI space, we're talking about 35% adoption.

In our case, I look at machine learning that is deployed through BigQuery. We're seeing 80% of our top 100 BigQuery customers use artificial intelligence. What that means is that they are getting value. They're getting to usability a lot faster, and that's not always bad news for your organization.

Michael Krigsman: This seems one of the core issues that chief information officers must grapple with because the mandate for many CIOs is to innovate and, at the same time, do more with less, right? We want you to be the driver of innovation but do it with less cost.

Bruno Aziza: It's a huge challenge, but it's also a great opportunity. Like I was just saying a little earlier, you don't want to be the executive who restricts access to data. You don't want to be the executive that slows down innovation. You want to be the executive who is lining up to the business objectives of the organization and provides a platform that is driving innovation.

If you think about it, innovation in any organization is going to come from the front-line folks, the folks that are in closer contact with the customers, and so forth. Enabling that model – that's why I keep going back to the data mesh, I think it's a great model to follow – is to really think about business goals, technology stack, and organizational structure.

Often, customers ask us, "Where should my analytics folks report under?" You've got to think about how you organize yourselves so you can get to innovation faster than any other organization.

I think, today, that's really the issue that we see is people do a lot of POCs (proofs of concept), but they're having an issue getting into production and then innovate on top of that production. We want to simplify that, and we want to get to a more liquid (if you will) relationship with your data.

Should the CIO own data and analytics?

Michael Krigsman: It's very interesting that when you talk about the data mesh, you talk about the business goals, the technology, and then the organizational structure. Why is the organizational structure so crucial?

Bruno Aziza: Things get done through people, and they get done through people that have shared goals. I could give you the best technology, but if it's deployed in the wrong system, it can't really help you.

We did a survey a few months ago where we asked, "Where should your data analytics powerhouse be?" We asked people, "Should it be under your CTO? Should it be under your chief product officer? Should it be under the CFO? Should it be under the CMO (chief marketing officer)?"

What we found is, one, the answer kind of depends, but it also is related to the types of executives that you have.

About ten years ago, I wrote a book called Drive Business Performance, and it was based on interviews of organizations that had experience, amazing success at driving a data culture inside the organization. The key was finding the right executive and getting the mandates into how we're going to make decisions in our organization. You don't want to discard that.

Sometimes, people look at their initiatives and say, "Oh, I just have sponsorship," and the sponsorship is good. It's necessary, but the mandate from your top executive is saying, "We will now make decisions based on data. We will now go out and look for opportunities to measure things that maybe we couldn't measure before. But because we know the business needs it, we're going to do it."

That's critically important and it's way more important, I'd say, than any of the nicest and latest technology you can acquire. If you don't have the organizational footprint, if you don't have the mandate from the CEO, you've got a great Ferrari, but you don't really have the keys to it, so why turn the engine on?

Michael Krigsman: You know it's kind of funny. What you're describing is almost a cliché and almost extremely obvious that we need to, again, align the work that we're doing to solve the problems that we think are important. But why is it so difficult to actually achieve that very simple goal?

Bruno Aziza: There are a few reasons for it. I think one is we live in the time where there's a lot of innovation, there are a lot of buzzwords, and there are a lot of vendors.

If you look at the data landscape that was just published, it's something I wrote about in my Forbes column, there is a high proliferation of solutions. There's a lot of innovation in technology. The cloud is a lot more available than it might have been just simply ten years ago.

I think there's a flurry of options being thrown in front of CIOs or chief data offers. It's really hard to kind of parse through that noise. Sometimes, as technologists (and I'm a technologist myself), we might get enamored by, "Oh, this is a cool concept. What if I deployed this?" I think this makes it a little bit hard for leaders.

Secondly is that chief data officers don't think about themselves as business leaders. They think about themselves technically as technical leaders.

What we work with organizations with is, when you drive your initiatives, do you have a brand for your initiative. Do you have a communication plan for your initiative?

As technical leaders, we really think that there's a part of marketing the solution, if you will, back into the organization that matters to succeeding. I think the combination of those explains why there's not a lot of success.

If you look at the latest research from Accenture, I think it's 68% of organizations can't find value from the data that they have. That's a huge number.

Also, it's just simply difficult for chief data officers to stay in place. I think I said earlier, I think the average tenure of the chief data officer is less than a thousand days. Primarily, it's related because we found that they rarely tie their technical initiatives to business goals and they might not think about the communication of those business goals. And so, that kind of hurts them in succeeding with their data strategy.

Michael Krigsman: All of this begs the question, who should own these data science efforts? When you say that there's a disconnect between the technology problem that is being solved and the business requirement, there has to be a cause. Who should own these efforts?

Bruno Aziza: What we found in some of the surveys that we did was primary two executives that this falls under. The First one, interestingly enough, is the CFO.

When we asked our community to tell us, "Where do you think data science and analytics should roll under?" a large percentage, 34% (not the majority) of people said primarily the CFO. The reason for that is because, I think, over the years, the CFOs have gone more from the back-office, cost retention type of role to innovating, using data as a way to power their organization and drive the operations of the organization. Data analytics and data science have a great opportunity to do that.

The CFO might actually be. Again, it depends on their objective. And so, you have to decide for your own organization. Actually, it might be a good group to own this.

The second one has been the CTO, the chief technology officer. Again, it depends on their style, their team, and so forth. I think the reason for that is because you want, on your bench, a good amount of very technical folks. We have seen, over the years, the data scientist was the sexiest data role. Now it's going to the machine learning engineer.

You think, wow, it's getting more and more technical. I think the reason for that is because the industry is innovating really, really fast. And so, you want technically savvy folks to enable you to deploy, but you want to couple them with your business folks.

It's probably harder to learn the business coming from a technology background than the opposite (at the moment) because we're now building technology that is taking over a lot of the tough steps that you might have needed to learn. I think about auto data preparation or auto data quality and all these steps that now business analysts can come in and start using.

What I would say is, what we see organizations do is they look at their blueprint. Say you have 100 people to handle all the data analytics issues. They tend to put half in the central business unit and then half into a central corporate unit like the CTO's organization or the CFO's organization.

Chief financial officer as the data owner?

Michael Krigsman: You said that CFOs could be the right folks to manage or be responsible for these data efforts. But my question is this. CFOs, in general, may understand technology but certainly, by and large, don't have the kind of deep expertise that's required from a technology standpoint. And so, how is it practical for a CFO to manage this? By the way, why didn't you say that the CIO should be responsible?

Bruno Aziza: There are two aspects to the answer to this question. Why not the CIO is because, at least from what we're hearing from our customers, typically if your CIO is focused on internal technology and infrastructure choices, data and analytics tend to be an application business.

We don't think about this today, but they are a business application consumed and directed towards value creation. I think that's why the CFO comes in here is because modern CFOs don't think about just budget and reducing costs. They think about opportunities for us to create value.

A great example of that is now organizations creating data products that they will monetize. I think about retailers, for instance, one of the great organizations I'm working with.

Carrefour is one of the largest retailers in the world. They've solved their data mesh issues for themselves and now have built around it, and they are now starting to create data products that they can sell back into their community, which is now talking about driving revenue for the organization.

I think the mindset of the most innovative organizations is that data is not a liability. Data is not something that I guard only. Data is something I build upon. It actually becomes an asset for me to manage up to a point where I can create products off of it and monetize these products. I think that's why customers are saying what they're saying around where it should fit.

Michael Krigsman: We have another interesting question from Arsalan Khan on an aspect of this topic. He says, "As organizations become increasingly reliant on AI and machine learning for decision-making, are some executives resistant to accepting data as the ultimate decision-maker?" In other words, if I can rephrase it, what should folks do if a business leader rejects the conclusions that the data presents?

"I don't think this is right. Yeah, sure. Your data says whatever, but I know from my experience it can't be right."

Bruno Aziza: This is the typical gut feeling that we deal with. The issue with a gut feeling is you never know if it's actual experience or if it's indigestion. You don't want to just rely on gut feel, but it is true that if you read a theory on this and books from Malcolm Gladwell and other folks that are very educated and well researched, the right decision is going to come from the combination of really good data and experience around the mistakes that maybe you've made or maybe others have made that you've been able to learn from.

I think, in general, it's never a good idea to decide 100% on your gut feel. You might get lucky every once in a while, but now we have technology that captures enough that you are able to not just understand but, in many cases, predict. And there many great stories like this in baseball, in the wine industry, and others like this that we can all relate to.

I think you're always going to get into a conversation with an executive that maybe might not believe the insights that you're bringing in. This is why, in the last few years, you might have seen the work from Nancy Duarte on storytelling, so connecting with the emotional aspects of how this executive might relate to the data.

In the book that we wrote a few years ago, Drive Business Performance, we talked about the example of Lego where the data analyst not only presented the information, the dashboards, but they actually had the voicemail left by the kids being played to the executive. The executive, as a parent, could relate to the customer feedback they would get and actually did change the strategy.

What I would advise our friend Arsalan here asking the question is don't think about just the binary logic aspect on how you're delivering the data. Think about the emotional aspect, the way people make decisions, even executives with great experiences, how they emotional connect with the data. That's really important as well to build into how you present your results to folks you're trying to convince.

Chief Digital Officer as the data owner?

Michael Krigsman: We have two interesting questions from LinkedIn. This is from Prashant Motewar. He says, "Number one, what about the chief digital officer as the owner of data and analytics? What do you think about that?"

Bruno Aziza: Given the past couple of years here where digitalization has really accelerated, we see certainly in organizations that would gather a lot of their information from physical locations – retailers and financial industries – nobody goes to the branches and nobody goes to the stores, so the person that's in charge of digitalization and taking this system into the future is certainly going to be interested in collecting and understanding data a lot faster.

What I would say, though, is it's not just the title that you have to look at it. It's also the organizational footprint. The people under this leader, is it the right talent? Is this the right organizational footprint? Do you have shared goals?

One of the important best practices that we see is that it's not just the CTOs job to innovate with data. It's the rest of the organization.

We worked with CIOs, CFOs, CTOs who share business goals that they actually don't have the direct impact into it, but they also have shared goals with the business folks who do not have direct impact on the metric itself. But the point here is to get them to get together, align, and collaborate.

The chief data officer and chief digital officer are great roles, but I wouldn't just be wedded to the title. I would look deeper into the organizational footprint of that organization.

Michael Krigsman: Another excellent question from Prashant, an important question. He says, "Data insights are extremely valuable when delivered at the right time to the right people with the right context. Any point of view on how to enable this?"

Bruno Aziza: A few years ago, I did a keynote at a data summit. I came up with this acronym, and not a very pleasant acronym but at least memorable, called RAT. R-A-T because data needs to be relevant, actionable, and timely.

You're absolutely right in your analysis. If I told you here's an umbrella because it rained yesterday, it's not very helpful.

There are a few best practices here. The first one is actual data literacy across the organization. It's one thing to deliver the data, but it's also another thing for people to actually understand the data. We also did a survey on how many data employees should you have inside your organization, so when you deliver the data, people understand what to do about it.

Keron Bourne, who is a member of the community, had (I think) the best answer where he said, "100% of your employees should be data literate." What that means is they should be able to recognize, they should be able to understand, and they should be able to talk data. I would make sure that at least all your employees understand the opportunity they have with using data so that when they get it in their context, they can use it.

Then a third of the organization should be data fluent. What does that mean? That means they should be able to analyze. They should be able to create arguments. They should be able to present results visually, emotionally to their management, to their peers.

Then 10% of your organization should be data professionals. A data pro here is someone who is paid to create value from assets.

The reason for why I'm saying this balance of roles matter is because the issue sometimes is you might be presenting data in the right context but to people that might not know what to do with it.

What we tend to forget is we're not in the business of building folks to become data specialists. We're in the business – you're in the business of doing your business. And so, to be able to be equipped with that, you have to deliver the data on time to an audience that is willing to or is equipped to act on that data. That's the most important thing is how do you act on that data.

Michael Krigsman: We have another question now from Twitter from Lisbeth Shaw who says, "In a busy company, should the person in charge of data science repurpose their old BI data strategy for today's data science needs? How should they do that?

Bruno Aziza: Too often, we feel like we have to hire outside folks to come and solve a problem because the other folks have a data scientist title and my folks don't have a data scientist. What I would say, I just talked to a chief data officer today who is looking at how she is going to upskill her team, and she is absolutely starting with the existing talent because institutional knowledge of your organization, knowledge of your customers, the knowledge of your organizational process is critically important.

Yes, you can bring outside talent that is technically gifted and so forth, but you'll never be able to hire enough of them so you can tackle the problems that you need to tackle and often in a very timely manner. The reality is that there are toolsets now that enable business analysts to step up into a data scientist type of role, and so I would never discard your existing talent. I am sure many of them are capable of doing more, either because they are motivated to get the training or because the toolset that is presented to them is making access to data and working with data a lot simpler.

I'll just give you an example on our data stack with this product called BigQuery. We have this embedded machine learning capability inside BigQuery, which means that you don't have to move the data. You don't have to set up an infrastructure. You can trigger models and run them with just a few lines of SQL.

Just these few lines make business analysts able to do work that in other platforms would require a lot of code from a machine learning engineer. And so, I think the good news here is that the industry – cloud vendors like us and the rest of the ecosystem – is really driving to making tools easier to use, which means (for you) you can use your people and upskill them into where you want them to be.

Value of low-code and no-code products to CIOs

Michael Krigsman: Certainly, when I talk with CIOs, the idea of low code, no code products is right, front, and center of how you can help your organization innovate while reducing costs.

Bruno Aziza: Absolutely. No code and no code is probably one of the most disruptive trends of 2021. It's primarily because it's enabling business users to create business purpose, domain-specific applications for themselves. I believe that the world of package applications is going to be disrupted over the next ten years because now these business users are able to just solve a problem by bringing in services, kind of in a composable manner, to create an application that's relevant to them and their community.

Definitely, the tools are getting easier. People are getting more skilled. There's more of an acute need for working with more data. I think the combination of those things doesn't mean that you should look for the answer outside of your organization. The answer often starts with you.

Michael Krigsman: We have another question again from LinkedIn. This is from Scott Beliveau. He asks, "What's the next data thing that organizations are not talking about and looking at but should be?"

Bruno Aziza: I look at a lot of technology. Recently, I bought myself these glasses, these smart glasses that I can talk to and they can do things for me and so forth. I think what we're not talking about enough is what is the future of the interface into data.

Today, we're used to our keyboards and we're used to our phones. But the reality is that, in the future – and not too distant of a future, I believe – our voice, our eyes are going to be how we interact with information.

There's a lot of investment going into natural language technology. It's a huge field around natural language processing, natural language understanding.

It reminds me back when I was at Microsoft where we shipped the Xbox. The Xbox had this amazing camera called Kinect. It was one of the most popular devices. It was this camera that would scan my body and the tagline was, "You are the controller."

I think that's where we're going with data. Ultimately, to broaden the appeal of data and make data more engaging for more people, we've got to change the relationship.

We're now in the mode where we have to talk machine. I think, in the next ten years, the machine is going to talk human, and that's going to empower new use cases, new experiences at home with smart home devices that you can call upon to set up your alarm or get answers and so forth. I think that's what's going to happen in the corporate context as well.

Michael Krigsman: Bruno, we're just about out of time. Any final thoughts or words of advice to folks who are listening regarding being successful with the data, aligning data efforts with the business, the things that are really hard?

Bruno Aziza: I do have one thing that I think is important for folks to consider is this whole domain of data culture.

One of my good friends Randy Bean just published a book. I think he calls it Fail Fast, Learn Faster. I might be misquoting the title, but I'll send you the exact link.

In it, he's interviewed data leaders and asked, what is getting in the way of being successful with data? It turns out that the majority of them are saying it's the data culture.

What I did is create my data culture checklist. We won't be able to go through all ten of them, but I'll give you a couple of best practices.

The first one, we talked about branding. Brand your initiative. Think about it seriously. Create a logo. Have certified data, so people recognize the quality of your data. It's really important in making sure that you take data seriously at your organization.

The second is have what I call a decision inspection. What I mean by that is often we do a postmortem. Often, we look at failure, and we say, "Okay. Let's try and understand how we failed."

Two best practices: Are you doing premortem? Premortem is, if everything goes wrong, what does that look like? Are we ready to react to a situation when it's going to go wrong?

Then the second practice is, analyze your successes. Often, you succeed and you move on, but do you truly understand why something really succeeded? Having this decision inspection mentality, I know it's hard because everybody is moving really fast. It's hard to prioritize. It is the best way for you to build this data culture that ultimately is going to make you one of the most innovative organizations in the industry.

Michael Krigsman: Bruno Aziza, thank you so much for taking time to speak with us today. It's been an action-packed 45 minutes. Really, really appreciate it.

Bruno Aziza: Thanks for having me, Michael.

Michael Krigsman: Everybody, thank you for watching, especially those folks who ask such excellent questions. Now, before you go, please subscribe to our YouTube channel, hit the subscribe button at the top of our website so that we can send you our newsletter and you'll get notified of our amazing upcoming shows, check out CXOTalk.com, and tell a friend.

Bruno Aziza: This is about looking across all your workload, so if you're using transactional or analytical workloads, how do you bring all of that together, reduce the cost, reduce the maintenance headache, and create an environment where now you can scale technically and also humanly?

Michael Krigsman: Bruno Aziza is the head of data and analytics for Google Cloud.

Bruno Aziza: When I started, not many people cared about databases and back-end issues when it came to data. They felt that it was a necessary evil. Now, organizations are realizing that if they're able to harness data faster than their competition, they really are able to do amazing things.

Great examples that you might be familiar with are things like anomaly detection, fraud analytics, or product recommendations. Think about, in your daily life, you go on a particular website and you're kind of shocked sometimes that a website knows you more than you might know yourself. They're able to recommend amazing content and products that you might not have thought about.

All these systems, all the work that goes behind serving you content that's highly personalized, highly relevant, and makes a great experience for you typically is powered by our technology.

How to choose a meaningful data problem?

Michael Krigsman: One of the significant challenges is choosing the right problem and the right set of data. How do you think about that? How do you go about making those selections?

Bruno Aziza: Think about the problems that are the most related to the business value that is driven for your organization. I think, if you look at the average tenure of the chief data officer, it's about a thousand days, probably a little bit less than a thousand days.

I think the reason for that is often the opportunity for data is so big that you tend to want to do everything. You end up just focusing on the business value, the business metrics. What is driving the bottom line with data? There's a huge opportunity there.

The two areas that we see people not fail but kind of lose their way is when you look at what I call the "why nots." Why not would we look at this use case? That sounds interesting but actually might not lead to a specific value.

Then there are the other use cases that sound interesting because they are highly innovative, but they're really not connected to some of the core issues that your organization is trying to fix.

What I would advise every chief data officer to prioritize their use cases with is not fix the existing or find net new ones. It's double-down on the use cases that your chief financial officer, your CEO, your COO, your CPO (chief product officer) is going to tell you, "I can rally behind that because it's going to drive bottom-line value to my organization." I think this is an important consideration.

Michael Krigsman: This seems easier said than done, and many data analytics folks, data science folks have a real tough time getting into the heart of the right problem. How do you winnow down, and what advice do you have for folks who are struggling with that?

Bruno Aziza: I think about what I call the five S's of the data opportunity. If you break them down, you realize that there are modern issues.

The first S is solve for speed. If you look at the opportunity with data, you're going to live in a world that is real-time. The ability for you to empower your people, drive value for your organization, is about delivering the right data at the right time and using speed as an advantage. That's your first S that you really should focus on.

This idea of discarding real-time infrastructure is really not a good idea. Real-time is now becoming just a must-have for every organization. That's the first S.

The second one is scale. Some organizations that we start working with tell us, "I'm not in the big data space." The news that I have for you is that everybody is.

In fact, we just call it data because if you look at what we're doing here, there's so much data that we are creating. Two-thirds of the data that any organization creates actually never gets analyzed. In fact, the economists, I think, came up with research showing that less than 5% of the data actually is a potential get in front of people that can make decisions on. This idea of building for scale is extremely, extremely important.

The third S is around security. We talked about speed. We talked about scale. Security is really important to build in the very first stages of your data strategy.

More data means more responsibility. We know sometimes people don't think about governance until they've started to kind of build their systems. You need to build with governance from day one.

Now, you don't need to just be focused on governance only. You don't want to be the guardian of data and people fear using data. But you do want to think about centralized governance to enable your people.

The fourth S is around simplicity. I know we'll talk a lot about this today, but Google was certainly late with this issue and this opportunity to deliver simple interfaces into your user community and still get the sophistication of the issue.

Customers that we work with are challenged with, "My environment and my problems are highly sophisticated, but I need an interface that's simple, so I can get adoption." If you're a data scientist, think about that.

Then the only way to get from sophistication to simplicity is my last S. It's what I call the smarts. Artificial intelligence is going to be the secret sauce on how you enable this.

The world of data is growing. We're not going to talk about how the volume of data is just out of control. But you have more data, more people that want it, more use cases, so you need to use artificial intelligence, automation to augment the capabilities you have today so you can deliver data with more simplicity. That's the framework that I think about when I work with organizations.

Michael Krigsman: Can you give us an example of a data problem that meets all of these criteria and tell us why?

Bruno Aziza: We talked about the issue of product recommendation, for instance. It does meet all of that.

For instance, you'd be a customer of, let's call it, a retailer. You're going to have a multi-channel relationship with this retailer.

You're going to go to their store. You're going to visit their site. You're going to visit their partner sites. This ability of thinking at a high level of scale is really important because the data points I'm going to get about Michael are going to be coming from multiple places.

Also, the ability that I have to project an option for you is related to this speed dimension that I talked about. It's not useful, Michael, if I suggest you a pair of shoes after you just went through the cart experience and you've already bought that pair of shoes. It's not really interesting to do that for you.

Also, building a platform that controls the way the data is being used. It's your data. It's not data that should be accessible by anybody else. Having a strong governance backbone to make sure that even though as a retailer I have a relationship with you, I've got to make sure that I've built an infrastructure so your data doesn't get away.

Think about not just your clickstream experience, but also your credit card information and so forth. You've got to think about that dimension.

Then artificial intelligence, like I just said, I can look at your profile and maybe what your clickstream information is, your cohort information, people that are just like you that I've made the right choices. Maybe they've taken advantage of discounts or group products in particular formulas where I can create an interesting experience for you.

Retailers, financial institutions, telecommunications organizations have this huge opportunity to use data and artificial intelligence to create compelling experiences for their customers and that return better value for them in the end.

Hopefully, I touched on all the dimensions that we just talked about here with this one example of product recommendation.

How important is data infrastructure?

Michael Krigsman: How much do you care about and focus on the infrastructure? The issue of prioritizing the business problems is at a pretty high level. Let's go all the way down to the other end, the infrastructure. How important is that and where does that come into play for you?

Bruno Aziza: Extremely important because we're now in a phase where organizations are moving to this modern data stack. You might have heard terms like the data mesh, for instance, an important term that has been popularized over the last few years.

The way to think about it is to step back and look at, what are the issues that your organization is trying to solve? We see, typically, three phases that organizations go through. I'll try to break them down for you.

The first one is what we call the data ocean. Actually, it's not me that calls that. It's customers like Vodafone, for instance, who popularized the term a couple of years ago.

The data ocean idea is that you want to broaden your perspective on where your data is as much as you can. This is capabilities around multi-cloud. Catalog technologies are important here.

This is about looking across all your workloads. If you're using transactional or analytical workloads, how do you bring all of that together, reduce your costs, reduce the maintenance headache, and create an environment where now you can scale technically and also humanly? As you have the opportunity of looking at so much more data, you're not going to hire a lot more people to match that scale, so you want to solve for the data ocean.

The technical infrastructure and the data ocean is different from the other two phases. What are the other two phases? The next one, as we just talked about, is this concept of the data mesh.

The problem you're trying to solve here in the data mesh is often people create data lakes. You hear this terminology saying they've become data swamps. The reason for that is because the data is stored, the data is observed, the data is cataloged, but it's not really acted upon.

What is a data mesh?

The concept of the data mesh is how do I create federated environments so now I can activate my business communities with data analytics environments that are relevant to them while I'm governing data century and that I can make sure that the data that people start with is the data that makes the most sense for the organization. Data mesh is about going from passive to active data. Data fabric technology is important here. And this idea of federation, access to analytics is important.

Finally, the third phase is where you are starting to think about you're evolving to now bringing new personas. In the first two phases, the data architects, the data scientists are important. In this third phase of what we call the data factory (popularized again by McKinsey) is this phase where now you're building data products.

Your chief product officers might come into play here. You might use universal semantic layer technology. It creates data-driven applications so you can repeatedly create products like the ones that we talked about, like product recommendation, anomaly detection, fraud analytics – all data products – all the way to creating your customer data platform so you can really have a 360-view of your customers.

I think the infrastructure is extremely important. What's also important is understanding your level of maturity and how you can accept this technology not just from the technological standpoint but also from an employee maturity standpoint. You're going to have to do a lot of training, a lot of communication to make sure people know what is it that they're trying to achieve.

Michael Krigsman: Again, you're always linking everything, all the technology aspects back to the business problem that you're trying to solve.

Bruno Aziza: It's the business problem and it's also the organizational footprint. The reason for why I really like this idea of the data mesh is because it describes what you're trying to achieve, how you should line up your infrastructure, but also how you should think about your organization.

We have lots of organizations that are asking, "Hey, should my data people report centrally into one organization? Should they be distributed?"

When you have these goals of the data ocean, the data mesh, and the data factory, you start now thinking, "How do I align my organizational footprint to best serve my business goals?"

I think those are important considerations.

As a vendor, I know you're probably expecting me to talk to you about products all day long, but I really think that the success of the data analytics and the data platform strategy is highly bound by your ability to galvanize your people, train them, and get them to work with you on achieving your business goals.

Michael Krigsman: Let's take a few questions from Twitter. First, we have Chris Peterson who asks, "In terms of data security and privacy, how does Google navigate the maze of different regulations internally? How do you as a company manage this stuff?"

Bruno Aziza: We internally probably have one of the most secure platforms for just managing data. You can look up this technology called Access Transparency and so forth. We also have specific industry teams that normally are working closely with customers on these issues, but also very versed with the issues for every specific industry.

I would say we build the security by design. Just like I was advising any company to think about it for their own data platform, that's what we've been doing on our own platform.

In fact, a lot of the issues that we're helping customers with today are issues that we have solved for ourselves. I think that's probably one of our competitive advantages (if I could talk about that). We relate very well with what is it like to create compelling experiences for the future of the analytics consumer.

If you think about your own organization, you're going to want people to consume information the same way that they go to Google.com today: A simple interface. It doesn't require any training. Provides high-level sophistication, high-level personalization, but still through an experience that's highly simple.

How can small organizations take advantage of data and analytics?

Michael Krigsman: Another question that's just come up on Twitter from Arsalan Khan relates to this complexity, to another aspect of this complexity. Arslan says, "Collecting this kind of data at this scale has real cost." He's wondering what smaller organizations can do to take advantage of data despite the costs.

Bruno Aziza: I forgot to mention that one of the important considerations is this domain of financial operations, particularly in the data mesh world. In the data mesh world, if you can imagine, now you have centralized data governance. You're creating your data hubs and your data neighborhoods (as some of the customers that we work with call them).

People start consuming and driving some compute costs, so how do you manage that? It's important that you think about choosing your platform that has flexible financial ops options.

What that means is we can create reservations. You could say, "Well, we're going to spending up to this cap. We don't spend more than that." Or you can allocate particular compute capacity to specific workloads that you can either predict or that you can kind of give a range to.

What I would say are two things. One is, there's the management of the costs, but there's also, when costs increase, it's not always bad news. It's also that your people are actually engaged, and they are actually using the platform to drive value.

I would not just look at costs by itself. I would look at price and performance relationships, and I would look at costs and value relationships because, if you look at our industry today – we've been working on it for the last 30 years – the adoption rates of technologies are very, very low. In the business intelligence space, we're talking about 30% adoption. In the AI space, we're talking about 35% adoption.

In our case, I look at machine learning that is deployed through BigQuery. We're seeing 80% of our top 100 BigQuery customers use artificial intelligence. What that means is that they are getting value. They're getting to usability a lot faster, and that's not always bad news for your organization.

Michael Krigsman: This seems one of the core issues that chief information officers must grapple with because the mandate for many CIOs is to innovate and, at the same time, do more with less, right? We want you to be the driver of innovation but do it with less cost.

Bruno Aziza: It's a huge challenge, but it's also a great opportunity. Like I was just saying a little earlier, you don't want to be the executive who restricts access to data. You don't want to be the executive that slows down innovation. You want to be the executive who is lining up to the business objectives of the organization and provides a platform that is driving innovation.

If you think about it, innovation in any organization is going to come from the front-line folks, the folks that are in closer contact with the customers, and so forth. Enabling that model – that's why I keep going back to the data mesh, I think it's a great model to follow – is to really think about business goals, technology stack, and organizational structure.

Often, customers ask us, "Where should my analytics folks report under?" You've got to think about how you organize yourselves so you can get to innovation faster than any other organization.

I think, today, that's really the issue that we see is people do a lot of POCs (proofs of concept), but they're having an issue getting into production and then innovate on top of that production. We want to simplify that, and we want to get to a more liquid (if you will) relationship with your data.

Should the CIO own data and analytics?

Michael Krigsman: It's very interesting that when you talk about the data mesh, you talk about the business goals, the technology, and then the organizational structure. Why is the organizational structure so crucial?

Bruno Aziza: Things get done through people, and they get done through people that have shared goals. I could give you the best technology, but if it's deployed in the wrong system, it can't really help you.

We did a survey a few months ago where we asked, "Where should your data analytics powerhouse be?" We asked people, "Should it be under your CTO? Should it be under your chief product officer? Should it be under the CFO? Should it be under the CMO (chief marketing officer)?"

What we found is, one, the answer kind of depends, but it also is related to the types of executives that you have.

About ten years ago, I wrote a book called Drive Business Performance, and it was based on interviews of organizations that had experience, amazing success at driving a data culture inside the organization. The key was finding the right executive and getting the mandates into how we're going to make decisions in our organization. You don't want to discard that.

Sometimes, people look at their initiatives and say, "Oh, I just have sponsorship," and the sponsorship is good. It's necessary, but the mandate from your top executive is saying, "We will now make decisions based on data. We will now go out and look for opportunities to measure things that maybe we couldn't measure before. But because we know the business needs it, we're going to do it."

That's critically important and it's way more important, I'd say, than any of the nicest and latest technology you can acquire. If you don't have the organizational footprint, if you don't have the mandate from the CEO, you've got a great Ferrari, but you don't really have the keys to it, so why turn the engine on?

Michael Krigsman: You know it's kind of funny. What you're describing is almost a cliché and almost extremely obvious that we need to, again, align the work that we're doing to solve the problems that we think are important. But why is it so difficult to actually achieve that very simple goal?

Bruno Aziza: There are a few reasons for it. I think one is we live in the time where there's a lot of innovation, there are a lot of buzzwords, and there are a lot of vendors.

If you look at the data landscape that was just published, it's something I wrote about in my Forbes column, there is a high proliferation of solutions. There's a lot of innovation in technology. The cloud is a lot more available than it might have been just simply ten years ago.

I think there's a flurry of options being thrown in front of CIOs or chief data offers. It's really hard to kind of parse through that noise. Sometimes, as technologists (and I'm a technologist myself), we might get enamored by, "Oh, this is a cool concept. What if I deployed this?" I think this makes it a little bit hard for leaders.

Secondly is that chief data officers don't think about themselves as business leaders. They think about themselves technically as technical leaders.

What we work with organizations with is, when you drive your initiatives, do you have a brand for your initiative. Do you have a communication plan for your initiative?

As technical leaders, we really think that there's a part of marketing the solution, if you will, back into the organization that matters to succeeding. I think the combination of those explains why there's not a lot of success.

If you look at the latest research from Accenture, I think it's 68% of organizations can't find value from the data that they have. That's a huge number.

Also, it's just simply difficult for chief data officers to stay in place. I think I said earlier, I think the average tenure of the chief data officer is less than a thousand days. Primarily, it's related because we found that they rarely tie their technical initiatives to business goals and they might not think about the communication of those business goals. And so, that kind of hurts them in succeeding with their data strategy.

Michael Krigsman: All of this begs the question, who should own these data science efforts? When you say that there's a disconnect between the technology problem that is being solved and the business requirement, there has to be a cause. Who should own these efforts?

Bruno Aziza: What we found in some of the surveys that we did was primary two executives that this falls under. The First one, interestingly enough, is the CFO.

When we asked our community to tell us, "Where do you think data science and analytics should roll under?" a large percentage, 34% (not the majority) of people said primarily the CFO. The reason for that is because, I think, over the years, the CFOs have gone more from the back-office, cost retention type of role to innovating, using data as a way to power their organization and drive the operations of the organization. Data analytics and data science have a great opportunity to do that.

The CFO might actually be. Again, it depends on their objective. And so, you have to decide for your own organization. Actually, it might be a good group to own this.

The second one has been the CTO, the chief technology officer. Again, it depends on their style, their team, and so forth. I think the reason for that is because you want, on your bench, a good amount of very technical folks. We have seen, over the years, the data scientist was the sexiest data role. Now it's going to the machine learning engineer.

You think, wow, it's getting more and more technical. I think the reason for that is because the industry is innovating really, really fast. And so, you want technically savvy folks to enable you to deploy, but you want to couple them with your business folks.

It's probably harder to learn the business coming from a technology background than the opposite (at the moment) because we're now building technology that is taking over a lot of the tough steps that you might have needed to learn. I think about auto data preparation or auto data quality and all these steps that now business analysts can come in and start using.

What I would say is, what we see organizations do is they look at their blueprint. Say you have 100 people to handle all the data analytics issues. They tend to put half in the central business unit and then half into a central corporate unit like the CTO's organization or the CFO's organization.

Chief financial officer as the data owner?

Michael Krigsman: You said that CFOs could be the right folks to manage or be responsible for these data efforts. But my question is this. CFOs, in general, may understand technology but certainly, by and large, don't have the kind of deep expertise that's required from a technology standpoint. And so, how is it practical for a CFO to manage this? By the way, why didn't you say that the CIO should be responsible?

Bruno Aziza: There are two aspects to the answer to this question. Why not the CIO is because, at least from what we're hearing from our customers, typically if your CIO is focused on internal technology and infrastructure choices, data and analytics tend to be an application business.

We don't think about this today, but they are a business application consumed and directed towards value creation. I think that's why the CFO comes in here is because modern CFOs don't think about just budget and reducing costs. They think about opportunities for us to create value.

A great example of that is now organizations creating data products that they will monetize. I think about retailers, for instance, one of the great organizations I'm working with.

Carrefour is one of the largest retailers in the world. They've solved their data mesh issues for themselves and now have built around it, and they are now starting to create data products that they can sell back into their community, which is now talking about driving revenue for the organization.

I think the mindset of the most innovative organizations is that data is not a liability. Data is not something that I guard only. Data is something I build upon. It actually becomes an asset for me to manage up to a point where I can create products off of it and monetize these products. I think that's why customers are saying what they're saying around where it should fit.

Michael Krigsman: We have another interesting question from Arsalan Khan on an aspect of this topic. He says, "As organizations become increasingly reliant on AI and machine learning for decision-making, are some executives resistant to accepting data as the ultimate decision-maker?" In other words, if I can rephrase it, what should folks do if a business leader rejects the conclusions that the data presents?

"I don't think this is right. Yeah, sure. Your data says whatever, but I know from my experience it can't be right."

Bruno Aziza: This is the typical gut feeling that we deal with. The issue with a gut feeling is you never know if it's actual experience or if it's indigestion. You don't want to just rely on gut feel, but it is true that if you read a theory on this and books from Malcolm Gladwell and other folks that are very educated and well researched, the right decision is going to come from the combination of really good data and experience around the mistakes that maybe you've made or maybe others have made that you've been able to learn from.

I think, in general, it's never a good idea to decide 100% on your gut feel. You might get lucky every once in a while, but now we have technology that captures enough that you are able to not just understand but, in many cases, predict. And there many great stories like this in baseball, in the wine industry, and others like this that we can all relate to.

I think you're always going to get into a conversation with an executive that maybe might not believe the insights that you're bringing in. This is why, in the last few years, you might have seen the work from Nancy Duarte on storytelling, so connecting with the emotional aspects of how this executive might relate to the data.

In the book that we wrote a few years ago, Drive Business Performance, we talked about the example of Lego where the data analyst not only presented the information, the dashboards, but they actually had the voicemail left by the kids being played to the executive. The executive, as a parent, could relate to the customer feedback they would get and actually did change the strategy.

What I would advise our friend Arsalan here asking the question is don't think about just the binary logic aspect on how you're delivering the data. Think about the emotional aspect, the way people make decisions, even executives with great experiences, how they emotional connect with the data. That's really important as well to build into how you present your results to folks you're trying to convince.

Chief Digital Officer as the data owner?

Michael Krigsman: We have two interesting questions from LinkedIn. This is from Prashant Motewar. He says, "Number one, what about the chief digital officer as the owner of data and analytics? What do you think about that?"

Bruno Aziza: Given the past couple of years here where digitalization has really accelerated, we see certainly in organizations that would gather a lot of their information from physical locations – retailers and financial industries – nobody goes to the branches and nobody goes to the stores, so the person that's in charge of digitalization and taking this system into the future is certainly going to be interested in collecting and understanding data a lot faster.

What I would say, though, is it's not just the title that you have to look at it. It's also the organizational footprint. The people under this leader, is it the right talent? Is this the right organizational footprint? Do you have shared goals?

One of the important best practices that we see is that it's not just the CTOs job to innovate with data. It's the rest of the organization.

We worked with CIOs, CFOs, CTOs who share business goals that they actually don't have the direct impact into it, but they also have shared goals with the business folks who do not have direct impact on the metric itself. But the point here is to get them to get together, align, and collaborate.

The chief data officer and chief digital officer are great roles, but I wouldn't just be wedded to the title. I would look deeper into the organizational footprint of that organization.

Michael Krigsman: Another excellent question from Prashant, an important question. He says, "Data insights are extremely valuable when delivered at the right time to the right people with the right context. Any point of view on how to enable this?"

Bruno Aziza: A few years ago, I did a keynote at a data summit. I came up with this acronym, and not a very pleasant acronym but at least memorable, called RAT. R-A-T because data needs to be relevant, actionable, and timely.

You're absolutely right in your analysis. If I told you here's an umbrella because it rained yesterday, it's not very helpful.

There are a few best practices here. The first one is actual data literacy across the organization. It's one thing to deliver the data, but it's also another thing for people to actually understand the data. We also did a survey on how many data employees should you have inside your organization, so when you deliver the data, people understand what to do about it.

Keron Bourne, who is a member of the community, had (I think) the best answer where he said, "100% of your employees should be data literate." What that means is they should be able to recognize, they should be able to understand, and they should be able to talk data. I would make sure that at least all your employees understand the opportunity they have with using data so that when they get it in their context, they can use it.

Then a third of the organization should be data fluent. What does that mean? That means they should be able to analyze. They should be able to create arguments. They should be able to present results visually, emotionally to their management, to their peers.

Then 10% of your organization should be data professionals. A data pro here is someone who is paid to create value from assets.

The reason for why I'm saying this balance of roles matter is because the issue sometimes is you might be presenting data in the right context but to people that might not know what to do with it.

What we tend to forget is we're not in the business of building folks to become data specialists. We're in the business – you're in the business of doing your business. And so, to be able to be equipped with that, you have to deliver the data on time to an audience that is willing to or is equipped to act on that data. That's the most important thing is how do you act on that data.

Michael Krigsman: We have another question now from Twitter from Lisbeth Shaw who says, "In a busy company, should the person in charge of data science repurpose their old BI data strategy for today's data science needs? How should they do that?

Bruno Aziza: Too often, we feel like we have to hire outside folks to come and solve a problem because the other folks have a data scientist title and my folks don't have a data scientist. What I would say, I just talked to a chief data officer today who is looking at how she is going to upskill her team, and she is absolutely starting with the existing talent because institutional knowledge of your organization, knowledge of your customers, the knowledge of your organizational process is critically important.

Yes, you can bring outside talent that is technically gifted and so forth, but you'll never be able to hire enough of them so you can tackle the problems that you need to tackle and often in a very timely manner. The reality is that there are toolsets now that enable business analysts to step up into a data scientist type of role, and so I would never discard your existing talent. I am sure many of them are capable of doing more, either because they are motivated to get the training or because the toolset that is presented to them is making access to data and working with data a lot simpler.

I'll just give you an example on our data stack with this product called BigQuery. We have this embedded machine learning capability inside BigQuery, which means that you don't have to move the data. You don't have to set up an infrastructure. You can trigger models and run them with just a few lines of SQL.

Just these few lines make business analysts able to do work that in other platforms would require a lot of code from a machine learning engineer. And so, I think the good news here is that the industry – cloud vendors like us and the rest of the ecosystem – is really driving to making tools easier to use, which means (for you) you can use your people and upskill them into where you want them to be.

Value of low-code and no-code products to CIOs

Michael Krigsman: Certainly, when I talk with CIOs, the idea of low code, no code products is right, front, and center of how you can help your organization innovate while reducing costs.

Bruno Aziza: Absolutely. No code and no code is probably one of the most disruptive trends of 2021. It's primarily because it's enabling business users to create business purpose, domain-specific applications for themselves. I believe that the world of package applications is going to be disrupted over the next ten years because now these business users are able to just solve a problem by bringing in services, kind of in a composable manner, to create an application that's relevant to them and their community.

Definitely, the tools are getting easier. People are getting more skilled. There's more of an acute need for working with more data. I think the combination of those things doesn't mean that you should look for the answer outside of your organization. The answer often starts with you.

Michael Krigsman: We have another question again from LinkedIn. This is from Scott Beliveau. He asks, "What's the next data thing that organizations are not talking about and looking at but should be?"

Bruno Aziza: I look at a lot of technology. Recently, I bought myself these glasses, these smart glasses that I can talk to and they can do things for me and so forth. I think what we're not talking about enough is what is the future of the interface into data.

Today, we're used to our keyboards and we're used to our phones. But the reality is that, in the future – and not too distant of a future, I believe – our voice, our eyes are going to be how we interact with information.

There's a lot of investment going into natural language technology. It's a huge field around natural language processing, natural language understanding.

It reminds me back when I was at Microsoft where we shipped the Xbox. The Xbox had this amazing camera called Kinect. It was one of the most popular devices. It was this camera that would scan my body and the tagline was, "You are the controller."

I think that's where we're going with data. Ultimately, to broaden the appeal of data and make data more engaging for more people, we've got to change the relationship.

We're now in the mode where we have to talk machine. I think, in the next ten years, the machine is going to talk human, and that's going to empower new use cases, new experiences at home with smart home devices that you can call upon to set up your alarm or get answers and so forth. I think that's what's going to happen in the corporate context as well.

Michael Krigsman: Bruno, we're just about out of time. Any final thoughts or words of advice to folks who are listening regarding being successful with the data, aligning data efforts with the business, the things that are really hard?

Bruno Aziza: I do have one thing that I think is important for folks to consider is this whole domain of data culture.

One of my good friends Randy Bean just published a book. I think he calls it Fail Fast, Learn Faster. I might be misquoting the title, but I'll send you the exact link.

In it, he's interviewed data leaders and asked, what is getting in the way of being successful with data? It turns out that the majority of them are saying it's the data culture.

What I did is create my data culture checklist. We won't be able to go through all ten of them, but I'll give you a couple of best practices.

The first one, we talked about branding. Brand your initiative. Think about it seriously. Create a logo. Have certified data, so people recognize the quality of your data. It's really important in making sure that you take data seriously at your organization.

The second is have what I call a decision inspection. What I mean by that is often we do a postmortem. Often, we look at failure, and we say, "Okay. Let's try and understand how we failed."

Two best practices: Are you doing premortem? Premortem is, if everything goes wrong, what does that look like? Are we ready to react to a situation when it's going to go wrong?

Then the second practice is, analyze your successes. Often, you succeed and you move on, but do you truly understand why something really succeeded? Having this decision inspection mentality, I know it's hard because everybody is moving really fast. It's hard to prioritize. It is the best way for you to build this data culture that ultimately is going to make you one of the most innovative organizations in the industry.

Michael Krigsman: Bruno Aziza, thank you so much for taking time to speak with us today. It's been an action-packed 45 minutes. Really, really appreciate it.

Bruno Aziza: Thanks for having me, Michael.

Michael Krigsman: Everybody, thank you for watching, especially those folks who ask such excellent questions. Now, before you go, please subscribe to our YouTube channel, hit the subscribe button at the top of our website so that we can send you our newsletter and you'll get notified of our amazing upcoming shows, check out CXOTalk.com, and tell a friend.