Data readiness for deriving business insight with analytics and AI/ML

Good afternoon, everyone. Are you all able to hear me clearly? All right, thank you. My name is Sash Wenger. I lead our Solutions Architecture team for US Commercial. And today I'm going to take some time to walk you through some simple tips on how to build a data driven strategy. Primarily to get more insights for analytics and business intelligence.

Here's the agenda for today:

• We're gonna talk about the importance of data.
• We're gonna talk about what does data readiness mean and how you can get there in small incremental steps.
• We're going to talk about the importance of culture to driving a data driven strategy.
• And then finally, we're going to show you some resources that AWS offers for free to help you create your own data strategy.

Um as the world becomes more and more digitized, we're seeing a staggering amount of data being produced by people by organizations, by governments, by things. However, when we ask organizations, if they are able to leverage their customer data and their operational data for competitive advantage, the answer is mostly no.

In fact, in a recent Gartner study, Gartner claims that a lot more organizations now talk about the value of data are investing in tools and technologies to capture more data, but truly haven't figured out how to extract insight out of them. In fact, 97% of company data stays unused untapped. And in their survey of 100 plus CX Os, 87% of the organizations showed up low on a maturity scale in terms of extracting insights from their data.

But we all know data is fundamental to deriving business insights, whether it's about your products, whether it's about your customers, campaigns are competition and modern technologies like generative AI further accentuate this capability gap in organizations.

When we talk to CDO's (Chief Data Officers), some of the insight we are getting is that a lot of them agree that a practical data strategy is needed if organizations are to succeed with technologies like generative AI. However, they also talk about a ton of challenges that they are experiencing.

There are organizations that are collecting a lot of data, there are organizations that don't collect enough data, there are companies that have the data, but the quality of the data is not really good. You have siloed and fragmented data sources primarily because of organizational silos. And I think one of the most important challenge is is your data integrated into your application strategy.

Finally, the cultural piece is probably the most important thing in driving a data driven strategy. And we see a lot of talking about data, talking about data driven strategy but not really using it on a day to day basis to make decisions.

Let's now walk through a few simple steps to create a data strategy and address some of those challenges. And when thinking about data strategy, the common issue is that folks start with data sources. It is extremely important that when thinking about a data driven strategy, you start with business outcomes. And when I say business outcomes, what are the most common questions being asked in the business? How frequently or how commonly are these questions being asked? And the reason to check on the frequency of these questions is so that you can prioritize the appropriate data elements that support those questions.

A simple technique that we use is something a simple map, a visual map to help you go through this process. Let's take a simple question. A very common question. What is the P&L for product X? Now a simple visual map for this would look something like this where you have the key data elements, revenue and cost on one axis and the business functions that rely on these data on the x axis. In this case, you will quickly see that you will understand which business functions either generate that data or rely on that data. And by going through this process through your questions, you're able to find out data elements that matter the most to most business functions. This allows you to prioritize a data platform rather than jumping in all in when you create a visual map, you want to prioritize data elements that drive value to multiple functions. You're maximizing the impact of putting a data strategy together.

The second you want to identify data elements that are needed to answer those common questions. Do you have gaps in your data or your quality or your access? A visual map allows you to quickly identify those gaps.

The next step for data readiness is about capturing what we call data black holes. When you recognize that you do not have the data to address some of these questions, you start to work backwards and to figure out what do we need in large corporations? A lot of transactions still happen in spreadsheets on the back of napkins or in work flows that are generally not captured in a system of record, which means all that data, which is very valuable is pretty much untapped. So as companies kind of think about, you know, going data driven, you want to make sure you're looking at these workflows, your processes, the small changes that you can make quickly so that you're capturing the data. If you don't capture the data, you can consume it, you can't use it.

The other piece is when you start having these conversations, it's very natural to encounter resistance. The first fear you probably will run into is the lack of confidence in your data it's very easy when somebody on a data team or an application team just dismisses a data, let's say a data artifact and says that does not look right. I don't think I can trust this data to get ahead of this. You want to be proactive, you want to look at your data owners and you want to set up a data reconciliation exercise ahead of time and do it periodically. When you're thinking data strategy, it is extremely important to validate and validate periodically because that's what builds confidence within the organization. It's important before you bring in any new data source, check the boundary cases validate or invalidate some of the constraints before you bring in a data set. And last but not the least, most companies do have some version of a single source of truth in their organization. When you're thinking about a data strategy and you're putting something together, you need to feel a little bold to go challenge that single source of truth sometimes because too often it's taken as granted and that's what everybody follows. You want to make sure you have certain experiments that are frequently validating that single point of truth. And the way you do that is one, make sure that your data and your application teams are actually integrated into your application strategy. You keep them separate, they think separately and it often leads to gaps in your strategy. Define a business reconciliation process, identify your data sources put a plan a quarterly thing where we say we run this exercise to validate these data sources and make sure that they reconcile.

Well, the last piece I would say is frequently challenge the single source of truth. It's very common for multiple systems in the company to be capturing the same data in different formats. Make sure that you're trying to avoid that duplication because it could create an alternate version of the truth.

The third tip for data readiness is talks about the right tooling and the importance of a PS data is a very fast evolving space. It is extremely important for you as an organization to be open to trying out different tools rather than forcing standardization. You want to give your developers the flexibility to experiment and pivot fast if there is a better tool out there.

The next piece I would say is for everyone stale data stinks and frankly, with the amount of data that we're creating these days, data becomes stale very quickly. And so your your transactional APS that are powering your application front end should be the same ones that are populating your data layer. So I want you to think about this. A lot of companies have applications capturing data, putting it in a certain data source, another system coming in doing some kind of transformation typically ETL putting it into another system and then another system coming in to run some kind of analytics or bad jobs, this creates a lot of gaps. More importantly, it prevents your business functions to get that insight in real time. And so if you can have your transactional APS on the front end populate your data layer, it can be significantly faster for your business functions to derive insight from that data

Last, but not the least, we create a ton of data reports and we distribute it across organizations, but we rarely check to see who's using these reports. How often are they actually being used to drive decision making? And so make sure uh that when you're looking at your dashboards or any reporting tools, check to see if they are actually being used, if they are not being used, deprecate them so that you're not creating this overhead in the system.

So the suggestions for tooling and APS be open to choosing the right tool do not for standardization mandate an integration layer for your data with APS. Anytime you create a new data source, you update it, make sure that there's an API layer that's creating abstraction that allows faster access and more experimentation. And finally deprecates stale data and reports by tracking usage.

Now, all these tips are great because they mostly address processes tooly. But there's a very important part to data readiness, which is the cultural aspect. And when I talk about data driven culture, what I mean is a company that actively uses data to make business decisions on a daily basis, they look at data as a strategic asset. They try to democratize access to that data across the organizations. They're facilitating more experimentation across the organization. And you could have the best tooling and a strategy. But what typically makes this sustainable long term is having a data driven culture.

Now, when you think about data driven culture, you, you start to think about how every layer of the organization, how do they view their role in capturing that data using that data and and and transferring that data to other parts of the organization. When you think about data driven culture, a lot of CSOs site culture as a huge impediment or a barrier to driving, you know, data driven, you know, uh strategies and to this is something that is incredibly difficult and long term. So I'm going to give you a small framework on where you can start and what are the important elements to drive that culture?

First, executive sponsorship of data driven initiatives is extremely important. The tone has to be set from the top. It is important that the organization sees leaders actively engaging in the data initiatives, showing the organization how decisions are being made based on that data and making sure that they are continuously engaging with different business units to validate or invalidate their beliefs.

A second part is the concept of a single threaded owner. This is something that we use at Amazon quite often data initiatives should be seen as a product, not as a by-product. When you have it as a product, you typically want a single threaded owner whose sole job is to make sure that that data initiative is successful. You want someone that wakes up every day and thinks about making this data initiative successful. And so beyond executive sponsorship, beyond a single threaded owner, it is in a unique position because you have visibility end to end across the organization, you can see the transactions and the work flows between different business functions. So make sure that your IT organization is structured in a way that drives accountability. At the same time, inclusiveness, you need to make sure you work with the different business functions to pull that data together consistently and maintaining a certain data quality standard.

Finally, governance, any time you're bringing data in there is always that risk of data being accessed by the wrong resource for the wrong purpose. So putting a certain level of access control, making sure that you're not replicating a lot of data sources just to provide access. These are fundamentals to making sure that you have a decent governance policy. Here is the remaining transcript formatted for better readability:

Finally, educate piece to see data, to see decisions being made based on data. The organization needs to have high level goals, very clearly stated, something that every business function and every employee in the company can kind of recognize and then you break it up into metrics for each functions. And there's a clear correlation between these metrics that an individual or a small team drives to the larger outcomes that the organization is delivering. When you speak a common language, it starts to become a habit.

So in summary, for a data driven culture, three things to remember:

• Executive sponsorship and engagement at the top
• Empowering your middle management to really make decisions on a daily basis. Leveraging the data
• And a language that even your front line employees can use consistently to talk about the business.

This is a simplified version of a strategy. It takes time, you need patience but done right. It's got amazing outcomes.

Now, we talked about the data driven strategy and the good part about this, you do not have to do this alone. Amazon has several initiatives, workshops, capabilities that come absolutely free to you to help create this data strategy.

I have a couple of links here:

• Gain insights which is specifically for small and medium business customers, allows you to talk to an analytics specialist to sit down and carve out an analytics road map.
• The data driven everything workshop is targeted at a specific use case or an idea that you bring to the table. And over a period of three days, we will chalk out the architecture and help you go implement it as quickly as possible.
• Finally, AWS for data, I would highly recommend spending some time on the portal. There's a massive trove of information about architectures, organization models, best practices that you should be able to take and apply it to your organizations.

With that I want to say thank you. But before I leave, if you are participating in the SMB mobile treasure hunt, the keyword for this session that you would need is readiness to get credit. So good luck with that. And thank you again for making the time.