Top 5 Mistakes Your Data and Analytics Team are Making in the Cloud

The following is a transcript from a webinar: Top 5 Mistakes Your Data and Analytics Team are Making in the Cloud hosted by EVP of Cloud, Greg Pierce.

[0:00:01] Thank you very much for joining today. What we're gonna talk through today are the top five mistakes your data and analytics team are making in the cloud. And of course, the reason that we would want to discuss this particular topic is so you can avoid those top five mistakes and find ways to accelerate your journey to the cloud, and make the cloud support your data and analytics workload in the most effective and efficient manner.

So what's the first mistake people make? It's a lack of planning. This probably resonates across a lot of things that we run into, both in our lives and our business lives as well, and sometimes the gut reaction, especially when you're looking at deploying your first workload, if it's something you're just experimenting with is to jump in feet first, just go ahead and start deploying, see what you can come up with.

And the mistake that's made there is that once you start down that path, it's really hard to unwind, and as you start developing something that's kind of in a silo and the architecture most likely doesn't align with business needs because you're moving so fast, you didn't take time to spend with other groups within your organization to determine exactly how it should be built.

[0:01:27] So decisions made in silos have a tendency to come back to bite us down the road, because if somebody isn't involved in the decision-making process, it's really hard for them to buy into it, and without buy-in, data and analytics deployments often fail. And the reason they fail is that one of the critical components of a data and analytics deployment is adoption.

If you can't get to a point where the company is going to adopt it, then everything that you've done from a planning perspective. From a design perspective, architecture, deployment, were all done for no good reason because it's not going to be used. That's not going to help lead you to the intelligent enterprise with data and cloud in the center of what you're doing.

Often when there is a lack of planning in place, the data baseline that you're looking at, in other words, the hygiene of your data is not good. If you start from a place where the data needs governance (it needs to be cleaned up), and you go down the path of looking to get new decisions out of it, the decision is going to be as good or bad as the data. So the proper planning, wrapped around making sure that your data is in the right place is super important as well.

[0:03:03] So the second mistake is where costs are misaligned, and I can give you an example here. We had a customer doing a very large global deployment and they anticipated that their monthly spend within the cloud was going to be $50,000 a month, and month one was right around $30,000. It looked like things were good, but then month two was 75, month three was 130,0000. So we were able to come in and over the course of 90 days, get them to a point where it was right-sized and at the right spot, which was the $50,000, but it cost them over $100,000 to realize and to fix that costly mistake.

Due to wrong cost estimation and lack of planning, this business lost $100,000 from cloud consumption and post-deployment clean-up.

When you think about it, that lack of planning, I talked about – It’s kinda like renting a house without knowing exactly what the price is ahead of time and just hoping at the end of the month that it's within your budget.

(iStock.com/scar1984)

So if you deploy from a cloud perspective, and there are several different methods that you can use. One is a pay-as-you-go where you pay with a credit card, for example, and at the end of the month, you pay for what your consumption was for that given month. Another would be that you do it through a partner, and the partner can help you do modeling. And oftentimes modeling is not performed when costs are looked at from a cloud perspective, and the important thing I think is to determine which makes the most sense for you but to do it pragmatically.

Would you rent a house without knowing how much it would cost month to month?

[0:04:52] In other words, it's okay to start with a pay-as-you-go type scenario as long as you have a plan and you know that your costs are not going to exceed a certain amount. So if you're just starting to deploy a little bit of a workload, you know it's gonna be let's say $1000 or $2000 a month, okay, that's fine. Let's make sure that this is going to work before we jump headlong into a more formal cost model.

Once you've designed out your model going forward and you know what it looks like, then it's time to leverage some of the optimized type scenarios in the cloud. Some of those are like CSP, which is through a Cloud Service Provider or reserved instances where you model out what your spend will look like over the course of a year, and then you commit to that year and you get it at a significant discount. So that helps you avoid the surprise like the customer that I spoke about earlier where month three, you're more than double what you thought it was going to be.

[0:06:02] The other part I think that goes unspoken oftentimes is that budget overruns can cause projects to slow or fail because there's only a certain amount of money available for the entire project. If you're inefficient in your use of the cloud and the infrastructure components that are required for that data and analytics workload to run, the money is going to have to be siphoned from the actual data and analytics project to be able to support it.

So oftentimes what we see is if this is not done properly upfront, the project is either shortened or truncated and you don't recognize the entire business value that you otherwise would have. I think another really important part to understand is that skillsets within the cloud are certainly different than they are on-premise, and oftentimes they do convey, which is good, but there are certain components within the cloud that are completely different and need to be something that's addressed.

When the cloud deployment isn’t done properly upfront, the project is either shortened or truncated and you don’t recognize the entire business value.

[0:07:18] So I think it's important to put together a learning plan that you would have in place for each individual that would support cloud and to either hire or bring in experts from outside that can help bridge those skill gaps until you get up to speed because the first time that things start to go south is when you really learn what those skill gaps are.

Description: CCGers whiteboarding as a training exercise.

There's enough graphical interface in the cloud that you can probably muddle your way through and find a way to deploy some servers, but once you get to a point of a problem occurring where there's a network issue or a security issue and you have to start using, if it's AWS, cloud formation scripts or PowerShell within Azure, things are much more advanced in that case, and you have to make sure that they're configured properly.

What ends up happening is if you don't have the proper skill sets in place, things are designed based on what you know and how you can deploy it. So basically a 101 or maybe a 201-level deployment, to deploy properly, you need those advanced skills ahead of time, and that will avoid leading to wrong decisions based on those skillset gaps. The team is going to deploy based on what they know, not the optimal way to do it.

[0:09:09] One organization that we worked with just recently ended up in a scenario where their entire environment was unavailable for almost 36 hours because it was improperly configured and then not supported properly afterward.

And unfortunately, we had talked to them beforehand about helping them from a skill set perspective and with some of the advanced components within public cloud, and they ended up having to engage us once the problem happened.

So what I tell everybody that's listening is, certainly avoid having to fix things by doing it right from the beginning. Plan it, and then make sure the skill sets are in place that you need to be able to support your environment.

An improper configuration led to the cloud environment being unavailable for almost 36 hours.

 [0:10:34] So mistake number four, security becomes secondary. And look, it's easy, especially if you're doing a (Proof of Concept) POC or something like that, to fall in the trap about, "We'll worry about security later once it's real."

But we have a customer that recently was deploying a proof of concept within public cloud, and they came to the realization that the data that they were populating their data warehouse instance with was actually subject to compliance and was very sensitive data. So not engaging the security team can end up putting you in a place that is not a good one from a compliance perspective, and even beyond compliance, from a security perspective, securing customer information, securing employee information, whatever it happens to be.

We end up working with some customers that are in some extremely regulated and sensitive data type industries. So security is always top of mind for us, and it certainly should be top of mind for anybody who is deploying a data and analytics solution in the cloud. By nature, people that are within data and analytics groups or that's their specialty, security isn't going to be the first concern.

That's why you have to make sure that you get somebody who understands security to either come in externally or from your internal group because we just see if you don't get this straight, a lot of times, you just have to start over from scratch, and that doesn't help anybody at all.

[0:12:10] So mistake number five deals with ongoing operations, and what we see from many organizations is that after the solution's deployed, they set it and forget it. They think that it's just going to run, and the assumption is that IT will just make it work.

Those sort of things lead to at a base level and sometimes it's just little things like performance issues, sometimes it ties back to cost, and there's cost overruns that have to be right-sized. A lot of times, though, what we see is that those sort of underlying things lead to a larger problem down the road like an inordinate amount of downtime when it could be avoided just by simply putting in place the right standards from a deployment perspective, the right governance and the right things that are needed from a process perspective.

For example, change management, the formalized process around anything that you're going to alter is logged within a central system. A change advisory board to regulate whether or not things should be done and when they can be done. Otherwise, if things are not running at an optimal level on a going-forward basis and on a continued basis, it's really tough to move on to the next data and analytics project so that you can really make decision-making a critical component because you keep having to come back and make the first one that you put in place work.

[0:13:54] So how do you get it right? Well, you probably guessed it based on some of the things that we talked about that can go wrong, but just make sure you absolutely have the right plan in place. Plan and then plan some more. Make sure that you are pulling in people from across the organization, build bridges and not walls. Make sure that the other people that you pull in have expertise in areas that you don't like security, like perhaps ongoing operational models, like cost optimization within public cloud. And if you don't have it within your organization, you need to go find it either through external vendors that can help you or by hiring people in that have that sort of experience, which means know what you don't know.

Understand that it's okay to have gaps that you need to fill in, and that you need to understand what your end result is from a data perspective and what you wanna bring to the organization. All the pieces that tie to getting you there are not necessarily pieces that I think anybody's gonna look at you cross-eyed if you tell them, "We don't have this internally right now. We need to build this and we're gonna need help until it's built."

It's okay to have gaps that you know you need to fill in.

[0:15:14] So that's the end of this webinar. Thank you guys for coming today. Really appreciate it, and if there's any questions anybody has, please reach out to us either through LinkedIn at CCG Analytics, through our web page. You can always reach out to me at gpierce@ccganalytics.com. Thank you.

[End of Transcript]

Want to optimize your cloud purchasing agreement for saved money and time? Learn the three ways to purchase Microsoft Azure with this guide.

Written by CCG, an organization in Tampa, Florida, that helps companies become more insights-driven, solve complex challenges and accelerate growth through industry-specific data and analytics solutions.