Business and IT leaders, once racing to get artificial intelligence (AI) and analytics workloads into the cloud, are beginning to take a more calculated approach — and for good reason. For many, the promise of the cloud once read like a slick commercial for an all-inclusive vacation, but without critical details such as how to get to the destination and the full cost of the package. Most assumed that the benefits would ultimately be worth it and took the leap.
The good news is that the benefits are real — and they are worth the effort it takes to realize them.
This cloud analytics maturity model, developed by SAS, acts as a “Maslow’s hierarchy of needs” for business and IT professionals responsible for AI and analytics. As organizations climb higher and higher within the model, they unlock more and more benefits from the cloud, including efficiency upgrades and cost savings.
Baseline needs (such as migrating and operationalizing current code in the cloud) are obvious and immediate. Higher-level needs emerge as demand for AI and analytics increases and an organization’s ecosystem matures.
The cloud maturity model
At a glance
According to a new survey conducted by Foundry on behalf of SAS, moving data analytics workloads from on-premises to the cloud improves performance across every function.
However, factors like…
42%
43%
43%
To better understand where organizations run into trouble when migrating AI and analytics workloads to the cloud, let’s take a closer look at what survey respondents shared about their experiences. For context, note that respondents have reported that moving analytics from on-premises to the cloud has delivered net benefits in the range of 7% to 25% improvement in key areas such as:
What’s going wrong
Factors inhibiting the upgrading of analytics platforms
Average impact on performance from moving data analytics workloads from on-premises to the public cloud
Read the full whitepaper
Everyone’s moving AI and analytics to the cloud. Are they realizing expected value?
Reaching the promised land
Phase 1: The initial cloud migration
Centrally monitoring and managing all models across the enterprise
High (9, 10)
Medium (7, 8)
Low (1-6)
1%
2%
3%
4%
4%
5%
7%
8%
Sponsored by
But in the real world, moving on-premises analytic workloads to the cloud can feel more like an extreme alpine climb: huge potential payoff if you reach the top, but unexpectedly high costs, technical complexities, and risks every step of the way. The pinnacle value, which SAS captured in its cloud analytics maturity model (see figure 1), is the draw: the opportunity to be a truly data-driven organization that relies on advanced AI and analytic models to make decisions, deliver automated customer services, trigger automated intelligent processes, and more (Level 7) — and being able to preserve this value over time by centrally monitoring and managing AI and analytic models (Level 8).
“When you’re implementing such a large change, IT must act as a guide by charting the best path forward and ensuring that everyone is properly prepared and equipped for the realities of the journey. If they don’t, people will get stuck — and they may even decide to turn around. When there isn’t a clear understanding of what’s ahead, we see companies run into issues and even pivot away from the cloud, which ultimately moves them away from all the benefits afforded to us by using the cloud.”
— Jay Upchurch, Executive Vice President
and Chief Information Officer, SAS
complexity
other priorities
and cost
Preserve
value
…are inhibiting migration.
For many, however, the pinnacle seems out of reach. According to the survey, which polled 408 global IT and line-of-business decision-makers across more than 20 industries, most organizations are getting stuck somewhere on their ascent, running up huge costs and never actually realizing expected business value. That’s why so many of those who have begun —
or have completed — a migration of their AI and analytics workloads to the cloud report being surprised at how public cloud costs are much higher than they expected. In addition, they report that they have yet to realize less strategic, but equally important, outcomes such as cost savings and efficiencies, increased customer and employee satisfaction, lower total cost ownership, and greater business agility and innovation.
Organizations expecting their cloud analytics journey to be a hassle-free all-inclusive vacation are surprised to find that their journey to cloud analytics is more like climbing Mount Everest.
8
7
6
5
4
3
2
1
Maximize value
and access
Upskill and democratize
Reduce development delays
Empower professional coders
Govern the ecosystem
Optimize cost performance
Migrate to the cloud
At levels 7 and 8, you are making business decisions supplemented by analytics across the business while maintaining effectiveness over time.
At levels 5 and 6, you are condensing collaboration cycles between data, analytics, and the business problem.
At level 4, you are empowering all professional coders to innovate on existing or net new use cases.
At level 2, you have moved data, workloads, and users to the cloud — paying by the minute. Observing and optimizing cloud costs is an early focus.
Respondents also reported that moving data analytics workloads from on-premises to the public cloud resulted in improvements in important ways, such as performance enhancements of up to 8% in key areas.
25%
16%
19%
15%
16%
13%
12%
Time needed to get the best models
Developer productivity
Time needed for professional coders to get to an explainable result
Revenue improvements
Time to rewrite code to another language
Operational cost improvements
Nontechnical employee productivity improvements
Accelerating new models production
Upskilling and democratizing analytics by empowering
non-technical business analysts with no-code/low-code tools
Reducing development delays
Accelerating the speed with which professional coders can build and improve models
Establishing guardrails for safe, responsible model innovation and development
Modifying workloads to consume less cloud infrastructure
Modifying or rewriting existing code to be operational in the cloud and driving business value
However, others report seeing no benefit or facing ongoing challenges regarding explainability, model optimization, code migration, and challenges and roadblocks in the following areas.
Those surveyed — all IT and line-of-business decision-makers at companies with 500 or more employees and representing more than 20 industries — said their organization is facing major challenges when upgrading its analytics platforms. Challenges include complex and difficult migrations of workloads, excessive time spent rewriting business application code, performance and scalability issues, lack of sufficient talent, unexpectedly high migration costs, and more.
Unexpected complexity and difficulty in migrating workloads
Scalability and agility
Performance and scalability
Lack of appropriate tools and processes
Data dispersion
Time required to rewrite business application code
Generative AI
Talent availability
Technical debt
Cost of upgrade
Lack of business priority
Complexity and difficulty to migrate workloads
Strongly disagree
Somewhat disagree
Somewhat agree
Fear of missing data value has led my organizations to prioritize data centralization over data organization
Moving workloads to public cloud has led to poorly-thought-out data swamps
Validation is the most time-consuming part of a migration
Moving workloads to public cloud increased the complexity of data-driven decision-making
Code that is not rearchitected to run optimally in the cloud can be slow and costly
Our current data analytics stack enables data users with low-code/ no-code application development tools
Responses show sporadic agreement with several of the key value propositions to leaving analytic and AI workloads on-premises.
41%
43%
43%
42%
41%
41%
40%
40%
42%
42%
39%
15%
16%
20%
41%
30%
44%
18%
30%
16%
0%
32%
36%
21%
11%
0%
28%
44%
17%
11%
0%
11%
27%
45%
17%
0%
17%
26%
37%
20%
0%
4%
22%
48%
26%
0%
4%
17%
50%
29%
16%
17%
44%
39%
20%
42%
38%
20%
42%
38%
21%
44%
35%
They also report many hidden costs in achieving a fully functional implementation — costs on top
of the basic public cloud services subscription fee. These costs include paying for additional cloud services; ongoing managed IT services; IT project consulting or systems integration; on-premises hardware or software; business consulting; additional bandwidth, VPN upgrades, and remote access services; and IT, end-user, and line-of-business training. On average, these costs add an extra 56% to the cost of moving data analytics workloads to the cloud, compared with keeping them
on-premises. And it’s why only 2% of those surveyed said they will save money by moving workloads to the cloud.
“With costs on the rise, companies must be careful in managing their cloud resources and reducing waste,” says John Gallant, Enterprise Consulting Director at CIO.com. “Some organizations are finding it a struggle to balance the value of cloud performance against the need for fiscal guardrails, and that can leave both the finance and IT departments facing new challenges.”
Higher-than-expected migration/upgrade costs
Given all these challenges — and the 56% higher costs (above and beyond public cloud service costs) to move data analytic workloads to the cloud — survey respondents reported that on average, it takes them 16 months just to recoup the costs of this investment. All things considered, the higher initial investment still renders a solid return on investment (ROI) — but at a slow pace.
Protracted time-to-value realization
16 months
5%
12%
20%
24%
30%
7%
1%
Less than 3 months
Estimated time to recoup investment in moving data analytics workloads to the cloud
3 months
to less than
6 months
6 months
to less than
12 months
12 months
to less than
18 months
18 months
to less than
2 years
2 years to less than 3 years
3 or more years
We will
not recoup that cost
Average time to recoup investment in moving data analytics workloads
1%
There is sporadic agreement on why it’s best to leave analytic workloads on-premises. Some say that their current data analytics stack enables users with low-code/no-code application development tools. Others find that code was never rearchitected to run optimally in the cloud, resulting in slow and costly analytic performance. This increases the complexity of data-driven decision-making, which is not their desired outcome. Others report frustration with tasks such as validation, which 62% agree is the most time-consuming part of a migration.
Excessive time to rewrite and validate application code
Strongly agree
Don't know
Engineers working on development projects for lines of business are generally paid to create new things, not manage old things or make them run more efficiently. In the quest for faster innovation, they write functional code but not “perfect code” that will run efficiently in the cloud. This creates technical debt for IT functions to deal with — debt that makes it harder and more costly to run these workloads in the cloud. In the end, IT entities under pressure to reduce cloud costs must either demand that code be refactored, do this work themselves, or make internal chargebacks for cloud costs so that lines of business bear the cost of their inefficient models.
Growing technical debt
“With costs on the rise, companies must be careful in managing their cloud resources and reducing waste. Some organizations are finding it a struggle to balance the value of cloud performance against the need for
fiscal guardrails, and that can leave both the finance and IT departments facing new challenges.”
— John Gallant, Enterprise Consulting
Director, CIO.com
These findings reveal a major disconnect between the real-world effort involved in moving analytics and AI to the cloud and the long-term value that companies expect to realize from these investments. This is a serious problem, because no value is realized until people actually use cloud AI and analytics to make better, faster decisions. This only happens at the final stages of cloud analytics
maturity — and until it does, it’s all effort and investment.
According to Jay Upchurch, Executive Vice President and Chief Information
Officer at SAS, one way to close the gap is to rightsize the tools and resources
IT has available at each stage of the journey. “When you’re implementing such
a large change, IT must act as a guide by charting the best path forward and
ensuring that everyone is properly prepared and equipped for the realities of the journey. If they don’t, people will get stuck — and they may even decide to turn around. When there isn’t a clear understanding of what’s ahead, we see companies run into issues and even pivot away from the cloud, which ultimately moves them away from all the benefits afforded to us by using the cloud.”
Closing the chasm between current efforts and expected value from cloud analytics
These issues tend to cluster around specific levels of SAS’s cloud analytics maturity framework. “We’ve grouped these clusters into three phases that act much like the grading systems used to classify the difficulty of alpine climbs,” explains Upchurch.
Phase 2: The “messy middle”
Phase 3: The promised land
This initial segment of the journey — moving existing code to the cloud and getting it operational — is equivalent to the low-lying foothills of an alpine climb: Risks are still present, but they can be planned for and managed with the right tools.
Most organizations today have progressed to somewhere between levels 2 and 6. These levels are all about costs, as companies are investing in building and deploying new models at scale. Many organizations get stuck, and some even move workloads back on-premises due to spiraling costs.
This phase encompasses levels 7 and 8, which are all about realizing business value. At these levels, businesses are truly data-driven and confidently reliant on analytics models to make decisions, provide customer service, automate processes, and more. The focus is on getting more models into production faster and managing them so they keep delivering value over time.
1
Migrate to the cloud
2
Optimize cost performance
3
Govern the ecosystem
4
Empower professional coders
5
Reduce development delays
6
Upskill and democratize
7
Maximize value
and access
8
Preserve
value
Phase 2: The messy middle — most organizations are stuck in this phase, unable to unlock the true value of the cloud.
Phase 3: The promised land — this is where organizations capture real ROI.
Phase 1: The initial cloud migration — most organizations have completed or mostly completed this phase.
IT can use these levels to assess and plan their organization’s journey, communicate plans and risks, and train and equip people for success. This includes adopting enterprise-grade, built-for-purpose tools and best practices that are designed for coordinated, efficient, large-scale initiatives for moving analytics to the cloud.
“For a transition like this, companies need their IT leaders to serve as their guide and lead them to the top of the mountain. That includes making sure the right tools and resources are accessible for people at each part of the journey to ensure success,” explains Upchurch. The most efficient way to do this is for IT to adopt an all-in-one modern data and AI platform that:
Moving toward maturity
Equips IT leaders with next-gen tools and resources that dramatically reduce time, risk, and effort (especially when moving existing workloads to the cloud); streamline tasks; and provide centralized visibility, monitoring and control, and model governance.
Gives developers the freedom to use their preferred tools and interfaces — including open source tools — to innovate and deploy new models more quickly, safely, and inexpensively but with corporate IT guardrails to protect the organization.
Provides non-technical employees, such as line-of-business analysts, with intuitive, no-code/low-code tools so they can help accelerate innovation — because there will always be more problems to solve than developers to solve them.
To learn more about the three phases of cloud analytics maturity, and how IT can accelerate their organizations’ journey to ROI
Learn more
On average, it takes 16 months to recoup investment in moving data analytics workloads.
The three phases of cloud maturity
Results from a survey of 408 global IT and line-of-business decision makers