Latest trends in SaaS deployment models: Moving towards multi-tenancy and split plane

Authored by Matthew Kropp, Armen Derkevorkian, Wendy Kadon, Ben Robison, and Naresh Govindaraj

This paper is a joint research study between BCG and Alteryx.

One of the biggest decisions that cloud software companies can make is the design of their SaaS solution architecture. Customers have varying needs and requirements, depending on the sensitivity of their data, the level of control they want to have over the software, budget sizes, etc. To assess customer preferences regarding deployment models, we conducted a survey of 476 IT decision-makers within their companies. We presented four deployment models — multi-tenant, split-plane, private, and self-managed — and found that a plurality of participants (43%) preferred split plane architecture, followed by 31% participants for a private SaaS model.

The four cloud deployment models

Among the four models presented, the main differences between them are driven by two considerations:

1. Where are the data plane and control plane hosted (multi-tenant or private tenant)?
2. Who manages the software (customer or software vendor)?

In the chart above, the models are arranged on a spectrum ranging from less cost and control for the customer on the left to higher cost and control on the right. On the far left is a multi-tenant SaaS design, in which both control and data planes are hosted on public infrastructure and the software is managed by the vendor; this is a good fit for customers without sensitive data. On the far right is self-managed SaaS, in which both planes are hosted on private infrastructure and the software is managed by the customer. This benefits customers with strong data isolation requirements and with large budgets.

The further left we go on the spectrum; the more cloud deployment benefits are leveraged by both vendors and customers. For example, the vendors typically maintain only one (latest) version of the software and can easily track and analyze usage trends to shape future product roadmap, while customers enjoy the latest cloud software updates and access to the newest capabilities. The further right we go on the spectrum, the more control and autonomy the customers enjoy. In some cases, the far-right option is the only viable alternative for customers given regulatory requirements, in which case, vendors likely need to maintain multiple versions of the software.

Split plane SaaS and private SaaS are the most preferred architectures

A plurality of participants (43%) prefers split plane architecture, followed by 31% for private SaaS, 19% for multi-tenant SaaS, and 7% for self-managed SaaS. A clear majority of participants (81%) prefers the data plane to be hosted on private infrastructure, as most companies work with some form of sensitive customer data. As one respondent commented,

“We often don’t care where the control plane is, as long as the data plane is on our private network.”

Within models that utilize private infrastructure for data planes, split plane is the most popular, followed by private and self-managed, respectively. This is likely due to software management preferences and cost of ownership: As we move left to right in the above chart, the total cost of ownership will likely increase across multiple dimensions such as monetary cost, operational costs, as well as reduced ability of the vendor to provide support and troubleshooting services. Split plane architecture appears to be a “sweet spot” since the data plane is on private infrastructure, the software vendor manages the software, and the solution is economically viable for the customer.

Large companies (>$10B) more receptive to split plane and less receptive to multi-tenant SaaS

While respondents favor split plane SaaS overall, there is a considerable difference in responses between large and mid-sized companies. Large companies (>$10B) prefer split plane over multi-tenant by 33 percentage points (henceforth pp) more), while mid-sized companies ($101M-$1B) prefer it by 15 pp more. Architectural nuances become more salient as companies get bigger, so they are more likely to look for more tailored solutions such as split plane SaaS. Large companies (>$10B) are also around twice as likely to prefer self-managed SaaS compared to overall respondents, since they usually have more business units, geographic locations, and data types with stricter requirements. Companies with revenues <$100M tend to be emerging new companies who are highly knowledgeable in latest cloud architectures. While their sample size (N=50) is low in this study, their clear preference for split plane architecture is another indication of increasing popularity of this option.

Architecture preference varies across industries

Across industries, there is a distinctly clear preference for split plane SaaS, with the exception of Professional Services & Consulting. Respondents from this industry indicate the highest interest in multi-tenant SaaS (30%), which is comparable to their interest in split plane SaaS (32%). This may be driven by the fact that they are less likely to handle sensitive data (e.g., Personally Identifiable Information or PII, Protected Health Information or PHI) subject to regulatory constraints, as compared to other industries such as Healthcare or Financial Institutions. Thus, they are able to choose a more economically attractive option. The retail industry is the least interested in multi-tenant SaaS (6 pp less than overall respondents) and more interested in private SaaS (7 pp more than overall respondents). This is likely motivated by their heavy use of PII data and their resulting preference to have both control and data planes on private infrastructure.

Companies are more willing to adopt full or partial multi-tenant SaaS solutions in the next three years

Currently, roughly a quarter of respondents utilize multi-tenant SaaS and/or split plane SaaS architecture. But another 64–68% would consider using them in the next three years. There seems to be an overall openness for moving towards multi-tenant environments in SaaS solutions (whether it is a fully multi-tenant SaaS, or split plane SaaS where only the control plane is multi-tenant).

This trend is likely driven by three factors:

• Increased confidence by IT buyers in the ability of major cloud service providers (e.g., Azure, AWS, GCP) to meet security, privacy, and uptime expectations. As one survey respondent said, “We trust cybersecurity configurations of major public cloud providers and prefer to fully utilize public cloud, which is much easier than managing our own infrastructure.”
• Creative ways of masking sensitive data by customers enables usage of multi-tenant solutions. As one respondent, a healthcare expert, said to BCG, “We often de-identify sensitive PHI data to comply with regulations; this allows us to move data outside company network for SaaS use cases.”
• Proactive steps by SaaS providers to share legal and contractual responsibility in meeting regulatory requirements that apply to their customers’ data. As one survey respondent said, “Sensitive data can leave company network, as long as SaaS vendors agree to comply with regulatory constraints and share risk (e.g., through shared governance partnership).”

There is a clear openness to considering multi-tenant SaaS, and respondents believe that their company policies will adapt accordingly, although at lower rates. Half of participants indicated that their company policy currently allows them to have data plane on public infrastructure, and 53% expect the policy to allow them to do so in 3 years. But 39% expect company policy not to allow them to host data plane on shared infrastructure in three years.

While trends are moving in favor of a multi-tenant data plane, there will clearly not be a rapid adoption across all companies in the next three years. The need for private data planes (i.e., split plane, private SaaS, self-managed SaaS) will very much still be there in three years.

How should SaaS companies think about their product roadmap?

There are five key considerations that SaaS companies should think about moving forward.

1. There is no one-size-fits-all deployment model

Nearly all companies utilize some type of SaaS cloud application today, but their requirements for architecture specifications vary widely across industries and even within a company. While companies anticipate being more open to multi-tenancy in the future (both multi-tenant SaaS and split-plane architectures), the speed at which they do so will be influenced by three factors:

• The business criticality of the application — the more central an application is to the business, the more control companies want over it
• The sensitivity of the data, which includes both business criticality and regulatory strictness
• The size and capabilities of the IT organization, as well as its willingness to change

2. Solutions will need to work on multitude of different platforms and access different data sources

Companies’ cloud infrastructure composition is complex and will continue to be so. Mixed infrastructure will remain the norm even as companies move away from on-prem. Heavy use of public cloud will continue, but private cloud infrastructure is not going away.

Today, not only do companies use a mixture of infrastructure types, but they also use a mix of cloud service providers. Despite the ubiquity of public cloud, most companies need to connect to data sources that are only accessible on the company network or virtual private cloud (VPC). This has profound design implications for the software, especially when designing solutions to be hosted by companies themselves.

3. Develop modular architectural designs when possible

As customer preferences evolve, it is important to build flexibility into the product development process to allow easier deployment of various cloud infrastructures as well as customization of architecture. This minimizes the likelihood of having to start from scratch for every design scenario.

4. Be more open to risk- and governance-sharing with customers

Sharing governance risk with customers will help alleviate their concerns regarding SaaS providers’ handling of sensitive data types (e.g. PHI). Addressing some of these risks via the product design and/or other processes and controls could also help expand the market for certain SaaS solutions, benefiting both customers and SaaS providers.

5. Elevate the roles of marketing, sales, and pricing

Product strategy cannot be made in a vacuum in order to be successful. Marketing, sales and pricing teams will have to be equipped with understanding of the key differences between various architectures and their business implications, as well as the latest cloud service provider (CSP) security and privacy protocols. Additionally, they will have to dig deeper to understand customers’ needs, budgets, and constraints in order to guide them towards the right solutions.

Note: This survey was conducted in November 2022 with 476 IT leaders and executives at the “Director / Senior Director” level and above: analytics leaders (35%), data leaders (29%), information security leaders (16%), and data/IT governance executives (16%). 68% of participants work in companies with annual revenues of $1B or more. 69% of participants are based in the US, 25% in Europe, and 6% from rest of the world.