All Email Is Not Created Equal: Understanding Email Economics When Considering the Cloud
The availability of seemingly low-cost cloud email services has put a laser-like focus on email spending. Organizations must delve deeply into the numbers to get a true picture of all the cost dynamics at work.
- Cloud vendors charge about $4 per user per month (pupm) for email, but we estimate that the fully loaded cost of that mailbox is about $9 pupm because customers must augment the service.
- Comparable, but not identical, email service via traditional on-premises deployment for 5,000 users is about $12 pupm. The higher cost generally can be attributed to smaller scale.
- There is a much broader range — from as low as $3 pupm to over $30 pupm — for on-premises email deployments because the range of features and operational efficiencies are very broad.
- Organizations need to be vigilant in ascertaining that the perceived cost savings from moving email to the cloud are, in fact, realizable cost savings, given the reliance of email on shared infrastructure.
- Organizations should perform a custom economic analysis — with special attention to shared-cost economics — to get a clearer picture of their own situation.
- The economic analysis should include all the capabilities of the email environment and a gap analysis with cloud email must be performed.
- Organizations should model a variety of scenarios, such as basic cloud email (missing certain features and integration points), and cloud email plus third-party services.
- Email decisions should always be made in the context of a larger strategy around collaboration.
Table of Contents
- Email Ownership Cost Components
- Calculating On-Site Email Costs
- Calculating Cloud Email Costs
- Comparing Service-Level Agreements
- Intangible Considerations
- Vendor Differences
- Future of Email Economics
With the ascent of cloud email, email economics has become a hot-button issue for many organizations. The topic is particularly thorny, since reliable information can be difficult to acquire. Challenges include:
- Accurately ascertaining current on-site email costs.
- Accurately estimating the total ownership costs for cloud email.
- Comparing costs when there may be differences between the two provisioning models in areas such as storage and SLAs.
- Factoring in intangibles such as risk, customization options, missing functionality and accounting strategies.
Economics, of course, should only be one element in considering moving email to the cloud. A longer discussion of other factors can be found in the document "E-Mail Is a Commodity and Other Fairy Tales" and in our "Toolkit: Cloud Email Readiness Assessment."
In the remainder of this report, we will explore in detail all facets of email economics, with the goal of demystifying this complex topic to ensure that IT leaders have the relevant data to make informed decisions.
The numbers used in this report are based on data gathered from hundreds of organizations. We've done our best to normalize and average out the numbers to be representative of this large population. No organization, however, will have the same ownership costs reported here. The numbers in this report should be used for illustrative purposes. Organizations looking for greater precision should perform their own ownership cost studies using data in this report to assist the effort.
We use Microsoft Exchange when discussing on-site email costs because it is the most widely used enterprise email system. The model presented here, however, will generally apply to most other client/server email systems such as the Attachmate Group's Novell GroupWise, IBM Lotus Notes/Domino and VMware Zimbra. Cloud email costs and discussion are largely based on the email components of Microsoft Office 365 and Google Apps for Business, which are the two most widely used cloud email services.
The large-grained elements that comprise the total cost of both cloud and on-site email are:
On-site. In an Exchange environment there are three components of the licensing costs: the Outlook client, the client access license (the fee charged for mailbox access) and server licensing. We do not include the cost of Outlook in our ownership costs since it is typically licensed as part of the larger Office suite. We do include the client access fee and the server licenses.
Cloud. Factoring licensing costs for cloud email is hard because the vendor license fee is bundled into the overall monthly fee. It might be said that Google does not have a license fee since there is no comparable on-site license charge. It appears to us that Microsoft is not monetizing the cost of a cloud license in the same way it does with an on-premises deployment, further increasing the economic appeal of Office 365. Microsoft may discount Exchange Online if the customer already owns the client access license, but that discount is typically less than what they pay for the on-site license.
On-site. Of all the email cost components, storage will have the greatest variability with an on-site deployment. The type of storage — highly available storage area networks (SANs) versus direct attached storage, for example — can make a 5:1 difference in cost. Organizations need to calculate the cost of storage based on managed costs, whereby the acquisition price of the storage is combined with utilization rates and the operational cost of managing the storage. Most organizations, however, don't offer more than 1GB of server-side storage to users. Users, however, typically have the ability to create personal archives (PSTs in Outlook), giving them unlimited local storage.
We do see some organizations reducing on-site storage economics via the use of JBOD (just a bunch of disks) and direct-attached storage, which, when combined with Exchange 2010's storage optimization, offers excellent storage economies. But full adherence to a Database Availability Group architecture can mean as many as three or four copies of each mailbox, eroding some of the storage savings. Other organizations are using "ready storage," whereby they pay the storage vendor for storage consumed, making it more of an operating expense. Backup is included in this line item.
Cloud. Both Microsoft and Google offer a breathtaking 25GB mailbox. In any cloud versus on-premises cost comparison, therefore, storage economics has to be a prime consideration. Organizations, however, should ask if users will, or should, utilize all 25GB. We generally recommend that organizations establish an email retention program that purges most email after two or three years to avoid unnecessary discovery costs (see "Building an Email Retention Strategy"). In these cases, the utility of 25GB mailboxes is lessened. Mailboxes with 25GB can enable organizations to eliminate local archives, which in most cases is a substantial (but non-monetary) benefit.
On-site. Organizations need a full complement of email staff that does daily care and feeding of the email system (level-two support) and offers deeper problem resolution, architecture, staff supervision and application integration (level-three support). Duties can also include email hardware and operating system maintenance.
There can be a large variability in the amount of staff required for email management, based on staff experience, age of email system, SLAs, topology and ancillary services. But generally speaking, we think a 5,000-person organization will have one or two level-two support personnel and one level-three technician. There is also a wide range of salaries for email personnel. Operational support is the biggest cost item in on-site email and we see some organizations outsourcing level-two duties to reduce expenses.
Cloud. Cloud email still requires customers to manage email on a daily basis. Level-two personnel must ensure that directory synchronization and single sign-on services are working and they must respond to configuration change requests and troubleshoot user issues. Level-three personnel still have application integration duties, as well as deep problem resolution and vendor management issues. Hardware and operating system duties largely disappear with cloud email, except for hardware required for directory federation and synchronization services, as well as a management server (required for Office 365). In the case of Google, organizations may continue to run Research In Motion's (RIM's) BlackBerry Enterprise Server (BES) on site for mobile services — and thereby incur operational fees. RIM offers BES services for no cost to Exchange Online customers.
We believe that most organizations will require about half of the email staff used for an on-site deployment when they move to the cloud for the first year of operations. Anecdotal evidence suggests that in year two and beyond, that number can drop to one third or one quarter of the original on-site staff.
On-site. Hardware costs for email have been driven down over the years — by virtualization, better software scalability and the overall hardware price/performance curve — but they still comprise a significant cost. A distributed topology will likely have higher costs compared to a centralized (and virtualized) topology. Extensive redundancy for continuity can double hardware costs. The test environment needs to be included in this line item.
Cloud. While the cost for hardware is bundled into the vendor fee for cloud email, customers are likely to have some slight hardware costs for directory synchronization and federation, plus, in the case of Microsoft, for the management console.
On-site. Organizations run at least one email utility — spam and virus filtering. Other common utilities are for services such as reporting, management or forensics.
Cloud. Cloud email services typically include spam and virus filtering in the base fee. While organizations may increase the number of third-party utilities in the cloud over time (as needs are identified and supply increases), at present there is little additional spending for third-party utilities.
On-site. Organizations need to incorporate costs for email server electricity, heating, cooling and floor space in the data center.
Cloud. There is a very minor data center fee for cloud email due to the requirement for on-premises directory activities. The cloud vendor bundles its own data center fees into the base fee.
On-site. Email is a major consumer of LAN and WAN bandwidth, but we don't factor it into ownership costs because internal networking costs are unlikely to drop when email is moved to the cloud.
Cloud. Cloud email traffic runs over the Internet and therefore may require installation of new router-based Internet access points, bandwidth expansion — as well as investments in redundancy, security and network acceleration. We also factor the cost of directory and sign-on activities as part of this line item. Theoretically, companies could generate savings if remote sites are able to downgrade the size of the WAN circuit, assuming that the cloud email goes out over a local access point rather than moving over the WAN.
On-site. Most organizations buy premium support from the email vendor. That support — sometimes in the form of a technical account manager (TAM) — is generally used for a variety of technologies, and is not exclusive to email. Consequently, we do not factor premium support into on-site email because those fixed costs are unlikely to diminish if email moves to the cloud.
Cloud. Although premium support is not required for cloud email, we strongly recommend that organizations buy additional support from the vendor to ensure fast problem resolution and effective change management. Microsoft offers cloud support via TAMs, premium support or the Cloud Vantage program. Google support is available via a TAM or from third parties.
On-site. Organizations typically upgrade their email infrastructures every three to five years. The labor involved in that upgrade is generally supplied by in-house personnel and included in our line item for personnel.
Cloud. For cloud email, we usually recommend using a third-party service and/or a third-party migration tool. Cloud migration fees, it must be noted, only apply once, since (at least in theory) the vendor will continuously upgrade the service with little or no impact on in-house personnel.
On-site. Directories typically service multiple applications and most large organizations have separate teams for adding, deleting and changing users, as well as for overall directory management. We don't believe this changes when email is moved to the cloud. Consequently, we are not factoring directory management into on-premises costs.
Cloud. Cloud email deployments normally require dedicated servers for directory federation and synchronization. Establishing the link with the cloud email system can be cumbersome, and the connection needs to be actively managed. Rather than calling that out as a separate line we chose to fold those activities into the networking line item. As with on-site directory, we do not monetize routine directory maintenance.
Vendor Operational Fee. In the cloud, the vendor charges a pupm fee for the core cloud service. This fee does not exist with an on-premises deployment. This fee includes hardware, software, some operational duties, phone support, community support, data center fees, storage, disaster recovery, and spam and virus filtering.
On-site Help Desk and Cloud Help Desk. This email ownership cost exercise does not include costs for level-one help desk because we believe level-one help desk costs are largely the same in the cloud versus on-premises. Problem escalation from level one to level two is included in the operational support line.
Organizations sometimes use email archiving, encryption, mobile device management and data loss protection services among other things. Each of these services costs roughly $2 to $4 pupm when deployed on-premises. These options are available for cloud email in the same price range from both the vendor (in the case of archiving and encryption) and from third parties (for mobile device management), but data loss prevention (DLP) is only available by routing traffic through an on-premises DLP gateway. Organizations using these services need to factor them into the total economic picture for both on-site and cloud deployments.
On-site email costs require a capital outlay for hardware, software, storage, networking and other items. In the example below we capitalized the hardware costs and amortized these costs over three years. Some organizations amortize capital costs over five years, which, of course, can have a big impact on ownership costs. We assume that the license costs were paid off years ago, and consequently we only factor in the cost for software maintenance. We assume the storage is part of a larger shared SAN and, consequently we blend capital and operational costs to come up with a per GB cost. As suggested above, we do not include networking fees since those costs are unlikely to diminish if email is moved to the cloud. As these assumptions demonstrate, extracting total email costs from a shared infrastructure can be difficult. Organizations need to be vigilant in ascertaining that the perceived cost savings from moving email to the cloud are in fact realizable cost savings, given the reliance of email on shared infrastructure.
We discuss our assumptions for each line item making up the ownership costs for on-premises email in Table 1 below. The example assumes a 5,000-person organization.
Licensing. We assume that a typical 5,000-seat organization spends an average of about $20 per user per year for the enterprise Exchange client access license (software assurance only) plus a fraction of the server license costs.
Storage. We assume the use of more expensive higher performance serial-attached or Fibre Channel hard-disk drives with three years' maintenance. Since email SAN storage is relatively lightly managed, the multiplier used to capture the operational costs is relatively small. The figure we calculate for the fully loaded cost of email storage is $2 per GB per month. In this case, we assume a 1GB mailbox.
Personnel. We assume that there are two full-time equivalents (FTEs) managing the email infrastructure at a fully loaded cost of $175,000 per employee (twice the salary). We assume there are four to five individuals that work on email, but their combined time adds up to two FTEs.
Hardware. We believe the average 5,000-seat company spends about $180,000 for the hardware infrastructure for fully redundant on-site mail encompassing: mailbox, hub transport, edge, proxy, and client access servers. These costs are amortized over 36 months. Storage acquisition costs are included in the storage line item. We are assuming a mostly centralized topology.
Utilities. The only utility we are including here is spam and virus filtering, which we believe costs an average of $6 per user per year. Operational costs for the filtering are included in the personnel line.
Data Center Costs. Data center costs are tricky to calculate so we used a somewhat standard metric of adding 10% of total costs to account for the data center costs.
Source: Gartner (June 2012)
It is unlikely that many organizations have the same cost allocations. We find a great variability in on-premises email costs for a variety of reasons such as:
- An antiquated version of Exchange, GroupWise or Notes — that has its costs fully amortized, with no software maintenance fees, running on old hardware, with very little disk space, part-time IT help and substandard uptime with no disaster recovery beyond tape — can operate for as little as $3 pupm.
- A new version of Exchange, GroupWise or Notes — with costs still being amortized, with software maintenance fees, on new hardware with 1GB or more of per-user storage, dedicated email technicians and best-in-class SLAs, with encryption, archiving, DLP and mobile device management — can cost from $20 to $30 pupm.
- Organizations with 20,000 to 100,000 users (and beyond) generate economies of scale far beyond those described in the 5,000-seat example. It is not unusual to see these mail systems operating in the range of $5 to $9 pupm. Very large organizations that move email to the cloud typically do so to avoid large capital expenditures associated with an email upgrade or migration, as opposed to a decrease in total costs.
Organizations also need to consider the economic impact of the timing of a move to cloud email. Organizations may still be in the throes of amortizing capital investments and those costs will continue unless the servers are re-allocated or the storage repurposed. License fees may still be in effect despite the move to another supplier. And if email technicians are redeployed to other duties then there are no labor cost savings.
The elements to consider in calculating the total ownership of cloud email include:
Vendor Operational Fee. The largest suppliers, Google and Microsoft, will offer very substantial discounts in competitive situations. Google's list price for Google Apps for Business is $4.16 pupm ($50/12 months), and Microsoft's list price is $4 pupm. Both vendors will offer volume discounts and resellers may also offer discounts. Microsoft may also offer discounts if organizations already own the Client Access License for Exchange. In the worksheet we use $3.75 pupm.
Migration Fees. For cloud email, we typically recommend using a third-party service and/or a third-party migration tool. Organizations may choose to use their own personnel for the migration. But in either case migration costs will be in the range of $10 to $25 per user, and in our worksheet we use $18 per user. This figure does not include training, and in the case of a migration to Gmail we strongly recommend end-user (and administration) training, which will raise transition costs.
Networking. Cloud email traffic runs over the Internet and, therefore, may require installation and operation of new Internet access points, which require firewalls, secure Web gateways and routers. Existing access points may need bandwidth expansion, as well as investments in redundancy and network acceleration. Some companies may use cloud-specific network-based acceleration services — such as those from the Akamai/Riverbed partnership — which target Office 365 and Google Apps. We also include costs associated with directory synchronization, single sign-on and federation services in this line item. We've priced the networking component at $1 pupm in our worksheet, which averages the costs from a variety of these activities.
Operational Support. For cloud email we believe most organizations will need one-half to one-quarter of the level-two and level-three email technicians required for their on-site email deployment. For the worksheet we assumed that the organization had been using two people to support email on-premises and for the cloud deployment we assumed one-half or 1 person would be required. We assumed that the fully burdened cost of that employee was $175,000.
Vendor Support. For the worksheet we assumed the cost of the extra support would be $75,000 and be delivered via a part-time technical account manager.
Table 2 assumes 5,000 employees and all one-time fees such as migration are amortized over 36 months.
Source: Gartner (June 2012)
Like on-premises email, there can be a wide range of cloud email costs. Microsoft, for example, sells a browser-only version of Exchange Online for $2 pupm called the Kiosk Worker Option. Organizations typically buy the Kiosk Worker Option for users that don't spend a lot of time in email, because they might pass on extra vendor support and they might not have any migration expenses beyond new account creation. And because this email service is not mission-critical, the organization might use fewer technicians to manage the service compared to a regular cloud deployment. In these cases, actual mail costs could be as low as $5 pupm for this constrained system. To match that service, we occasionally see Google offering (custom bid only) a Gmail-only version of Google Apps for about $2.50 pupm.
At the other extreme, organizations might buy the base service plus archiving, encryption, a third-party mobile-device management service and route the outbound traffic back on-premises to run it through a local DLP engine — in this case ownership costs can be as high as $20 pupm.
Organizations need to be cognizant of a variety of SLAs when comparing cloud and on-site email systems. For an on-premises deployment, higher SLAs require more hardware (due to extra redundancy) and create a greater operational burden (due to added complexity and more system components).
Cloud vendors provide a 99.9% uptime SLA and they are generally meeting the target. Our model for on-premises email also assumes a 99.9% uptime, although on-site uptime SLAs are more difficult to meet — because all system components (including network gear, for example) are part of the SLA — compared to cloud SLAs, which are for the email service only.
Cloud services typically promise a near-instaneous replication of data for close to no data loss, which translates to a five-minute recovery-point objective. We assumed a similar near-instanteous transfer of data with our on-premises model. Cloud services, though, generally don't offer backup services beyond 30 days, but our on-premises model assumes six months of backup.
Cloud vendors typically promise a two-hour recovery time objective (the time to get the system running in case of a catostrophic failure). But we don't know of any instances where cloud email data centers have been forced to failover, so we can't vouch for it. Some vendors seem hesitant to initiate a failover since we have seen outages certainly go longer than two hours. Our on-premises model assumes a six-hour recovery time objective.
Cloud suppliers typically have a rigid classification system for problem resolution, with incidents rated by severity level with a corresponding committment to address the problem. Vendors, understandably, don't commit in advance to a time-to-resolution of a given problem, but only the time to apply resources to the problem. Organizations comparing cloud with on-site email should scrutinize problem resolution approaches.
Making email decisions based solely on price would be a mistake. There are many other factors that should be considered in the decision-making process, such as:
Customization Needs. Organizations may need to integrate a variety of processes into the email system such as inbound/outbound faxing, voice mail, or automatic inbound routing. These services may not be available from a cloud supplier. A full range of customization scenarios, such as content control and security, is available in "Email Is a Commodity and Other Fairy Tales."
Functionality Loss and Gain. Cloud platforms may lack the important functionality available with an on-premises system. The Office 365 cloud version of Exchange, for example, does not support public folders or Outlook online mode. A chart comparing Exchange on-premises functionality with cloud functionality can be found in "Understanding the Microsoft Exchange Hosting Landscape."
Risk. One of the benefits of a close economic comparison of on-site versus cloud systems is the ability to quantify risk. If an organization concludes that cloud email is $3 pupm less expensive than on-site email, it then has a rather simple way to examine risk. So the question might be: is saving $180,000 per year ($3 pupm x 12 month x 5,000 users) worth the risk of going to a relatively new platform or going to a vendor that generates less than 2% of revenue from enterprise sales?
Opportunity Cost. Organizations routinely tell us that one of their goals in moving email to the cloud is to repurpose staff freed up by the move. The theory is that former email technicians can move to jobs where they can add more value to the organization. This is a valid goal, but it should be pointed out that retaining the employee will skew the economic analysis, since there will be no personnel cost savings and retraining can be costly.
When comparing Microsoft and Google cloud email economics it is important to keep several things in mind.
- Google sells Gmail as part of a larger package of workgroup applications including instant messaging, conferencing, shared workspaces and personal productivity tools. This is the service it sells for $4.16 pupm. Microsoft's $4 pupm Exchange Online service is email and calendar only. Microsoft offers a kiosk worker option for Exchange for $2 pupm. Google unofficially will sometimes offer Gmail only for $2.50 pupm.
- RIM offers BlackBerry operational support to Office 365 customers for no fee. Customers continue to pay for devices and airtime. Gmail users can use fee-based third-party BlackBerry services or they can run the BES on-premises (thereby increasing costs). The BlackBerry savings with an Exchange Online subscription should be factored into the economic analysis.
- Gmail users are more likely to use third-party management and audit services compared to Exchange Online customers (thereby increasing costs) due to a less comprehensive management and reporting infrastructure.
It is a buyer's market for cloud email services as Google and Microsoft compete aggressively for small and enterprise business. In January 2012, Microsoft expanded the Kiosk Worker Option to include Active Sync mobile support for iOS, Windows Mobile and Android devices, and it doubled the storage allocation from 500MB to 1GB. In March 2012, the company cut the list price of Exchange Online Plan 1 by 20% from $5 pupm to $4 pupm. We expect Google to maintain its $4.16 pupm fee, but we think it will add new features such as social networking, and voice and file sharing at no extra cost, thereby increasing the value. There is no such competitive pressure in the on-premises email market.
The vendor fee for cloud email only comprises about 40% of the total cost of ownership, and we see dynamics at work that will drive down other cost elements. We believe cloud email costs will drop the longer an organization has email in the cloud based on several factors:
- Elimination of migration costs. Vendors will perform all future upgrades as part of the base fee.
- Reduction in vendor support costs. We anticipate that support tools will improve, vendors' phone support and problem escalation will improve, and the systems will become more stable, all of which will enable the customer to reduce extra spending for vendor support.
- Reduction in operational support costs. Due to the factors listed in the line item above, plus internal skills development and the ability to use less skilled resources for day-to-day operational support, we believe operational support can be reduced by over half.
In this case, the economics are more favorable (see Table 3).
Source: Gartner (June 2012)
In this case, the vendor fee increases from 40% of the total to 55%.
On-premises email will take on more cloud attributes — which will have a positive economic impact — such as:
- Greater utilization rates and use of direct-attached and JBOD storage
- More server virtualization and centralized topologies
- Simpler and more uniform configurations
In both cases — cloud and on-site deployments — cost reductions will be off-set by additional spending required for mobile device management, records management, archiving, security compliance and other needs. Depending on the business requirements, therefore, email costs may actually rise as the sophistication of the system increases. But the cost of a generic mailbox will certainly drop.
Cloud email has already had a huge impact on the email market, even for those organizations running email on site. We have observed business people — fired up by their knowledge of consumer cloud email — forcefully demanding larger mailboxes, better uptime and mobility, and reduced chargeback rates. If those needs are not met internally (the argument goes) the business will move its email to the cloud, with or without the IT group's support. Cloud email has meant new accountability for IT, which will find it increasingly difficult to defer investments in the email infrastructure. The IT group, of course, may not be at fault — it is executive suite management that sometimes turns down the capital requests for upgrading email.
Economics, as we suggest above, is but one factor that needs to be considered in determining if a move to cloud email is warranted. But it will increasingly take on a more important role as cloud email services mature.