Survey Analysis: Customers Rate Their Business Intelligence

Platform Ownership Cost

27 August 2015 ID:G00290015

Analyst(s): Rita L. Sallam, Josh Parenteau, Cindi Howson, Ehtisham Zaidi

VIEW SUMMARY

BI leaders often focus solely on license metrics when comparing BI platform vendor and
analytic tool costs, even though it makes up only a small portion of overall cost of
ownership. Here, we compare total three-year BIPOC and its components across
vendors, vendor types and deployment sizes.

Overview

Key Findings

Buying low license cost platforms does not always translate into low business
intelligence platform ownership cost (BIPOC) over time, or into higher business
benefits.
Open-source platforms, often purchased due to low license costs, have the
highest three-year BIPOC per user due to high IT full-time-equivalent (FTE) costs
per user.
Data discovery leaders have the lowest three-year BIPOC per user, largely driven
by a reduced need for IT FTEs and external implementation services.
Cloud BI vendors have a below average three-year BIPOC per user with a very
favorable (the lowest) hardware cost per user profile and low initial implementation
costs.
Large independents and megavendors have higher-than-average initial license,
maintenance and hardware costs, as well as initial and ongoing implementation
costs, but lower-than-average IT FTE costs per user — due to the efficiency of
supporting large, viewer and report-centric deployments — result in a favorable
overall three-year cost per user profile.

Recommendations

Balance any cost consideration with functional requirements, expected adoption

and potential business benefits when evaluating vendors.
 Manage the cost of incremental pricing as deployment sizes grow and manage
maintenance increases over time, rather than focus solely on initial price
negotiations.
Consider cloud BI as an option to reduce infrastructure costs and to take
advantage of product innovations without the need to upgrade existing
infrastructure.
Include switching costs as part of your evaluation if you are considering moving
from a high-priced vendor to one with lower overall costs.

Table of Contents

Contents

Survey Objective
The 2015 Magic Quadrant Customer Survey
Data Insights
Overall BIPOC Observations
License Costs
Maintenance and/or Recurring Annual License Costs
Initial Implementation
Initial Implementation Cost per User for Internal Resources
Initial Cost for Infrastructure (Hardware)
Ongoing External Service Provider Cost
IT FTE Costs
Methodology

Tables

Figures

Survey Objective

This research details the business intelligence platform ownership cost (BIPOC) results
from Gartner's 2015 BI and Analytics Magic Quadrant customer survey of 2,083
companies using BI platforms.

Figure 1 shows the key cost drivers of the seven license components analyzed in this
report. They are divided into three overall categories:
 1. License/additional infrastructure (hardware)/maintenance and recurring fees:
License costs
Maintenance and recurring fees
Additional infrastructure (hardware)
2. Initial internal implementation and ongoing IT FTEs:
Initial implementation (internal resources)
Ongoing implementation, IT FTEs
3. Initial implementation and ongoing implementation (external service provider):
Initial implementation (external service provider)
Ongoing implementation (external service provider)

Figure 1. Three-Year BIPOC Components and Drivers

BIPOC = business intelligence platform ownership cost; FTE = full-time equivalent;

OOTB = out of the box; SI = system integrator

Source: Gartner (August 2015)

The different pricing and packaging models of BI platform vendors make direct
comparisons difficult for BI leaders evaluating products. Using our survey data, we
divided total costs for each cost component by reported average deployment sizes to
calculate an average cost per user to create a basis for comparison across vendors,
vendor types (Magic Quadrant versus Other Vendors surveyed) and the vendor
categories, as shown in Table 1.

Table 1. Vendor List by Category

Vendor Vendor Vendors
Type Category
MQ Cloud BI Birst, GoodData
vendors
Data Qlik, Tableau, TIBCO
discovery
leaders
Large Information Builders, MicroStrategy, SAS
independents
Megavendors IBM, Microsoft, Oracle, SAP
Open source OpenText (Actuate), Hitachi Data Systems (Pentaho)
Small Alteryx, Board International, Datawatch (Panopticon), Logi Analytics,
independents Panorama Software, Prognoz, Pyramid Analytics, Salient
Management, Targit, Yellowfin
Other Other Adaptive Insights, arcplan, Bitam, Chartio, Decisyon, Dundas Data
vendors vendors Visualization, eQ Technologic, InetSoft, Infor, Jedox, Jinfonet
Software, Kofax (Altosoft), Lavastorm Analytics, Looker, Manthan,
Phocas, Sisense, SpagoBI
BI = business intelligence; MQ = Magic Quadrant

Source: Gartner (August 2015)

Source: Gartner (August 2015)

The 2015 Magic Quadrant Customer Survey

The 2015 Magic Quadrant customer survey results used in this analysis included a total
of 1,783 (2,083 responses less OEM responses) received from:

Vendor-provided references (1,535, or 74%)

Survey responses from BI users from Gartner BI Summits (48, or 2%)
Respondents from last year's survey (190, or 9%)
OEM references (310, or 15%) were excluded from this analysis as it is focused on
end-user buyers
We include vendors in this research that did not meet the Magic Quadrant inclusion
criteria for an actual position in the Magic Quadrant, but that received at least 12
completed customer survey responses. These are reported in aggregate as "Other
Vendors" in the figures depicted throughout the analysis (see Note 1 for the inclusion
criteria for a position on the "Magic Quadrant for BI and Analytics Platforms").

Note that not all survey respondents completed the BIPOC questions. Hence the
responses for each cost category are different and less than 1,783. We note the number
of responses in red next to the segmented cost numbers. Results with less than 30
responses should be considered directional. This data was collected as part of the 2015
BI and Analytics Magic Quadrant process between October 15 and November 15 in
2014.

Also note that data quality was addressed by removing outliers. We could not control for
differences in deployment types. For example, TIBCO Spotfire and SAS deployments
may have advanced analytics included, or not, or Microsoft deployments that include
SQL Server may or may not have data warehouse FTEs counted in the total.

Data modeling and preparation is a large part of a BI and analytics deployment, but the
degree that organizations include this in their BI cost, or treat it separately as a data
warehouse and storage cost, varies significantly. The survey did not control for this.

Data Insights

Overall BIPOC Observations

Buying low license cost platforms does not always translate into low BIPOC over time
or into higher business benefits.

Figure 2 shows that, while overall three-year BIPOC per user goes up as deployment
sizes increase, as you would expect, per user costs go down substantially from $83,228
per user for deployment sizes of less than 100 users to $4,061 for deployment sizes
with greater than 500 users. The average BIPOC cost per user across the survey is
$47,525.

Figure 2. BIPOC per User by Deployment Size

BIPOC = business intelligence platform ownership cost

Source: Gartner (August 2015)

Licenses and recurring maintenance represent the most visible components of BI

platform cost, and most enterprises (shortsightedly) focus the bulk of their vendor
comparison effort here. However, Figures 3 and 4 show that the ongoing cost of IT FTEs
to manage the deployment and support content creation makes up the bulk of three-
year BIPOC — 78% on average across the survey and 79% of BIPOC for deployments
under 100 users, but decreasing to 55% of three-year BIPOC for deployments above
500 users. This suggests there is the potential for significant IT FTE efficiencies per
additional user supported.

Conversely, for the largest deployments above 500 users, the cost per user for license,
maintenance and recurring license fees, initial implementation costs for external service
providers and ongoing external service provider implementation costs are at least
double as a percentage of three-year BIPOC when compared to deployments below
100 users.

Lower than average IT FTE costs per user — due to large content consumer-intensive
deployments — result in a favorable overall cost per user profile.

IT FTEs are more cost-efficient per user served than external service providers.

Figure 3. BIPOC Cost Components by Deployment Size

BIPOC = business intelligence platform ownership cost; FTE = full-time equivalent

Source: Gartner (August 2015)

Figure 4. BIPOC Components as a Percentage of Total BIPOC by Deployment Size

BIPOC = business intelligence platform ownership cost; FTE = full-time equivalent

Source: Gartner (August 2015)

Figure 5 compares BIPOC components by vendor category, while Figure 6 details cost
components both by vendor category and vendor. BIPOC components by vendor
category, vendor and deployment size are detailed in the remainder of the report.

Figure 5. Three-Year BIPOC Components per User by Vendor Category

BI = business intelligence; BIPOC = business intelligence platform ownership cost; FTE

= full-time equivalent

Source: Gartner (August 2015)

Figure 6. Three-Year BIPOC Components by Vendor Category and Vendor

BI = business intelligence; BIPOC = business intelligence platform ownership cost; FTE
= full-time equivalent; MQ = Magic QuadrantThis data represents customer survey
responders' views, not Gartner's.

Source: Gartner (August 2015)

While it is often assumed that open-source vendors offer low-cost solutions, in fact they
have the highest three-year BIPOC per user at all deployment sizes largely due to a
higher IT FTE cost per user and implementation costs. They are often deployed by IT for
embedded application development, which could in part explain this result. Notably,
license costs are also above average, as most buyers use the commercial versus free
versions of open-source software. OpenText (Actuate) has higher IT FTE costs than
Hitachi Data Systems (Pentaho), but lower cost per user results in all other cost
categories (see "Benefits and Compromises of Open-Source and Corporate Software
Suites for Advanced Analytics").

Modern BI platforms require less IT skills. Data discovery leaders have the lowest three-
year BIPOC per user, largely driven by a lower requirement for IT FTEs and external
implementation services. This is likely due to more favorable product ease of use.
Unexpectedly, customers also report paying less for data discovery tool licenses than
other types of vendors. Qlik has the lowest overall three-year BIPOC result with the
lowest IT FTE costs per user, while TIBCO has the highest, perhaps due to the inclusion
of Jaspersoft, a more IT-centric traditional reporting tool. The frequent use of TIBCO
Spotfire for more advanced analytics could also explain a higher BIPOC. Tableau falls in
between the two.

The cloud BI vendors in the survey also feature easier-to-use, modern BI platforms.
They have a below-average three-year BIPOC with a very favorable (the lowest)
additional infrastructure and hardware cost per user profile — as you would expect —
and low initial implementation costs (second to the data discovery vendors). With cloud,
while users need to develop initial content, they do not have to set up and deploy the
software. While Birst customers report lower license, maintenance and initial
implementation costs per user than GoodData, ongoing external service provider and IT
FTE costs are higher, resulting in an overall higher BIPOC per user.

Like the data discovery vendors, on average, Figure 7 shows that cloud BI vendors also
have favorable ease-of-use scores and their customers report achieving above-average
business benefits — despite a slightly higher BIPOC, Birst has higher scores than
GoodData on both measures. The often centralized deployment model (IT or a power
user creates dashboards for business users) of the cloud BI vendors may explain why
costs per IT FTE are below average, but higher than for the data discovery vendors.
Cloud BI IT FTE per user costs are also higher than for megavendors and large
independent vendors, even though they are easier to use and result in higher business
benefits — perhaps in part because of smaller deployment sizes with fixed costs being
spread over a smaller user base.

Figure 7. Ease of Use, Achievement of Business Benefits and Three-Year BIPOC by

Vendor Category
BI = business intelligence; BIPOC = business intelligence platform ownership cost; FTE
= full-time equivalent
This data represents customer survey responders' views, not Gartner's.
Composite Ease of Use Score is on a scale of 1 to 10 and is an average of Ease of Use
for Content Authors, Ease of Use for Content Consumers and Ease of Use of
Implementation and Deployment scores.
Composite Business Benefits Score is on a scale of 1 to 10 and is an average of 11
business benefits scores.

Source: Gartner (August 2015)

Large independent and megavendors have a below-average three-year BIPOC per user.
While initial license, maintenance, infrastructure/hardware, initial and ongoing
implementation costs are at or above average, lower than average IT FTE costs per user
result in a favorable overall cost profile. Even though these vendors are IT-centric and
are more difficult to use, they have the largest and often most complex deployments,
which can explain their higher implementation costs. They are also deployed largely for
traditional BI, where a small number of centralized content developers (often in IT) serve
a large number of content consumers or viewers. The incremental license — as a result
of server and role-based pricing — and the IT cost of serving an additional viewer are
often very favorable. These platforms are also designed to support large enterprise-
scale deployments.

Of the megavendors, while many customers buy Microsoft because of low cost of
license, it in fact has the highest three-year cost of ownership; SAP, meanwhile, has the
lowest. This can perhaps be explained by the fact that, until the introduction of Power
BI, Microsoft BI components have been one of many uses for SQL Server, SharePoint
and Excel (for example, organizations use SQL Server for data warehousing and
SharePoint for knowledge management in addition to BI). It may be difficult for
Microsoft customers to separate out and report the costs and IT FTEs associated
specifically with BI versus other solutions using the same components. SAP also has
larger-than-average, report-centric consumer deployments that tend to be at a lower
cost per viewer served.

The small independent vendor category is higher than average across all cost per user
components with the exception of license and maintenance/recurring fees. This can in
part be explained by smaller deployment sizes, but there is also a high degree of
variability among specific vendors. For example, Alteryx has the highest overall BIPOC,
despite high ease of use scores, but it also has an average deployment size of only 55
users. Its use by primarily power users for self-service data preparation and advanced
analytics in support of broader data discovery deployments could explain this result.

Datawatch, a niche data discovery vendor, has the second highest BIPOC in this
category in part due to higher license costs per user and to higher IT FTE costs per
user. It often deals with unstructured, messier and more difficult data. Like Alteryx,
Datawatch also has among the smallest deployment size in the survey.

Prognoz is another high BIPOC vendor despite strong ease-of-use scores and larger
deployments. This is likely due to costs for external implementation services and IT
FTEs for building end-to-end analytic applications, included ETL and advanced
analytics (versus simple reporting), which are in the bulk of Prognoz deployments.

At the other end of the spectrum, Panorama Software and Pyramid Analytics have the
lowest BIPOC in this category, but higher than the data discovery vendors. This could in
part be due to data modeling (mostly in cubes) often happening outside these
Microsoft-centric platforms.

Other vendors are second only to open source in three-year BIPOC. All cost categories,
with the exception of initial internal implementation and hardware cost per user, are
higher. Like the small independent vendors, the individual vendor results have wide
variability. They also have smaller deployment sizes. That said, many are also newer
vendors with limited external marketplace expertise. Many of these platforms are also a
work in progress. These factors can contribute to higher external implementation costs
per user. Within this category, SpagoBI, Lavastorm Analytics and Adaptive Insights
customers report the highest three-year BIPOC, while eQ Technologic, Decisyon and
Jinfonet customers report the lowest. Moreover, Lavastorm Analytics, like Alteryx, is
primarily used for self-service data preparation for a small number of users.

Recommendations:

Manage IT productivity through ongoing skills development and fit-for-purpose

tools. Deliver self-service capabilities to business users, along with the necessary
user enablement and support, to make them successful with minimal IT
involvement.
Don't focus solely on initial price negotiations; also manage the cost of
incremental pricing as deployment sizes grow, and manage maintenance increases
over time.
Evaluate the cost of making large initial purchases with high discounts, but then
paying maintenance on software not deployed for a period of time, if ever, versus
buy-as-you-go at a prenegotiated, but perhaps higher-per-user, price (although
you could negotiate a lower price as volume increases over time). The latter
approach has additional value as it is more agile and keeps your options open if
business requirements and technologies change. You might be willing to pay more
for that option.
Develop hybrid internal/external strategy that leverages the skills and
competencies of internal resources as much as possible and augment with
external service providers as needed.
Consider agile approaches to external service engagements, which can get more
complex as deployment sizes grow. Assess trade-offs and accelerate knowledge
transfer between internal resources that are fixed but appear to be more efficient
as deployment sizes grow versus less permanent, more costly external service
providers. Often, the external providers are more expensive because of specialized
skills (like install, performance tune) versus a building reports/outsourcing
 approach.

License Costs

Licensing cost is one of the most visible and easy-to-define components of total cost of
ownership. However, as vendors price and package their products differently — some
as named user, some as server-based, and some as enterprise solutions — assessing
cost per user can be a challenge. Figure 8 shows that the overall average BI license
cost per user is $2,128.

While there are perceptions that certain BI products are cheaper than others, the survey
results do not support some of the perceptions. For example, open-source software
may be free for limited capabilities, but customers who subscribe to the premium
capabilities have similar license fees to commercial BI vendors. Customers often
perceive data discovery vendors as being significantly cheaper than BI platform
vendors; on average, this appears to be true, with the average cost for data discovery at
$1,195 versus megavendors at $2,263. There are further differences based on the size
of the deployment across all vendor categories, with higher licensing costs at
deployment sizes of 100 users or less, and greater discounts with deployments over
500 users.

While overall license costs go up as deployment sizes increase, as you would expect,
per-user costs go down substantially. This could be due to any combination of higher
discounting, a higher percentage of "viewer"-type licenses in the deployment (which are
less expensive on a per-user basis than content developers), and/or more scalable
pricing models such as role-based pricing and server- and core-based pricing that bring
the economics of per-user pricing down as user volumes increase.

Figure 8. Initial License Cost per User per Vendor Category

BI = business intelligence; MQ = Magic Quadrant

Source: Gartner (August 2015)

Figure 9 shows interesting differences in license prices per region expressed in USD
terms. While some variation can be explained by currency conversion, Western Europe
and Asia/Pacific customers appear to be paying almost 80% more per user than their
Central and Eastern European, Latin American, and Middle East and African
counterparts. North American customers are in between.

Figure 9. License Cost per User by Region

BI = business intelligence

Source: Gartner (August 2015)

Figure 10 shows differences in license costs paid by industry. The education industry
pays lower per use license fees on average, but government customers do not,
although both in theory would have access to the same General Services Administration
(GSA) pricing, at least in North America. Meanwhile, analytics-intensive industries like
communications, banking, insurance, retail and utilities appear to pay the highest
license fees on a per-user basis.

Figure 10. License Cost per User by Industry

Source: Gartner (August 2015)

Figure 11 shows average license cost per user by deployment size by vendor. Within the
cloud BI vendor group, average license costs for Birst are approximately half the
average cost reported by survey respondents for GoodData and approximately half the
cost for the survey average. TIBCO customers showed much higher license costs than
other data discovery vendors, but it's important to note that most of the TIBCO survey
respondents were on a perpetual licensing model; TIBCO Spotfire transitioned fully to a
subscription model in 2015 with lower licensing rates.

In Gartner inquiries, Tableau customers express frustration at the vendor's unwillingness

to discount prices and while Tableau does not discount, its net per user pricing is the
lowest for data discovery vendors at $953, approximately 10% lower than Qlik's.

Of the large independent vendors, MicroStrategy is the most competitively priced at

$1,323 average per user, which indicates that MicroStrategy's new pricing and
packaging implemented in mid-2014 is making the vendor more competitive with both
larger independents and data discovery vendors.

SAS customers have cited high licensing costs as a concern and the survey shows that
the average SAS license cost per user is the highest of any MQ vendor at $6,531,
although lower in the "Greater than 500" user deployment range, with an average per
user cost of $619. Note that SAS did introduce new pricing for visual analytics in mid-
2014, which appears to be more competitive.

Of the megavendors, customers often perceive Microsoft to be "free" because they

already own many of the components that comprise the BI capabilities (SharePoint,
SQL Server, and Excel); however, when these components are purchased specifically for
BI purposes, the license cost is higher than the overall average at $2,286 per user, with
one of the highest licensing costs for less than 100 users at $6,375. Microsoft
customers who were on a subscription license such as Power BI show a lower licensing
cost. New Power BI pricing for the July 2015 release of $120 per user per year should
have a downward impact on pricing moving forward.

Recommendations:

Use the averages per vendor group and deployment size to develop an initial
budget.
Consider vendor pricing models, product packaging, product scalability (which
determines how many cores in server-based pricing) and deployment size, which
affects discounting when evaluating the license costs of the specific products and
projects.
Plan your deployment in phases, ideally locking in discounts for future expansion.
If you are a global company operating in many geographies, consider requesting
quotes in lower cost regions.
Use average license cost per vendor as a point of reference and negotiation in
vendor contract negotiations.
Consider additional costs beyond licensing that impact cost of ownership; do not
let a low license cost negate higher costs to implement and support.

Figure 11. Initial License Cost per User by Vendor and Deployment Size Range
BI = business intelligence; MQ = Magic Quadrant
This data represents customer survey responders' views, not Gartner's.
Number of responses is shown in parentheses in red. Results with less than 30
responses should be considered directional.

Source: Gartner (August 2015)

Maintenance and/or Recurring Annual License Costs

Figures 12 and 13 show average annual maintenance and/or recurring license costs
paid by survey customers by vendor category, vendor and deployment sizes. Annual
maintenance fees for perpetual licenses are charged either as a percentage of list price
or percentage of the negotiated discount, generally ranging from 22% to 28%. With
subscription-based licensing, the maintenance fee may be lower in subsequent years or
may be a full subscription price. For example, GoodData customers are essentially
paying the full license fee in subsequent years, whereas Birst customers are paying
51% of the first-year license and SAS customers 43%.

Figure 12. Maintenance and/or Recurring Annual License Costs per User by
Deployment Size Ranges by Vendor Category

BI = business intelligence; MQ = Magic Quadrant

Source: Gartner (August 2015)

Figure 13. Maintenance and/or Recurring Annual Costs per User by Vendor and
Deployment Size Ranges
BI = business intelligence; MQ = Magic Quadrant
This data represents customer survey responses views, not Gartner's.
Number of responses is show in parentheses in Red. Results with less than 30
responses should be considered directional.

Source: Gartner (August 2015)

Recommendations:
 Understand various support options in your maintenance fees.
When possible and publicly available, consider vendors' spending of total revenue
on research and development and on continuous product improvement.
When negotiating with cloud vendors, understand the differences in first-year
contract, and whether subsequent years are discounted. Also understand and
factor in options.
Negotiate caps on maintenance increases where possible, as this is a key driver of
this cost component the longer the relationship with the vendor.

Initial Implementation

Figures 14 and 15 show initial implementation costs per user for external service
providers by vendor category, vendor and deployment sizes. Implementation costs
include installing the software and hardware (if on-premises), configuring and tuning
software, designing a metadata layer, building reports and dashboards.

Organizations may do this with internal resources or external service providers, provided
either by the vendor or by a partner. Data modeling and preparation is a large part of a
BI and analytics deployment, but the degree that organizations include this in their BI
cost or treat it separately as a data warehouse and storage cost varies significantly. The
survey did not control for this. For example, Tableau can readily integrate with existing
data warehouses; QlikView can integrate with an existing data warehouse but may also
act as a type of data mart with merging of data from multiple systems. This may in part
explain why services to implement Tableau are significantly lower at $230 per user
versus Qlik at $1,043 per user.

While there may be variances in degree of data modeling and application building that
account for differences in external consulting costs, the overall complexity to install and
configure the application, as well as build the content, also accounts for differences (see
Figure 7 Ease of Use scores). Of the megavendors, Oracle has one of the lowest ease-
of-use scores and one of the highest per-user cost for external services at $3,159 per
user.

Figure 14. Initial Implementation Cost per User for External Service Providers by Vendor
Category and Deployment Size Range
BI = business intelligence; MQ = Magic Quadrant

Source: Gartner (August 2015)

Figure 15. Initial Implementation Cost per User for External Service Providers by Vendor
and Deployment Size Range
BI = business intelligence; MQ = Magic Quadrant
This data represents customer survey responses views, not Gartner's.
Number of responses is show in parentheses in Red. Results with less than 30
responses should be considered directional.

Source: Gartner (August 2015)

Recommendations:
 Realistically assess the availability and expertise of internal resources to provide
implementation and delivery services; when there is limited bandwidth or expertise
from the existing BI staff, budget accordingly for external assistance.
Compare consulting costs from the BI vendor as well as certified partners.

Initial Implementation Cost per User for Internal Resources

As shown in Figure 16, the average initial implementation cost per user reported by
reference organizations is $1,244 — approximately 60% of the average per user license
cost noted above. Internal implementation costs are highest in the open-source vendor
category, most likely due to the considerable development effort required for upfront
manual configuration required in these products compared to commercial offerings. It
could also be because open-source tools are often specifically chosen for extensibility
and desire to customize. This is the case for the open-source vendor category overall
and also for deployments of 500 users or less.

On the other end of the spectrum, data discovery products required the lowest level of
internal resource investment for initial implementation at $337 per user, which is
expected as products in this category focus on enabling rapid deployments and quick
time to value without the need for significant upfront work to prepare the platform for
use. The above noted assumption — that initial implementation may include more than
software/hardware installation and configuration — is highlighted in the per-user cost
$752 reported for cloud BI products, which typically require little, if any, implementation
effort by internal resources.

Figure 16. Initial Implementation Cost per User for Internal Resources by Vendor
Category and Deployment Size Range
BI = business intelligence; MQ = Magic Quadrant

Source: Gartner (August 2015)

As shown in Figure 17, references with fewer than 100 users report a wide range of
initial implementation costs across the vendor categories and individual vendors. The
highest initial implementation costs per user for deployments of 100 users or less are
reported in the large independents, megavendors and open-source categories, with the
lowest being reported in cloud BI and data discovery.

However, references using platforms in the high-cost categories for small deployments,
such as Information Builders (large independents), Hitachi Data Systems (Pentaho)
(open source) and OpenText (Actuate) (open source), which support large external
customer-facing deployments, also report lower than average spend in deployments
greater than 500 users for these platforms. This is not the case for platforms in the
megavendor category, in which references, on average, report spending $250 per user
— twice the MQ vendor average for deployments with over 500 users.

Vendors that focus on self-service data preparation, specifically Alteryx and Lavastorm
Analytics, are used in a different capacity than other BI vendors and are generally
deployed to a small number of users internally, which makes comparison to platforms
that are deployed more broadly difficult. This should be considered when evaluating
these platforms and the seemingly high per-user initial implementation cost.
Figure 17. Initial Implementation Cost per User for Internal Resources by Vendor and
Deployment Size Range

BI = business intelligence; MQ = Magic Quadrant

This data represents customer survey responders' views, not Gartner's.
Number of responses is shown in parentheses in red. Results with less than 30
responses should be considered directional.
Source: Gartner (August 2015)

Recommendations:

Develop a hybrid internal/external strategy that leverages the skills and

competencies of internal resources and augments with external service providers
as needed.
Plan for growth in the implementation phase and proactively consider scalability
requirements both from a user and data perspective to avoid incurring additional
postimplementation costs.

Initial Cost for Infrastructure (Hardware)

Infrastructure cost is generally inclusive of servers that are needed to operate the BI
platform required for metadata management, Web applications, content sharing, in-
memory analytics, etc. The server requirements vary significantly by vendor and depend
on several factors that include underlying platform architecture, memory requirements,
and support for self-contained extraction, transformation and loading (ETL)/data storage
within platform.

The organization's deployment type and user demographic will impact the infrastructure
cost dramatically. Some vendors will opt for multiple environments to support their
initiative, which requires a development, test, QA and production server. Some
organizations require high availability and redundancy to support disaster recovery,
which will also increase initial infrastructure costs. A higher mix of complex analytical
workloads and concurrency will require some organizations to scale up or out in order to
meet demands of their end users. In addition, costs related to enterprise data
warehouse data storage, interim datamarts and online analytical processing (OLAP)
cubes have (potentially) been included as initial infrastructure costs attributed to the BI
platform. These should be taken into consideration, particularly for products where the
presence of a data warehouse or OLAP environment is a prerequisite to using the
platform.

As shown in Figures 18 and 19, survey references using megavendor BI platforms report
initial infrastructure costs of $1,387 per user, which is nearly twice the overall survey
average of $786. Platforms in this category are often designed and sized to support
high levels of concurrency across large deployments in support of system of record
reporting.

As megavendors continue to introduce more advanced and interactive analytic

capabilities through data discovery offerings, the need for servers with sufficient
memory and compute power will increase the initial level of investment for new
deployments and existing customers seeking to benefit from new platform capabilities.

The initial infrastructure costs reported within the data discovery category with
deployments of fewer than 100 users is higher than expected, given that many
organizations start with desktop deployments to build momentum in this deployment
size range, which does not require infrastructure investment.

Figure 18. Initial Cost for Infrastructure (Hardware) per User by Vendor Category and
Deployment Size Ranges

BI = business intelligence; MQ = Magic Quadrant

Source: Gartner (August 2015)

Figure 19. Initial Cost for Infrastructure (Hardware) per User by Vendor and Deployment
Size Ranges
BI = business intelligence; MQ = Magic Quadrant
This data represents customer survey responders' views, not Gartner's.
Number of responses is shown in parentheses in red. Results with less than 30
responses should be considered directional.

Source: Gartner (August 2015)

Cloud vendors, as you would expect, have the lowest per user initial infrastructure cost.
The higher than average initial infrastructure cost noted above in the megavendor
category is driven primarily by Oracle and Microsoft, both with average initial
infrastructure costs exceeding $2,000 per user compared to a vendor average of $786
per user.

The highest cost vendor overall is Prognoz, which delivers a wide range of capabilities
and offers an end-to-end platform spanning ETL and data warehouse development
tools on the back end, and a full spectrum of analytic tools on the front end.

SAS is an example of a vendor that has products requiring robust in-memory

capabilities to support the level of complex analysis delivered by its products, such as
SAS Visual Analytics, which uses the SAS LASR Analytic Server to deliver performance
that is expected from a tool that offers interactive visual analytic capabilities. The
expectation for "speed of thought" interactive response through interactive data
discovery tools is a trend that will certainly continue and will increase the level of initial
and incremental infrastructure investment for organizations.

Recommendations:

Plan for increased investment and server upgrades as vendors introduce more
advanced analytic capabilities requiring additional memory and processing power.
Consider cloud BI as an option to reduce infrastructure costs and to take
advantage of product innovations without the need to upgrade existing
infrastructure.

Ongoing External Service Provider Cost

Organizations that have engaged an external service provider initially to implement a BI

platform often extend the engagement beyond deployment further into the life cycle of
the program. In some cases, organizations may decide to engage with an external
provider to manage maintenance and small enhancement activities while the internal
team focuses on larger development efforts. Depending on the nature of work, the costs
may be incurred under a master service agreement (MSA) where covered work is
defined and performed as part of the agreement, or on a time and materials (T&M) basis
where resources are provisioned and assigned on an as-needed basis by the service
provider or consulting firm.

As shown in Figure 20, the average annual spend on a per-user basis for external
service providers for ongoing involvement is $696, which is roughly half of the initial
external service provider implementation cost. Consistent with the pattern observed in
the cost analysis of external providers for initial implementation, survey references using
data discovery products spend the least of all categories on ongoing external service
provider engagements.

Per-user cost is relatively consistent across the other MQ vendor categories, ranging
from the megavendor average of $533 at the low end to the large independent average
of $696 at the high end.

Within the megavendor category, however, it should be noted that the per-user cost of
external service providers does not follow a similar pattern as the other categories in the
101 to 500 deployment size, all of which observed a significant drop in per-user cost
compared to deployments with 100 or fewer users. This particular segment reported
significantly higher-than-average spend on external service providers for initial
implementations, so it is likely that many organizations decide to engage in ongoing
support given the complexity and technical expertise to support many of the platforms
included in this category.

Figure 20. Ongoing Cost per User for External Service Providers by Vendor Category
and Deployment Size Ranges

BI = business intelligence; MQ = Magic Quadrant

Source: Gartner (August 2015)

Figure 21 shows annual ongoing external service provider costs per user.
While the data discovery category incurred the lowest overall cost of any vendor
category for ongoing external service provider cost per user, per-user costs vary
considerably across the vendors within the category itself.

Tableau references report very low ongoing external service provider costs across all
deployment sizes, which indicates that most organizations find that the platform can be
managed appropriately internally without the need for external services.

Qlik generally requires more postimplementation involvement given its data modeling
and dashboard development capabilities, which tend to require more technically skilled
resources than Tableau. The contrasting external service provider costs for these two
vendors could also be explained by the difference in how both are used — Tableau is
often used when a data warehouse already exits while Qlik is often used as datamart
and dashboard application, with ETL. It should be noted that Qlik Sense, which was not
generally available (GA) when the survey was conducted, may change this dynamic and
reduce the external service provider dependency for Qlik customers.

Surprisingly, vendors with a rich network of implementation and system integration

partners, such as IBM and Microsoft, are among the lowest in terms of cost related to
ongoing support from external service providers.

References using other large vendors with similar multiproduct, complex portfolios —
such as SAS, Oracle and SAP — report significantly higher overall spend on external
ongoing support. This may be due to greater availability of internal resources for
ongoing maintenance of IBM and Microsoft solutions, or a different mix of users and use
cases across these vendors requiring a different level and type of ongoing support.

Figure 21. Ongoing Cost per User for External Service Providers by Vendor and
Deployment Size Ranges
BI = business intelligence; MQ = Magic Quadrant
This data represents customer survey responders' views, not Gartner's.
Number of responses is shown in parentheses in red. Results with less than 30
responses should be considered directional.

Source: Gartner (August 2015)

Recommendations:
 Extend engagements with external service providers only if the initial
implementation was a success and there continues to be a skill or bandwidth gap
postdeployment.
Focus efforts on knowledge transfer from external service providers to internal
staff to lessen long-term expenditures and develop internal support competencies.

IT FTE Costs

IT full-time equivalent (FTE) staffing costs make up for close to 78% or more of the
overall three-year BIPOC and these costs are recurring, hence BI leaders need to focus
considerable attention here.

Contrary to the popular myth and practice of companies shortlisting and engaging with
BI vendors solely on the basis of the license costs, our survey analysis clearly shows
that the IT FTE costs make up for a sizeable portion of the overall BI spending of an
organization over a three-year period across all vendors and, most importantly, across
all deployment ranges, making this the most significant component and driver of the
overall BIPOC.

IT FTE costs are often the most difficult to define, quantify and capture as different
organizations may have their own methodology and definition of whom constitutes an IT
FTE. We define IT FTEs as the number of full-time equivalents who author BI content
(including reports, analysis, visualizations, dashboards and data mining models) and/or
administer and support users (for example, support, semantic layer/data models,
configuring and tuning software, defining a metadata layer, or performing testing of
analytic workloads). We use an average salary of $114,121 to calculate annual IT FTE
costs (see Note 2).

It should also be noted that the "initial implementation cost" for internal resources
components of BIPOC (defined earlier) may include some of the same resources that
were involved in the IT FTE cost calculations, depending on how the reference
organization defines, captures and reported these costs.

Figures 22 and 23 show the average FTEs broken down by vendor category, vendor and
deployment size ranges per year. The average number of FTEs across all deployment
size ranges is close to six FTEs per year. Survey references for the cloud BI vendors
mentioned in our survey report the lowest number of FTEs per year with an average of
2.7. Among the cloud vendors, GoodData users report the lowest FTEs per year across
all deployment ranges and noticeably the number of FTEs per year reduces to 1 in the
101-500 deployment range.
It should also be noted that in many organizations, BI content developers reside in the
line of business and would not necessarily be counted in these numbers.

Figure 22. Annual IT FTE Counts by Vendor Category

BI = business intelligence; FTE = full-time equivalent; MQ = Magic Quadrant

Source: Gartner (August 2015)

Figure 23. Number of FTEs by Vendor and Deployment Size Ranges per Year
BI = business intelligence; FTE = full-time equivalent; MQ = Magic Quadrant
This data represents customer survey responders' views, not Gartner's.
Results with less than 30 responses should be considered directional.

Source: Gartner (August 2015)

At the other end of the spectrum, the user references of megavendors report the highest
number of FTEs needed per year on an average — 11.2 FTEs — and also report very
high FTEs across the higher deployment ranges of above 100 users. Notably, for
megavendors in deployments ranging 101-500 users, the number of FTEs is
approximately double the survey average. This could be because most of the
megavendors have high FTE use cases and projects in these higher deployment size
ranges, and economies of scale per additional user added really accelerates after 500
users. Another reason could be that, for some megavendors like Microsoft, customers
often attribute their own data warehouse or information management FTEs as BI FTEs,
since data warehouse is also a part of the SQL Server product. Finally, this high number
of FTEs could also be attributed to the fact that most deployments using megavendors
or large independents require the building of OLAP cubes for each BI project, which
would mean that organizations would require more IT FTEs to build them, thereby
driving up overall IT costs. These reasons can explain the sudden jump in the FTEs per
year for vendors like Microsoft in the megavendor category for 101-500 users, or
MicroStrategy in the large independents category in the "Greater than 500" deployment
category.

Despite this, Figures 24 and 25 show that, because of their large deployment sizes at
the high end (on a per-user basis at deployments above 500 users), per-user IT FTE
costs for megavendors are actually significantly less than the survey average. These
results stem in part from the enterprise administration features of traditional BI platforms
for supporting large numbers of users and large deployments. Very large user volumes,
particularly the incremental cost of supporting a viewer or report consumer, allow
economies of scale for administration.

Data discovery leaders report low average FTEs per year across all categories — with
an average of 4.6 (Figure 23) — which is expected given that they promote self-service
BI and data discovery and generally have platforms that are easy to build, use and
deploy by IT, as well as business users with limited BI skill sets.

One obvious outlier in the data discovery leaders is TIBCO, for which reference
customers report close to 16 FTEs per year in the "Greater than 500" range, which is
significantly higher than the overall average in this deployment category, but is more in
line with Qlik in this deployment range on a per-cost user basis. This is because the
average deployment sizes for TIBCO in this range are larger than for either Tableau or
Qlik.

Figures 24 and 25 show the IT FTE costs broken down per user by vendor category,
vendor and deployment size ranges. The average IT FTE costs per user reported by
reference organizations is $12,316 — approximately 78% of the total three-year BIPOC
costs per user noted in the above sections.

Annual IT FTE costs per user are highest in the open-source vendor category at $18,118
(almost 1.5 times the average annual IT FTE costs across all vendor categories), most
likely due to the considerable development effort required for upfront manual
configuration needed for these products compared to commercial offerings. This is also
the case for the open-source vendor category on the whole and also for midsize
deployment ranges of 100 or less users and 101-500 users.

At the other end of the spectrum, data discovery products required the lowest level of
annual IT FTE cost investments at $4,246 per user, which is expected as products in
this category are known to be user-friendly in terms of self-service and also focus on
enabling rapid deployments and quick time to value without the need for significant
upfront work to prepare the platform for use.

Cloud BI vendors, which are otherwise perceived to be relatively inexpensive due to

their subscription-based licensing, have above-average IT FTE costs per user of
$15,329 in the "100 or less" user deployment category, which means that even though
they are being rapidly used by organizations for new innovative BI use cases, the overall
annual IT FTE per user costs turn out to be higher than average, rendering the cloud BI
project more expensive on a per-user basis over a longer period of time. Cloud BI
vendors, on average, currently have smaller deployment sizes than other vendor types.

Another noteworthy category is the large independent vendors, which exhibit above-
average annual IT FTE costs per user across all deployment ranges and categories. This
could be due to several factors including complex use cases and deployments being
undertaken by these large independents. Alternatively, it could be because the vendors'
BI platforms are not as easy to use, which would require additional training to be
provided to the internal IT FTEs. It could also be due to a shortage of skill sets or
trained IT FTEs for the vendors' BI platform, which would make the overall procurement
of skills more costly.

Figure 24. IT FTE Cost per User by Vendor Category and Deployment Size Ranges
BI = business intelligence; FTE = full-time equivalent; MQ = Magic Quadrant

Source: Gartner (August 2015)

As shown in Figures 24 and 25, references with fewer than 100 users report a wide
range of IT FTE costs per user across the vendor categories and individual vendors. The
highest IT FTE costs per user for deployments of 100 users or less are reported in the
open source, small independents and large independent vendor categories, with the
lowest being reported in the data discovery and cloud BI categories.

IBM and Microsoft, however, have much higher IT FTE costs compared to the other
megavendors in the survey average and in the "100 or less" user category, and
Microsoft also in the 101-500 user category. However, the IT FTE costs per user for
these megavendors go down substantially as deployment sizes go up to over 500 users,
mainly due to economies of scale and efficiencies to serve larger numbers of consumer
users. Large independent vendors including SAS and Information Builders, which, like
megavendors, are frequently deployed in large enterprise standardization projects, have
similar IT FTE cost profiles to megavendors, albeit their costs are a bit lower on a per-
user basis at the midrange deployment levels of 101-500 users. These results stem in
part from the enterprise administration features of traditional BI platforms for supporting
large numbers of users and large deployments. Another reason could be that, unlike the
data discovery BI platforms, the platforms of most megavendors and large
independents are perceived as less user-friendly for complex deployments, which
require additional training of IT FTEs or more time in order to produce reports and
perform analysis.

Open-source vendors including OpenText (Actuate) and Hitachi Data Systems

(Pentaho), which are perceived to be high cost by reference users in the "100 or less"
deployment range, report lower-than-average IT FTE costs per user in deployments
greater than 500 users. The same is the case with certain small independents including
Alteryx, Datawatch (Panopticon) and Logi Analytics, which have IT FTE per-user costs
well above average in the "100 or less" user deployment category but exhibit a great IT
FTE per-user cost in the larger deployment scenarios, which is below average. However,
this is not the case for platforms in the large independents vendor category in which
references, on an average, report spending $1,000 per user — almost twice the vendor
average for large deployments with over 500 users.

Among the data discovery leaders, which are the most cost-effective in this category,
Qlik references report very low annual IT FTE costs per user across the "100 or less"
category, whereas Tableau users report the lowest cost for annual IT FTEs across the
"Greater than 500" deployment range, which indicates that most organizations find that
these platforms can be managed appropriately internally without the need for external
services or the need to find highly trained and skilled IT staff. Qlik (particularly QlikView)
generally requires more postimplementation involvement given its data modeling and
dashboard development capabilities, which tend to require more technically skilled
resources than Tableau. This explains its relatively higher costing compared to the other
data discovery vendors in the above 100 users categories.

While there may be variances in the degree of data modeling and application building
that account for differences in IT FTE costs, the overall complexity to use, install and
configure the application, as well as the complexity, of the use cases also accounts for
differences (see Figure 7). Of the open-source vendors, OpenText (Actuate), for
example, has one of the lowest ease-of-use scores and one of the highest per-user cost
IT FTEs at $22,635 per user.

Figure 25. IT FTE Cost per User by Vendor and Deployment Size Ranges per Year
BI = business intelligence; FTE = full-time equivalent; MQ = Magic Quadrant
This data represents customer survey responders' views, not Gartner's.
Number of responses is shown in parentheses in red. Results with less than 30
responses should be considered directional.

Source: Gartner (August 2015)

Recommendation:
 When evaluating the overall costs of BI platforms, extend your analysis beyond
initial license and hardware fees to include IT FTE costs. By far, IT FTE costs
represent the most significant portion of BI platform costs. Managing these costs
is the key to lowering total costs.

Methodology

The online survey was developed and hosted by Gartner to support the BI platforms'
Magic Quadrant analysis. More than 5,280 unique companies were invited to
participate. Vendor-provided references (direct customers and OEMs), participants in
Gartner's BI summit conferences globally, and respondents from last year's survey also
provided data. OEM responses were excluded for the purposes of this analysis, which
focuses on cost of ownership for companies as opposed to independent software
vendors (ISVs).

This data was collected as part of the 2015 BI and Analytics Magic Quadrant process
between 15 October and 15 November in 2014.

This report represents the views of survey respondents.

To ensure the integrity of the survey data, each survey response was verified by
company-respondent email. For survey responses from nonidentified email accounts,
such as Gmail or Yahoo accounts, the respondent was contacted and had to provide
Gartner with a company email address, a company role and other contact information
(this amounted to fewer than five responses), all of which were further vetted and
ultimately included. Only completed surveys were included in the results.

© 2015 Gartner, Inc. and/or its affiliates. All rights reserved. Gartner is a registered
trademark of Gartner, Inc. or its affiliates. This publication may not be reproduced or
distributed in any form without Gartner’s prior written permission. If you are authorized
to access this publication, your use of it is subject to the Usage Guidelines for Gartner
Services posted on gartner.com. The information contained in this publication has been
obtained from sources believed to be reliable. Gartner disclaims all warranties as to the
accuracy, completeness or adequacy of such information and shall have no liability for
errors, omissions or inadequacies in such information. This publication consists of the
opinions of Gartner’s research organization and should not be construed as statements
of fact. The opinions expressed herein are subject to change without notice. Although
Gartner research may include a discussion of related legal issues, Gartner does not
provide legal advice or services and its research should not be construed or used as
such. Gartner is a public company, and its shareholders may include firms and funds
that have financial interests in entities covered in Gartner research. Gartner’s Board of
Directors may include senior managers of these firms or funds. Gartner research is
produced independently by its research organization without input or influence from
these firms, funds or their managers. For further information on the independence and
integrity of Gartner research, see “Guiding Principles on Independence and Objectivity.”