A Primary Matters Total Cost of Ownership/ROI Analysis . . .
|On Software Not Included:||The analysis in this chart above only addresses voice services platform costs that vary between solutions. It does not include the costs of the voice recognition and text-to-speech software. These costs are identical across both solutions. It does include central office & operations costs.|
|On High Availability:||The above chart is not focused on high availability configurations. If this is desired, the costs of both solutions will increase. Due to the high availability features built into the HP OCMP environment, the incremental cost of addressing high availability will be lower than that of alternative platforms.|
|Configurations:||OCMP including call and resource control, SS7 stack, VXML2.0 interpreter and DS3 card on HP-UX, PA-RISC; Mid-range platform; IA32 Windows, Quadspan T1 and associated hardware DSP boards and 3rd party call and resource control middleware, VXML interpreter|
To understand the impact of HP's OCMP on a Service Provider's voice service business, a Baseline Business Plan has been established. The Baseline Plan includes a mid-range, 96 port platform based on Intel/Windows environment with quad-span T1/E1 card including hw DSP and 3rd party call and resource control and VXML interpreter. The impact of using the HP solution is determined by comparing the attributes of the Baseline business with those from a business using HP's platform. Two major comparisons are made:
The main assumptions of the business plan for the Service Provider's voice services business are:
The details specifying the data underlying the Baseline and Scenarios are in Appendix 1, 2 and 4 of this paper.
This analysis compares a Service Provider's Total Cost of Ownership of deploying its voice services business based on the HP OCMP versus the cost of ownership associated with the typical mid-range solution currently deployed. If one views the Total Cost of Ownership associated with this alternative platform, the HP OCMP solution provides a significantly lower long-term total cost of ownership as shown in the following table. Appendix 1 provides the detailed assumptions that drive the conclusions shown in this table.
In this table, a narrowly defined Total Cost of Ownership has been used, with the major cost elements being:
The following table provides a closer look at the underlying dynamics of these total costs of ownership summaries.
HP’s OpenCall Media Platform is designed especially for Service Providers. In particular, it uniquely offers:
The large difference in cost is driven primarily by the scalability of the HP OCMP and the softDSP. The high port density available on the HP platform significantly reduces the space, number of communications links, and operations personnel required (up to 1/3 less than alternate platforms). In addition, the large nature of the HP Solution takes advantage of the ability to have very large pools of ports available to handle applications instead of smaller ones; therefore the port utilization is higher by 5% to 6% for the same levels of service compared to the low-end platforms.
The softDSP translates into lower cost of ownership for two reasons. First, there is no need for specialized DSP chips and boards in the servers since the standard CPU handles the voice processing. Thus there is less complexity in the platform and lower operations, support and upgrade costs. Second, the softDSP, running on standard computer server technology, can take advantage of Moore's Law and the rapidly dropping costs associated with commodity servers.
These factors result in a significant advantage over traditional media platforms in creating a cost effective, high-end voice services business with literally millions of dollars being saved for other business functions.
By adopting a distributed voice service architecture, therefore changing the way an application is deployed, a Service Provider's business opportunities are transformed. In particular, the Service Provider is able to:
These business attributes change the way a Service Provider is able to approach its voice services business, providing major improvements to the Service Provider's customer relationship, revenue and profitability.
The potential impact of these themes is captured in the following table describing this business scenario. Here, the same Service Provider Interactive Voice Services Business is used to compare platform total value of ownership as the foundation of these scenarios (growing from start-up to slightly over 22,000 voice ports and 3.5 million users over a 3-year period for the Baseline). In capturing the influence of these platform advantages, the following assumptions are made about the impact of an OCMP hosted voice services environment.
Business Impact of HP's OCMP High End Platform on a Voice Services Business
|Attributes of HP's Platform that affect the Service Provider's Customer Relationship||Affect on Customer Relationship||Data used in Scenario Comparing Impact of Platform on Voice Service Business|
|Mid-Range System as Voice Services Platform||HP OCMP Platform based business|
|HP's Platform enables a business to lead the market in terms of:
||Increases Average Customer Life Cycle||30 months||Increases steadily to 33 from 30 months over the planning period, leading to an increase in total customers from 1.29 to 1.41 million at the end of the planning period|
|Due to the enhanced ability to pilot applications and the lower deployment costs, a Service Provider can target more market niches and increase usage||Greater Number of Applications per Customer||At start of business, 1.0 applications per user, growing to 1.1 over the planning period.||At start of business, 1.0 applications per user, growing to 1.3 over the planning period.|
|Since a portion of users will use more than one application, and the HP solution makes it easier to offer many applications, users will acquire multiple applications more often.||Higher Average Revenue per Customer||At the start of business, $8.00 per month per user, growing to $9.00 over the planning period||At the start of business, $8.00 per month per user, growing to $10.00 over the planning period|
All of these factors result in the evolution or creation of a voice services business that has an improved chance of success, with greater revenues and profits.
As shown in the following table, the impact of the above listed assumptions on the customer relationship leads to a large positive improvement in the performance of a Service Provider's business. Due to the scalability of the platform, and its ability to easily market test and deploy many different applications, a Service Provider has the flexibility to move quickly in the market taking advantages of business opportunities and turning them into revenues.
HP's OpenCall Media Platform, designed specifically for the Service Provider marketplace, makes a major difference in the ability of a business to increase its revenues and quickly reach profitability.
Platform sizing, standards, the capacity to rapidly deploy applications, and increased reliability combined with the vision to turn these advantages into better customer relationships are paramount to a Service Provider's ability to capitalize on business opportunities available in the voice services market.
HP's development efforts, which are focused on platforms and solutions for the Service Providers marketplace, enables the provider to experience the following:
For service providers who are looking to increase their revenue streams through the introduction of enhanced voice-enabled services, HP's OpenCall Media Platform is an open (adheres to current industry standards, VXML, MRCP), flexible (utilizes HP's SoftDSP technology), carrier grade platform which provides the greatest scalability in terms of voice access and lowest Total Cost Of Ownership unlike alternative solutions in the marketplace. In addition, HP's OpenCall Media Platform allows rapid creation and secure deployment of new services directly in the carrier's network. These are fundamental platform benefits that will allow exponential growth in the voice services market.
The following tables and explanations specify the assumptions used in this Total Cost of Ownership analysis.
Platforms used in this comparison:
Total Investment for 480 Ports: This is the investment in the voice service platform (not including central office expenses) required to acquire 480 ports of voice services capacity.
Ports per Platform: This represents the maximum number of ports that can be supported for a typical configuration.
Number of Platforms for 480 VXML Ports: The number of platforms that must be acquired in order to have 480 ports of capacity.
Total Investment per Port: This is the sum of the hardware and middleware software investment. The middleware software performs the call control, runs the DSP processing environment and provides the VXML interpreter. The cost of the Voice Recognition and Text-to-Speech are included in this price.
Hardware Investment per Platform: This is the cost of the servers and operating system, DSP boards and communications boards required to support the voice processing software.
Percent Uplift for Spares: Most SPs ensure that they have provisioned spare parts to reduce any downtime that may be caused by failures. This percent allocates an additional investment for maintaining such spare components.
Platform Price Change per Year: This shows the change in platform price per year into the future. Since the HP solution is using standard processor platforms, the change in price is quite aggressive. Since the non-OCMP product require addition DSP boards and other communications equipment, the price change is not as aggressive.
Hardware Maintenance Percent: This is the annual maintenance contract expenses paid to the suppliers for hardware support and service.
Platform Software Investment per Port: This software controls the voice processing hardware, thus playing the role of middleware software, and provides the VXML interpreter. It is used by the voice services applications as well as the voice recognition and text-to-speech software resources
TTS License Fee per Port: The software license fee for Text to Speech features.
Percent of Ports with TTS: When one configures a solution, the ports are only using the TTS function part of the time, so one acquires fewer port licenses than the total number of ports. This percent varies by the application that is being supported.
ASR License Fee per Port: This is the software license fee for the Automated Speech Recognition software.
Percent of Ports with ASR: Again, licenses are not necessarily acquired for the all the ports depending on the application.
Total Software Investment per Port: The software investment per port combines the Platform Software Investment with the weighted average of the voice recognition and text-to-speech software licenses.
Software Maintenance Percent: This represents the percent of the software investment on which the annual maintenance fee is assessed by the software suppliers.
These costs are incurred by a Service Provider who installs the solution in a central office environment.
Central Office Foot Print Costs/Sq Foot: This is the monthly cost per square foot assigned to resources for using central office rack space.
Square Foot/Platform: This identifies the number of square feet required for each voice service platform. It is used to determine the monthly cost assigned by the central office to each platform.
Switch Communications Resource Type: This is the type of data communications resource linking the platform to the central office.
Ports per Communications Resource: The number of voice conversations, each requiring a port, that are carried by the communications resource type.
Switch Communications Resource Cost/Month: This is the monthly cost assigned to the voice services platform for the central office switching activities for each of the communications resources used.
Monthly Telecom Cost per Resource Type: This is the monthly cost assigned for the bandwidth and the communications resource linking the central office and the voice services platform.
This represents the Operations Personnel Full Time Equivalent (FTE), i.e. cost of one full-time personnel resource assigned to managing, maintaining, operations and performing back-ups and updates to the voice processing solution. In determining the FTE, the complexity of the platform, in terms of the number of components, software drivers and data communications connections, is the primary factor. The HP OCMP is a very simple configuration, since there are no special-purpose DSP cards required, there is one high bandwidth data connection, and all of the software runs on standard processors. In addition, only a few platforms make available a very large number of ports, thereby reducing the number of Central Office sites required to house voice platforms and minimizing operations personnel required to support the platforms in the CO.
The typical mid-range platform is more complex. The hardware system consists of a number of specialized boards, the solution can be provided by several different vendors, there is a requirement for a greater number of platforms, data links and data communications cards in order to provide an equivalent number of ports. Software updates are more complex and time consuming. These platforms require significantly more skills and knowledge to successfully operate and support.
Minimum Number of FTE: This is the minimum Full Time Equivalent operations people required to fully deploy, manage and support the solution on a 24x7 basis.
Minimum FTEs per Site: This is the allocation of personnel required to handle all of the operations tasks at each site where the voice services platforms are installed.
FTE per Platform: This is the time allocated for handling the operations issues for each platform. For instance, .05 FTE represents 1/20th of a full time operations personnel allocated per platform.
See the discussion in Appendix 2 on traffic engineering and efficiency of port usage due to Erlang principles. This factor focuses on the typical network configurations used for providing access to voice applications. As a result of the significantly higher port size and simplicity of configurations supported by OCMP, assuming the same Quality of Service level, the total port sizes available for dialing applications will be higher than for the mid-size platforms, leading to slightly higher efficiencies due to network traffic characteristics. This factor is ONLY used in the TCO analysis of a business, not in comparing the TCO of platforms with the same port size.
Currently, most SP's acquire a voice services application, to offer to the market, from an application supplier. In doing so, the voice platform comes with the application and it can be used only for that application. The SP estimates the demand and peak usage and then calculates how many ports are going to be required to meet the Quality of Service goals.
The result of this calculation is determined by the size (for instance in number of ports) of the peak demand and the Quality of Service needed by the callers. The Quality of Service (QOS) measure is defined as the percent of callers that, when dialing in, are able to obtain the service they want instead of getting a 'busy' or 'not available' signal. For example, when a SP configures their solution, the decision may state, "I want 99.8% of my calls to be able to obtain service when they dial and only 0.2% to receive a busy signal."
There is a class of equations, Erlang, which enables a SP to determine how many voice ports it needs to meet the desired QOS. If a SP has only a small number of ports providing a service, a voice platform of 96 ports providing access to a single application for example, only 80% of those ports can be in use at peak in order to meet the Quality of Service goals stated in this example of 1 in 10,000 calls receiving a busy signal. If the application specific configuration supports 500 ports, there need to be roughly 60 extra ports in order to meet the quality of service level (a maximum of 88% of the ports can be in use and 12% must be 'waiting for the next call').
VXML-based portals provide access to multiple applications from any port, i.e. voice ports are no longer bound to a particular application. Service Providers benefit from using voice infrastructure that can scale. High end configurations allow upwards of 96% to 97% port usage during peak demand; only 3% or 4% of the ports are 'waiting for the next call' to achieve the same QOS levels as described above (1 in 10,000 calls receiving a busy signal).
For any SP moving from application-bound platforms to a distributed VXML Portal environment, Cost of Ownership will decrease as platform size increases because of efficiencies gained through Erlang traffic models.
There is a wonderful excitement surrounding Voice Services. These services, ranging from messaging, custom greetings and ring tones, information acquisition, customer support, voice mail, and mass alerting, are on the cusp of mass adoption and major growth. The confluence of many different technologies, arriving from different places, is being incorporated into standards-based platforms able to support many different solutions for a variety of users.
Some of the excitement results from the fact that the model of the Web is being used to direct the platforms and standards for creating voice services. Today, someone can create a voice services-based 'web-site', or add voice access to existing Web Sites, within a business framework resembling the Web. The necessary changes in platform architectures are being embodied in products that are available to Service Providers (SPs) as a foundation for new business opportunities.
Two major applications established an industry and foundation
Two major 'killer applications' were the foundation of a market explosion establishing the automated voice services market. Voice Mail and Integrated Voice Response (IVR) customer service moved from R&D and early adoption into industry-wide deployment in the mid-1980s. After battling the sales issues of "Who would want to talk to a machine?" that confronted voice mail in the early 1980s, this communications service became a basic requirement of all white collar workers and most homes by the latter half of the 1990s.
A similar objection faced Touch-Tone user interfaces as well. However, financial pressures on companies in the early '90s forced companies to no longer ignore the cost savings available from automating as many customer interactions as possible. The promised savings were exceeded in most cases.
These markets continue to grow, and there are still aggressive deployments of Integrated Voice Response platforms and applications. As a result, there is now a solid marketplace sized at more than $1 billion a year for equipment and software manufacturers, making investments in automated voice technology and services viable.
Improvements in voice recognition technology is now making it a deployable solution
Many view the Touch Tone User Interface as difficult. Even with its spectacular success, it pales in comparison to talking directly with someone. In the last several years, voice recognition has reached a new peak of quality and cost that, for the first time, makes it a viable technology for industry-wide deployment. Although there has been availability of voice recognition systems since the 1970s, there has been very little deployment. The two major reasons for low penetration have been the limited ability of these solutions to provide speaker-independent speech recognition over the phone, and the high processing requirements needed to support this technology.
Breakthroughs in the quality of speaker independent, speech recognition via the phone have been significant recently, as well as the ability to interpret the caller's queries and link these with relevant responses. These breakthroughs have been significant enough to result in quality voice recognition with accuracies exceeding 98%, with solutions currently in production. In addition, the ability to uniquely identify the caller through name identification and authentication have reached a level of accuracy such that account access is protected to the same degree as is available in other methods. With these two issues solved, and continuing to improve, voice recognition is now at a very acceptable level for major deployment.
Evolving from Specialized Voice Processing Platforms to Standard Servers - the SoftDSP
On the computing side, specialized Digital Signal Processing (DSP) chips, which are dedicated processors used for interpreting voice, have been the norm. These are expensive not only to acquire but also to own, maintain and upgrade, thus leading to a expensive voice services platforms.
The most recent innovations in processing voice are focused on removing the need for the specialized DSPs, instead relying on the power of the host processor (PA-RISC, IA-32 and IA-64 2004) on the computer server for handling these compute intensive tasks. The Soft DSP eliminates the requirement to install specialized DSP chips and boards, thus enabling general purpose servers to handle the processing of voice. This means that
HP’s R&D efforts are at the leading edge of voice processing using industry standard servers and commodity processors for voice processing with its SoftDSP. This technology enables a company to acquire standard servers, running either UNIX or LINUX, to fully support its voice processing requirements, no longer needing to acquire, support and maintain specialized processors for these tasks.
In creating a business to address opportunities in the voice services market place, typically the SP surveys its market and chooses the applications that it wants to use to address the market. These applications are provided by an Application Vendors. Occasionally, an SP will acquire and platform and build their own application.
The Current Approach to Building and Deploying Service Provider Voice Solutions
Currently, voice applications are created, via Voice Application Programming Interfaces (APIs), specific to the underlying voice processing hardware platform chosen. These applications are typically written using telephony-specific, proprietary Service Creation environments which shield developers from these voice specific APIs. However, to enable an application to run on a different hardware platform, it requires development and testing to link the Application Solution to a different hardware platform and set of proprietary APIs. This has led to the stovepipe, monolithic, voice services architecture so commonly seen in today's incumbent service providers.
The Voices Services "Killer Application" may be a successful VMXL Services Environment with Supporting High-End Platforms
VXML facilitates SP's ability to leverage voice resources in their network by enabling the separation of application/service logic from the underlying media resources. Thereby facilitating the reuse of these resources for many applications and reducing the need for application specific voice hardware. VXML enables voice applications to be written in an industry standard language rather than to proprietary API's. The advantage being that application vendors will be able to more easily place their solutions on a variety of platforms, picking either those that are used by the Service Provider as well as those that perform best in the application developer's view. In addition, because VXML enables the use of standard IT development tools, the community of developers creating voice-enabled applications is growing significantly.
HP's OCMP supports this distributed voice services architecture discussed above, so that:
Impact of Platform Size and Port Density on Total Cost of Ownership
When a Service Provider's goal is to create a business that requires tens of thousands of ports to support a user base of several million subscribers, the size of the platform makes a very big difference.
Impact of Port Size and Communications Density on interactive voice platforms
|HP OCMP||A Low End Solution||A Mid-Range Solution|
|Number of Ports||Up to 2,000 per Platform||24 Ports per Platform||96 Ports per Platform|
|Components to maintain||Typical standard multi-processor server components||Large number, including DSPs||Large number, including DSPs|
|Communications Links||High Speed DS3 or OC3, or E1/T1, supporting all ports with one link from the CO||T1 or E1||T1 or E1|
|Density||Very high, due to both high speed communications link||Low Density, requires many servers||Moderate Density|
|Rack-mounting||Can rack up to 5 Servers on 1 central office rack, up to 10,000 ports/rack depending on application||Depending on applications, between 14 and 48 servers per rack, or 336 to 1,152 ports/rack||Typically, about 14 Servers per rack, supporting up to 1,344 Ports|
In addition to the sizing benefits available from a scalable platform, HP's OpenCall Media Platform meets the requirements of a true distributed voice architecture and is specifically designed for Service Providers. It does this by:
An Example Configuration using the HP OCMP VXML Voice Services Platform