VoIP technology is gaining ground in financial and other service industries as a cost-effective means of provisioning additional, enhanced services unsupported by traditional circuit-switched technology. However, enterprises seeking to reap the benefits of VoIP technology must acquire a solid understanding of VoIP-based call-center components and their behavior to plan, predict, and control performance and QoS levels.
VoIP-based call centers present many challenges:
- Integrating traditional telecom infrastructures with the wide array of protocols required to support new, add-on services.
- Quality and reliability issues that are essential but often difficult to control.
- Dependency on network specialists who are expensive and hard to find.
- Achieving high QoS, GoS, and media quality.
Despite these complexities, an increasing number of enterprises are turning to VoIP call centers as a cost-effective solution for delivering the kind of customer experience that strengthens relationships. Growing at a rate of 20% per year, call centers are taking center stage as the critical connection between a company and its customers.
A company’s call center empowers its customers to summon service in real time or near real time via phone, fax, e-mail, website, and IVR. However, the QoS experienced by customers can substantially strengthen or severely try their loyalty.
The key to the success of call centers is the quality and performance they provide. With an understanding of the system’s components and the right test tools, VoIP-based call centers can be implemented with confidence and maintained with integrity.
What Components Require Testing?
Predictive environment testing can preempt many of the problems associated with various call-center components. Following is a look at the vital components and the kind of testing they require:
ACD
The ACD automatically distributes incoming calls among agents assigned to provide customers with requested services. Primarily dealing with voice sessions, it can be expanded to support online chat such as ICQ between customers and agents. During rush hour, the ACD mechanism should maintain functionality and withstand stress, managing call distribution in conjunction with a queue management program.
When testing the ACD, it is important to simulate incoming calls according to the Poisson model to verify system efficiency, determining the maximum, minimum, and average response time per call; the time each agent was busy/available; and the number of calls handled per busy hour.
IVR
The IVR mechanism enables customers to receive voice-based services using DTMF to indicate the type of service required. The main consideration when testing the IVR is to determine whether it reacts according to the DTMF streams sent (in- or out-band) using VoIP and whether it functions properly under stress.
AS
The AS enables the provisioning of add-on services such as unified messaging and receipt of account status by e-mail, fax, or voice message for customers. To provide these services efficiently, the agent accepts the customer’s name and account status online several seconds before the agent answers the call, which then is transferred by the ACD mechanism. This saves time for both the agent and the the customer.
Naturally, one of the most important challenges is security. Customers should only be able to receive information on their own individual accounts. This can be achieved by users initiating sessions using a PIN code and the agent asking the customer several questions in random order from a predefined list entered into the system at the registration stage.
Testing the AS should cover fault insertion tests to verify immunity of the system against intruders and unauthorized users; functionality under stress; testing of the pop-up screen mechanism for incoming calls; and prediction tests of system behavior, traffic calculations in compliance with the Erlang C model, performance, and level of service.
VRS
Voice recognition is a fast-growing method for securing the system to avoid entry of unauthorized personnel. Using a powerful, dedicated DSP, VRS is a sophisticated conversion mechanism that represents voice characteristics such as accent, pronunciation, frequencies, harmonies, and amplitude and enables customer authentication according to voiceprint.
An adaptation mechanism should follow modifications of the human voice over time. The adaptation mechanism learns these adjustments and identifies the person behind the voice.
The VRS mechanism must overcome the challenges presented by VoIP-based systems, which affect voiceprints and increase the complexity of ensuring VRS integrity. These challenges underscore the importance of testing the VRS under real environment conditions using an IP network emulator.
To this end, tests should be conducted under stress conditions to verify the behavior, reliability, and efficiency of the VRS.
- Compression and decompression mechanisms can cause quality degradation, which can be measured by objective means correlated to MOS.
- Silence suppression and VAD can cause both front-end and back-end clipping, affect background noise, and result in the metallic voice phenomenon.
- Echo and latency should be measured and presented by return loss and echo delay parameters. Efficient echo cancellation with a predefined echo tail can decrease this phenomenon.
- Jitter, packet loss, and packet misorder are inevitable because VoIP is a real-time application combining continuous phenomena over a deterministic transport layer (packets). Latency differs per packet since the path of each packet is unique.
In real-time applications such as voice, jitter is an impairment that should be treated. According to RFC 1889, the relative transit time of two sequential packets, i and i-1 (expressed as Di), is defined as:
Di = (Tri – Tri-1) – (Tsi – Tsi-1) = (Tri – Tsi) – (Tri-1 – Tsi-1)
where:
Tri = time the ith packet was received
Tsi = time the ith packet was sent
- Quantization and harmonic distortion could look like background noise. Quantization distortion may appear in the process of converting voice from an analog signal into PCM and encapsulating it into packets. Harmonic distortion can be caused by use of nonlinear components for the conversion.
The SNR affects the probability of successfully carrying out the voice-recognition process.
DBC
A huge database should be able to access customer data during sessions. The most important issues are capacity and response time, a backup mechanism to ensure continued operation, and protection against intruders and unauthorized users. The testing of database functionality and the integrity of the redundancy mechanism require the simulation of many incoming calls that send queries to the DBC.
Management Tools for Online Provisioning
Management tools enable call centers to evaluate the level of service received by customers, providing statistics and online details concerning agent status, the number of calls answered in a given time, the time each agent was inaccessible, and the response time to incoming calls. These tools accommodate the analysis of various parameters, such as queue status; the maximum, minimum, and average time customers have to wait for a response; and the number of abandoned customers during a predefined period.
Management tools also are responsible for sending alarms concerning malfunctions, traffic blocks, congestion, and bottlenecks. In many enterprises, the call center is a distributed, network-based implementation with the customer routed to an agent who can be located anywhere, such as an office or private home. This enables the management tools to monitor the routing process.
PDS
Using manual methods to reach customers is expensive and inefficient. Statistically there is a 12% probability of fulfilling sessions that are dialed by agents.
The solution is PDS: sophisticated systems for telemarketing and responding to abandoned customers. These tasks are achieved by conducting calls when the probability of reaching the customer is high, enabling enterprises to initiate calls to customers with a high probability of call completion. Additionally, when the call is initiated, an agent, to whom the customer will be routed immediately after reply, must be accessible. A PDS also can include a follow-up mechanism that can locate a customer from among different phone numbers.
DLS
These systems are integrated in call centers to collect data online about the calls, the instructions given, and the actions made to accumulate records and history.
IP Phone/Internet Telephony Gateway
Access to call centers can be through traditional PSTN or directly through the IP net using an IP phone or a POTS through a gateway. To save bandwidth, the VoIP path can be implemented using media compression, silence suppression, multiframes per packet structure, or CRTP.
Use of these mechanisms in call centers presents several challenges, such as QoS, GoS, and voice-quality degradation. Functionality testing requires simulation of this type of traffic using multiple IP modes, enabling enterprises to generate traffic using the Poisson model, the most realistic representation of phone traffic based on call-center user behavior.
AC
ACs are implemented in call centers to track customer actions and ensure that only authorized customers use the system. Access to call-center functions only is possible after entering a PIN code. New customers are routed to registration centers to receive a PIN code, record their voiceprints, and enter other relevant information.
Testing Performance and Integrity
Performance and integrity are keys to successful call-center operation. To ensure operation with a predicted amount of traffic, calculations and tests should be conducted based on the Erlang B and Erlang C traffic models. These models provide call-center planners with the ability to calculate the number of resources required for the system during each hour of the day.
These calculations are based on the assumption that call arrivals comply with the Poisson statistic model when the user population is large enough and when users place their calls independently (Figure 2). Consequently, when deploying a call center, one of the main tasks is to test whether the estimated number of agents and trunks required per hour per day can handle a predefined number of calls and provide an acceptable GoS.
Dynamic tests should be performed to evaluate call-center performance vs. the number of agents required. This is an ongoing process requiring continuous reassessment as call-center circumstances change and considering various factors such as peak hours, special time periods (end of the year), and special events (marketing campaigns).
The fundamental parameters that should be known or anticipated include the number of users, the average duration of a call including the wrap-up time, the number of incoming calls, the average delay for incoming callers, the average wrap-up time, and the GoS target expressed as a fraction of the total calls that will be lost because of insufficient line provisioning.
Traffic probably will be lower during lunch breaks than during rush hour (Figure 3). However, this might change according to geographical location and business type. For example, stock-market activities can transform the lunch break into rush hour.
A VoIP Test Solution
RADCOM’s VoIP Performer™ Test Suite guarantees carrier-grade quality. Providing full end-to-end testing and mapping of quality vs. stress, it introduces a conceptual method for testing VoIP networks (Figure 4). The Performer includes emulation and simulation components as well as measurement and analysis components.
The system is controlled from one management console and allows dynamic scripting and automation of testing procedures. Using this system, enterprises such as call centers run by banks and financial institutions can activate an optimization process to find the best match between parameter configuration, network impairments, and desired quality.
The suite consists of the following solutions for testing the various components that make up typical call centers:
- TMS: central management tool that controls all Performer components.
- QPro: scalable, multiple PSTN client emulator, including stress and objective quality measurements, to enable customers to simulate thousands of incoming and outgoing calls.
- 323Sim: multiple H.323 client emulator including registration, signaling, and media; capable of simulating up to 80,000 calls per busy hour.
- SIPSim: multiple SIP client emulator including registration, signaling, and media; capable of simulating up to 700,000 simultaneous calls and up to 3 million calls per busy hour.
- Capture: testing for interoperability and compliance to standards.
- NetSim: network cloud emulator and impairment simulator.
- MediaPro: real time, packet-based, objective- and subjective-quality measurements.
- MasterScript: dynamic scripting tool for automation and optimization processes; capable of controlling multiiple consoles and servers.
About the Author
Eyal Tomer was the former products department manager at RADCOM LTD. in Tel Aviv, Israel.
About the Company
RADCOM Equipment, 6 Forest Ave., Paramus, NJ 07652, 800-723-2664
FOR MORE INFORMATION
on VoIP technology enter this rsleads URL
www.rsleads.com/305ee-179
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2003 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
May 2003