New SLAs and Traffic Rules for UC

Introduction

Service Level Agreements, aka SLAs, in the IT business are a BIG THING. All carrier and services contracts carry SLAs with associated penalties for non-performance. Vendor maintenance contracts also carry similar SLAs.

In the contact center, SLAs are used to measure Average Speed of Answer (ASA) and Abandoned Call Rates. These two basic SLAs are used as a baseline for performance measurement for outsourced contact center contracts.

And as multi-channel is gaining ground as the "new" way of handling calls in the contact center, multi-channel brings together voice calls, web chat, texting, video, e-mail, faxing, and social media integration. A "call" can be any one of these actions and needs to be identified and measured within the contact center - each is particularly different, with different expectations and different SLAs. Customers can now connect with the contact center in just about any form, from any device, including:

• Desk phone
• Home phone
• Laptop, desktop PC
• Mobile Tablet
• Mobile Smartphone

Any device connectivity to a contact center is now here, and any contact center needs to be designed with such in mind.

Typical SLAs

Because customers want to get answers quickly, here are baseline considerations for identifying SLAs or time to answer based on the "type" of contact center call:

• Voice Calls
  • The defacto voice call SLA of 80/20 (80 percent of calls answered in 20 seconds) is just the starting point (note that different industries may have different SLAs, either more lenient or stricter than the defacto stated here)
  • Typical call length - 2 - 4 minutes (3 typical average)
• E-mails and Faxing
  • Typically carry a response SLA of anywhere from one (1) hour to 72 hours depending on the contact center and industry
  • Typical response - 2 - 5 minutes (depending on the level of research needed and number of questions asked in the e-mail)
• Video chat/calls
  • Are categorized, in my opinion, as "instant response" and have little tolerance for any response longer than 10 - 15 seconds. This is thanks to, in part, the Amazon Kindle Fire Mayday button for "setting the bar" for video callbacks, a 15-second SLA that was surpassed last Holiday season down to 9 seconds.
  • Typical SLA - 5-15 seconds
  • Typical video call length - 4-16 minutes, depending on industry and type of transaction
• Web Chat, Texting
  • Similar to video, the expectation for a Web chat is 'instantaneous' as well
  • Typical SLA - 5-15 seconds
  • Typical chat session length - 1-3 minutes, again depending on industry and type of transaction. Most web chats are built around multiple chats at once
• Social Media
  • SLA will vary depending on "urgency" and "importance" of the discovered social media response
  • Length of a social media response - typically falls into "call" lengths similar to Web Chat sessions, commonly 1-3 minutes, also depending on industry and type of transaction.

Each of the above needs to be defined and designed for internal SLAs and staffing models to support these multiple channels. These are also needed for proper traffic design for number of channels, licenses, etc. when designing a new contact center.

Traffic Design for UC

So what are typical traffic design elements and expectations for UC in an enterprise? All of those areas cited below have an impact on total and QoS-based bandwidth requirements. Any of these elements can apply to SIP trunking:

• Voice calls - still average today about 3 minutes in length for a live call
• Voice mail - typically average 30 sec - 2 minutes
• Ad-hoc conference calls - can last typically between 5-15 minutes
• Meet-me audio conference bridges - typically last 30-60 minutes
• Webinars and training staff - commonly 30-60 minutes
• IM/chat - can be an excellent "quick hit" for answers to queries - typically can be as little as 15 seconds to 2 minutes from first text to last
• Videoconferencing - commonly 15-60 minutes, depending on the type of video conference and number of participants

WAN Design

Each of the above "call types" requires WAN characteristics to provide a best quality user experience and design characteristics, including:

• Availability - percentage circuit uptime on an annual basis - minimum of 99.9% - 99.99% (three 9s - four 9s)
• Latency (Delay) - Allowable packet delay between 2 sites, a maximum of 80 - 120 ms
• Jitter - Variations of packet delay - less than 3 ms
• Packet Loss - Packets not correctly formed - less than 1%
• Bandwidth - including all data, voice and video traffic, never to exceed 80% of the total bandwidth available
• QoS Traffic (Voice, Video) - needs to be prioritized in front of all data traffic, and the total bandwidth available is based on all data, voice and video traffic and overhead

Voice traffic typically runs at 30-35K per conversation (G.729 CODEC) compressed or 80-85K per conversation (G.711 or G.722 HD CODEC) non-compressed. Any HD video traffic can require as much as 1Mb per conversation non-compressed, or 12x that of a non-compressed voice conversation.

Lastly, there is a significant difference between a private data network (private Ethernet or MPLS), and a public data network (Internet). The key difference is that the public network is Best Effort only, and no QoS prioritization exists to prioritize voice and video in front of data packets. Relief can always be provided, to some degree, by throwing enough bandwidth at such a connection, but will never "hide" a problem if a circuit gets filled via excessive traffic or a network broadcast storm or anomaly.

As time goes on, the significance of a difference between private (QoS controlled) and public (non-QoS controlled) network is less so, however the basic QoS-controlled private circuit is always desired over a non-QoS controlled public circuit.

Conclusion

As we continue the journey to begin to deliver multi-channel in the contact center and deploy UC throughout an organization, the basic rules of customer service tied to SLAs and traffic design and engineering still apply. We will need to design around an "expected customer experience" for each of the contact center channels you may consider to deploy for your organization. You will need to:

• Discover and design around an SLA that is applicable to your industry and your specific environment.
• Design for WAN bandwidth and characteristics that meet or exceed those described in this post for a great UC experience.

Without consideration for these two areas, a full contact center or UC implementation may never get beyond "marginal" or "tolerable" and will dampen the entire customer or user experience. So make SLA design and traffic design a crucial part of a new contact center or new UC deployment; it will make the end user experience a "delight" in the end.