There is no question that email has become a preferred support channel for many customers.  While Web self-service, Web chat and communities receive more ink and visibility, email continues to be a growing support channel for SSPA members, with triple-digit growth over the last three years.  As seen in Figure 1, according to the SSPA Benchmark database the average monthly volume of email incidents grew from 1,257 in 2001 to 1,840 in 2003; a 46% increase.  Over the last three years, email incident volume ratcheted up a whopping 101% to 3,708 in 2006.

Figure 1:  Average Monthly Email Volume Increases Dramatically

 

 

 

 

 

 

 

 

 

 

As email volume has increased, service levels for these incidents have dropped.  As seen in Figure 2, email incident response and resolution times are much higher than those for phone.  As an asynchronous support channel, email incidents provide flexibility in agent staffing and availability.  Unlike phone calls, in which a live body must be available at that moment, email incidents carry with them an expectation of a short delay before the customer receives a response.  Most surveys now report that customers expect an email response within 20 minutes to 2 hours, and while this certainly provides more flexibility in agent staffing than a ringing telephone, clearly member companies are not meeting customer expectations for email response times.

Figure 2:  Service Levels Dramatically Lower for Email Incidents

 

 

 

 

 

 

 

 

 

 

Why then have companies not tried to improve service levels for email in an attempt to entice more customers to use this channel?  A common response from members is that customers were originally told to use email for ‘non critical’ incidents, and a decade later some companies still treat emails as noise.  However, email has become a common, if not the most common, channel for business communications today, and now customers are less willing for their emails to be treated as ‘second class citizens’ by support organizations.

Streamlining Customer Email Handling with ERMS

Email Response Management Systems (ERMS) first emerged in the late 1990s, originally targeting the new breed of online businesses that didn’t have large support teams or even published phone numbers.  Early ERMS case studies were primarily from online travel agents and retailers, with early ERMS vendors Brightware, eGain and KANA showing dramatic improvements to email service levels by deflecting some inbound emails using intelligent auto-response, and assisting agents with auto-suggest to cut handling time for each interaction.

As seen in Figure 3, at a high level ERMS technology works as follows:

• Inbound emails are fed into a classification engine, the ‘brain’ of an ERMS.  The email text is analyzed and compared against sequential rules in the engine, looking for matches.
• The classification engine looks for text matches in rules, such as error messages or product codes, but ancillary email information can be analyzed as well.  For example, using the email address to identify the customer account, integration to a CRM or billing system can also feed the classification engine meta data such as products purchased by the customer, which version they are using, etc., in order to aid in accurate rule matching.
• If an email matches one or more rules, a personalized auto-response can be sent to the customer.  If this happens, a case is automatically created to show the inbound email and the outbound response in the customer’s history.
• If no rule match is found, or the system’s confidence level of the match is low (most systems allow you set to set a confidence level on each rule, so it will not fire an auto-response unless the match is strong), the email is routed to an agent for handling, generally with a set of suggested responses for the agent to review, hopefully cutting the time involved to respond.

Figure 3:  The Classification Engine is the “Brain” of an ERMS

 

 

 

 

 

 

 

 

 

 

 

Member Adoption of ERMS Remains Low

ERMS offers excellent ROI potential:  resolving customer issues with no agent involvement.  The cost of auto-response incidents are typically listed as less than $1, about the same as issues deflected with Web self-service.  ERMS vendor case studies also show that agents using auto-suggest can effectively process and respond to customer emails considerably faster.

However, though the potential for cost savings is great, adoption of this technology remains low in many segments of customer service, including technical support. In Figure 4, we see that only 45% of SSPA members >$1B have an ERMS in place; 11% are considering an ERMS purchase in 2007, and the remaining 44% have no interest in the technology.

Figure 4:  SSPA Member Adoption of ERMS

 

 

 

 

 

 

 

 

 

 

 

Why is adoption so low?  SSPA Research offers these likely reasons:

• Bad flashbacks. Support management that was around for the early versions of ERM systems, which were only accurate for simple, repetitive questions, may have had bad experiences.  Early systems also could not handle multiple questions in a single email, which is a common occurrence.  (“Can you help me with this error message; and when is the new release version available?”)
• B2C vs. B2B.  With the publicized case studies and ROI stories all focused on high volume, low complexity consumer contact centers, using this technology for tech support may not be an obvious option.
• Inapplicability.  Automated email responses will not solve complex technical problems requiring diagnostics, and support management may feel an ERMS simply won’t work in their environment.  But even complex technical support teams receive simple, repetitive questions, and “how do I?” questions which a detailed procedure emailed to the customer would resolve. 

While the percentage of emails handled by an ERMS is lower in a technical support environment than in a consumer contact center, the cost per case is dramatically higher in the B2B situation.  Even with fewer interactions deflected, the ROI can be fast.

Innovations in ERMS

Those members with less than dazzling results using early versions of an ERMS should know that the technology has matured a great deal since the late 90s, and today’s systems deserves a second look  Innovations in ERMS include:

• Natural language processing.  The single biggest development in ERMS accuracy was the advent of natural language processing (NLP).  Instead of relying on exact matches between inbound email text and rule definitions, NLP can identify questions by concept, allowing rule matching to occur regardless of how the question is worded.
• Support for multi-question emails.  This was a major limitation with early systems, which only allowed a single rule match for each inbound email.  If customers asked multiple questions within the same email, either all but one question were ignored in the auto-response, or the multiple question text confused the system and the response was inaccurate.  Today’s systems recognize multiple questions or concepts are contained in a single customer correspondence, and each thought can be analyzed separately.
• Access to external data.  With Web services integrations now supported by leading eService platforms, tying an ERMS classification engine to external data sources is easier, and vendors offer packaged integrations to common front office and back office applications.  Real time data integration means the system can make calls to 3rd party systems and knowledgebases to further define a question or retrieve a real time answer to a question (“When did my order ship?” or “Do you have model XYZ in stock?”).

The SSPA Recommends

Obviously I am a proponent of service and support technology.  When it comes to ERMS, I feel this is a “must have” for any support organization, and when members complain to me about their challenges in handling customer emails, often their issues would be resolved by an ERMS.  If auto-response just sounds too good to be true, consider these other ERMS benefits.

• Auto-Acknowledge.  You can buy more time to work on a customer’s email issue if you provide them an immediate “meaty” acknowledgement.  More than just replying with a case tracking number and a promise to follow up at some point, automated acknowledgements can also provide a list of possible answers, a link to the self-service website preloaded with search terms from the email, and with integration to your CTI or CRM system, even provide an accurate estimate of when someone will contact them.
• Auto-Suggest.  With all the attention on auto-response, many companies never hear about the benefits of auto-suggest.  I have talked to many companies using an ERMS who are afraid to use auto-response, but they still have great ROI results using auto-suggest.  When the agent opens the email incident, the classification engine has already decided what the most likely responses will be.  In a great example of streamlined user interface design, agents can peruse the list of suggested responses, click on the one or more which apply, do any text editing that is required, and hit send.  The response goes to the customer including complete personalization (customer name, company, product, etc.).
• Right Channeling.  If the classification engine doesn’t find a perfect fit for auto-response, a skills-based routing engine determines who the email should go to.  Instead of dropping everything into a generic queue, the routing rules will send the email to a specific agent or group depending on any rule criteria, such as product, expertise level required, or level of service contract.

Stay tuned for more on automating email.  Later this month I will publish a report on best practices to increasing effectiveness and accuracy with your ERMS.

About John Ragsdale…………………………………………………………

John Ragsdale is Vice President of Research for the SSPA.  Ragsdale spent 10 years managing tech support operations before moving to Silicon Valley where he held product management and marketing positions at eService and CRM vendors. He spent 5 years at Forrester Research as VP and Research Director before joining the SSPA.