Live Chat Performance Benchmarks A Statistical Analysis 2012 Edition
©2012 LogMeIn, Inc, Version 2
Highlights • Thousands of customers, worldwide • Reliable 99.99% uptime • The industry’s best value • Rapid deployment • 24/7 support • Proven installs across vertical markets Contact BoldChat, a LogMeIn brand 2024 N. Woodlawn Suite 350 Wichita, KS 67208
Phone: (866) 753-9933 (316) 462-7190 Email: [email protected]
In the fall of 2009, we published a benchmarking report which queried the data across our entire customer base in order to produce statistics related to live chat usage. Prospective customers, members of the media, analysts, and existing clients had asked for data from the vantage point of Internet retailers, e-service providers, and other website owners who use live chat software for sales and services engagements. The initial publication marked the first time that a leading live chat provider openly shared statistically relevant benchmarking data and, as expected, the report was immediately popular. The document, entitled Live Chat Performance Benchmarks: A statistical analysis remains one of our most popular collateral pieces. Because our products are provided through a software-as-a-service model, our infrastructure includes aggregated data from thousands and thousands of live chat customers. With tens of millions of chat records, hundreds of millions of website visit records, and billions of unique page-URL recorded visits, we have access to one of the largest live chat user communities in existence. This massive base of real-world data has enabled us to extract the findings presented in both the original report, and this follow-up version. While many of the goals of this analysis remain consistent with the inaugural version, there are three additional goals which themselves were born from readers of the initial report. The original goals remain: 1.
Present benchmarks for a wide variety of live chat operational practices
Uncover implementation differences affecting key live chat operational issues
Test the statistical causality of live chat operational best practices
The additional goals in this updated version are: 4.
Monitor and report on changes to key benchmarks since the first report
Calculate benchmarks for higher traffic sites
Identify and present benchmarks for “elite” users of live chat
These last two goals are worthy of discussion. In the first benchmarking report, we noted that the key determinant related to many critical live chat statistics was quite simply the amount of website traffic a business had. As the volume of visitors went up, the averages and ranges associated with the benchmarks shifted in meaningful ways. Higher traffic sites, rightly, have asked us to contemplate benchmarks for them distinctly and in this updated report we do so explicitly. The last goal comes from the fact that many of the benchmark ranges presented in this type of analysis are, by their nature, very wide. In many engagements with customers and others familiar with our benchmarking report, we’ve had conversations that began mostly like this: “Thanks for providing these benchmarks; it’s helpful for us to understand the ranges one could expect on average. But, we don’t want to be average! We want to be exceptional! If we apply best practices and rigorously test things as you suggest, then what can we expect?” In this new version of the benchmarking analysis, we attempt to provide guidance here. In several sections throughout the report, we analyze the “best of the best” – smaller groups of customers achieving exceptional results – so that readers can discover not only what average is, but what’s superlative.
In several sections throughout the report, we analyze the “best of the best” – smaller groups of customers achieving exceptional results – so that readers can discover not only what average is, but what’s superlative.
Terminology For those unfamiliar with live chat technology some key terms are defined here which will help readers to better understand the data in this report. Other terms related to customer groupings are also defined. Conversion: The BoldChat system includes a conversion tracking mechanism allowing customers to specify and report on customized conversion types. For this analysis, only financial conversion types were considered (i.e., purchases). Pre-Chat Form: A set of questions posed to a potential chatter before the chat starts but after they click a chat button or accept a proactive invitation. Unavailable Email Form: A form presented to the website visitor after they click on a chat button during a time in which no chat operator is available to answer the chat. Reactive Chat: Chats initiated by website visitors by clicking on a chat button. Proactive Chat: Chats which occurred due to a website visitor’s acceptance of an invitation to engage in a chat.
Unavailable Chat: An attempt by a website visitor to initiate a reactive chat during a time in which the website’s agent is not available to answer the chat. Abandoned Chat: A chat that makes it to the pre-chat form, but never actually starts. Unanswered Chat: A reactive or proactive chat begun by a website visitor but never responded to by a website’s agent. Higher Traffic Sites: The group of websites in our customer base that are well trafficked. These sites, on average, experience almost four times the website traffic as our average customer. Elite Customer Segments: A smaller group of websites for which traffic is significant and for which all the sites in the segment exhibit benchmark performance at the highest level.
Benchmarks vs. Best Practices While the data provided in this report is certainly aimed at making customers more successful, they cannot themselves be called best practices. Certainly, part of this project aims to uncover areas that can positively influence a live chat implementation, but best practices are not only derived mathematically. There is a human component – an expert human component – necessary for any best practice to yield results. So, while the data in this report may inform best practices and even uncover new areas that can become best practices, they cannot be labeled as such.
There is a human component – an expert human component – necessary for any best practice to yield results.
While “benchmarks” is not an entirely accurate term either, it does more correctly describe the statistics presented herein. A benchmark is a snapshot of data against which progress can be measured over time. Using aggregated metrics of thousands of customers should give new and existing live chat users alike something against which they can compare their own performance.
Approach To create this report, we employed three tactics iteratively. Common Questions During sales and professional services engagements, our staff is frequently asked a similar set of questions. Again, we are able to answer these questions based on expertise developed over years of firsthand experience, but the questions themselves were directionally useful as we crafted specific queries to our vast databases of information. Some of the most influential questions posed to us are: • What kind of chat traffic can our company expect? • Does proactive chat positively influence conversion? • What percent of chatters will convert to sales? • What is the best time to engage a visitor proactively? • Should we start with chat for support or chat for sales? • Where should we put the chat button? Known Best Practices By working directly with customers over the years, our professional services team has developed a playbook of implementation standards that seem to work time after time. This project afforded us the opportunity to statistically verify or refute the observed efficacy of many of these practices.
Data Interrogation Many people inside our organization, because of the nature of their jobs, have an intimate knowledge of our data structure. The insight brought to bear on this project by our database administrators, developers, professional services personnel, quality assurance team members, and many others proved an invaluable resource. Their understanding of the interdependence between seemingly disjunctive data points coupled with a dogged curiosity for discovery often led us in new and important directions.
Live Chat Benchmark Statistics This section of the report, divided into sub-sections, shows and explains the findings from this research effort. Live Chat & Conversions In this section’s analysis, we examined tens of thousands of individual purchase conversions. It yielded some of the original report’s most intriguing findings. The same holds true today. • Chatters are 7.5x more likely to convert than visitors who don’t chat. Up from 4.1x in 2009.
• Chatters buy, on average, 24% of the time.
This section includes a discussion of the statistical decisions made during this project.
• Chatters spend about 55% more per purchase than non-chatters. • Chatters who engage via proactive invitation are 8x more likely to convert than visitors who don’t chat. Up from 6.3x in 2009.
Median vs. Mean Many of the benchmark statistics presented in this analysis utilize the median rather than the mean. The median, distinct from the arithmetic mean, is the middle value of the data. Half of the data set is less than or equal to the median and half is greater than or equal to the median. The median is still a type of average, however, so it is appropriate for a given benchmark to say that, “the average BoldChat customer…”
Knowing that higher traffic impacts benchmarks, here are some of the above statistics recalculated against our higher-traffic customer base.
We decided to use the median because it corrects for outlying data while an arithmetic mean does not.
• Chatters with higher traffic websites buy, on average, 17% of the time.
Confidence Interval Where appropriate, we also provide statistical ranges within a 95% confidence interval. Ranges allow people to see the upper and lower boundaries that can be expected. Intervals also allow us to make the following types of statements: “We can be 95% confident that the average BoldChat customer will experience between X and Y…”
• Chatters are 5.5x more likely to convert than visitors who don’t chat.
• Chatters who engage via proactive invitation are 9.8x more likely to convert than visitors who don’t chat. And while both result sets are impressive, here is a calculation for the “elite conversion segment” set of customers. These businesses have turned chat conversion into an artform worthy of study, and accolades. • Chatters with the elite buy, on average, 55% of the time.
Chat Engagement These data points are focused on actual live chats occurring between a website agent and a visitor. • The average percentage of website visitors who engage in chats is 1.7%. This is essentially unchanged from 2009. • Within a 95% confidence interval, the engagement rate ranges between <1% and 15%. • For our higher traffic websites, less than 1% of visitors engaged in chats. Readers are advised to remember that this statistic represents the middle of the data set; half the sample experiences engagement percentages equal to or above this figure and half experience engagement percentages equal to or below this number. If a site aims to maximize engagement, we know it’s possible. We know by looking at the range associated with the “elite engagement segment” of our customers. This group chats with 9-22% of their site visitors. Proactive Chat Proactive chat is the issuance of a form, image, or other component that generally appears on top of a website and invites the visitor into a chat interaction. • The average percentage of website visitors who accept proactive invitations to chat is 8.5%. Up from 6% in 2009.
• The acceptance rate for proactive invitations continues to vary widely. Within a 95% confidence interval, the acceptance rate is between 1% and 29%. Previously, this range had acceptance rates between 1% and 22%. • The ‘Chat Form’ type of invitation is still more widely accepted. It’s nearly 37% more likely to be accepted over other types. However, in 2009, it was 50% more likely to be accepted. Proactive appears to be critical for higher traffic websites. Among these sites, proactive had an average acceptance rate of 9.7%. Proactive chat is an area where many customers attempt to optimize. Looking at the “elite proactive segment” of our customer base, proactive acceptance ranges from 13% to 46%, but the Chat Form type of invitation is no more or less likely to be accepted.
Expected Chat Volume The possible range of the chats to visit ratio is substantially affected by the size of the website and the use of proactive invitations. The chart below gives averages on what one can expect for chat volume based on the visit volume a site receives each month. It appears that for larger sites, it is more difficult in general to get the same proportion of visitors to engage in a chat compared to smaller sites. Adding proactive invitations, however, can substantially increase one’s range of influence over chat. The expected chat volume for any sized site should increase as proactive chat becomes a larger part of the mix. By charting the median values of engagement, one can see the importance of proactive chat. On average, it will more than triple a site’s engagement rate.
5.00% 4.50% 4.00% 3.50%
3.00% 2.50% 1.93%
Figure 1: Expected Chat Volume
Unavailable, Unanswered, and Abandoned Chats These data points revolve around missed chat opportunities. Pre-Chat Survey Before a chat begins, many customers collect some information from the visitor. They do this because it either helps them provide better service or to filter out unwanted engagements. • The average percentage of website visitors who abandon chats when a pre-chat form is presented is 47%. That’s up from 39%, which hints that visitors are becoming less willing to complete a form prior to receiving assistance.
It is clear that showing certain fields and/or making fields required affects the average abandonment percentage. Requiring a phone number appears to make a significant difference worthy of note. Median 47%
Status Not Shown
Figure 2: Pre-Chat Form Fields
• For higher volume sites, the abandon rate was 44%.
Unanswered Chats We regularly tell customers as a best practice to answer chats within 10 seconds. Initially, the data from this project challenged that advice. • The average wait time before visitors give up on a chat is 20 seconds when a pre-chat form is not in use which is up from 15 seconds but, • The average wait time before visitors give up on a chat is 54 seconds when a pre-chat form is in use. This is down significantly from 113 seconds in 2009. So, while visitors over time seem to be becoming less likely to fill out a pre-chat form, they are becoming more patient when one isn’t in use, but less patient when one is. If they bother to fill out the form the chat needs to be answered today, faster than in 2009. Overall, with each passing second, more and more visitors give up:
100% 90% 80% 70% w/Prechat
40% 30% 20% 10% 0%
Figure 3: Unanswered Chat Survival
Unavailable Email Form When website operators are not available to chat, an email form can capture lead information for later follow up. • The average percentage of website visitors who will submit an email form if one is presented to them is 17%. This is down from 23%. • Within a 95% confidence interval, the average percentage of website visitors who will submit an email form if one is presented to them is between 0% and 57%.
Chat Satisfaction Aggregated customer data confirms that live chat engagements are highly satisfactory for website visitors. • The average percentage of chatters who will fill out a post-chat satisfaction survey when presented to them is 25%. • Using the ‘window close prompt’ offered by BoldChat increases survey response by 39%. It’s even more effective than the previous benchmark of 31%. • The average satisfaction score given on a 1 to 5 scale across all survey questions is 4.4 which is identical to the 2009 findings. • On average, 10% of the people who submit a post-chat survey also include comments, down from 29%. • The average satisfaction scores for all categories are between 2.75 and 3.5 for chats lasting less than 10 seconds. • The satisfaction average goes up gradually from 3.5 up to 4.0 for chats that increase from 10 seconds up through 30 seconds.
Satisfaction scores based on chat duration in seconds
5.00 4.50 4.00 3.50 2.50
2.00 1.50 1.00 0.50 0.00
Figure 4: Chat Satisfaction and Chat Duration
• The average satisfaction level increases up to 4.3 at 80 seconds and up to 4.5 at 180 seconds. • Between 180 seconds and 1000 seconds for chat length, the average satisfaction fluctuates slightly between 4.3 and 4.5. Chats that are too short score low from a satisfaction standpoint but, interestingly, the inverse appears not to be true. Even chats lasting nearly 20 minutes are still scored highly.
Chat Satisfaction & Conversions We were interested in the relationship between conversions and satisfaction. As such, we looked only at the group of customers who both use conversion tracking for purchases and offer a post-chat satisfaction survey. What we found is telling: • The sites in the top 20% for conversion percent averaged 4.50 on their overall satisfaction scores. • The sites in the top 50% for conversion percent averaged 4.48 on their overall satisfaction scores. • The sites in the bottom 50% for conversion percent averaged 4.25 on their overall satisfaction scores. • The sites in the bottom 20% for conversion percent averaged 4.24 on their overall satisfaction scores. Another way to look at these results: • Sites in the top 20% for satisfaction scores had a 266% greater chance to convert a chatter than an average site. • Sites in the top 50% for satisfaction scores had a 63% greater chance to convert a chatter than an average site.
• Sites in the bottom 50% for satisfaction scores had a 29% less chance to convert a chatter than an average site. • Sites in the bottom 20% for satisfaction scores had a 35% less chance to convert a chatter than an average site.
Conclusions The data presented herein quantitatively supports the following conclusions: Live Chat Is an Increasingly Effective Sales Channel For the average website, adding live chat will increase conversions and average order size. A chatter is 7.5 times more likely to convert than a regular website visitor and this is up since the last time we reported this benchmark. That, coupled with the fact that on average a buyer through chat will spend 55% more, indicates that live chat – and those employing it – is improving for sales throughput and value. Proactive Chat Invites are Critical for Highly Trafficked Sites Chatters who engage via proactive are, for the average site, 8 times more likely to buy. That should encourage all sites to explore the technology with a serious intent for implementation. But, for highly trafficked sites it appears even more important. Chatters engaging on those sites through proactive are 9.8 times more likely to buy. Additionally, proactive chat appears to be a website’s best tool to impact overall chat engagement. Highly trafficked sites, on average, have 9.7% of their proactive invites accepted. Missed Opportunities Are Controllable, But Methods May Change Over Time Website owners are not helpless. Through the implementation of pre-chat forms (or not), optimization of answer
speed, and inclusion of unavailable lead capture mechanisms, Internet businesses can signi cantly change the number of engagements and follow-up opportunities created by live chat. Business should keep an eye on these statistics because over time, chatter behavior has changed. Visitors are increasingly less likely to fill out a form before chatting, but if they do, they want the chat answered more immediately than in 2009. Chat Satisfies and is Related to Conversions Not only do chatters participate in post-chat surveys but the average score they give across the variables of professionalism, responsiveness, knowledge, and other criteria shows that these interactions are overwhelmingly positive. Additionally, there appears to be a direct relationship between high satisfaction scores and the likelihood a site has to convert a chatter. Those in the top 20% for satisfaction scores had a 266% greater chance to convert a chatter. 7 Key Benchmarks Based on the findings presented in this document, live chat users should diligently manage, or seek the expertise of professionals in order to ensure that their implementations achieve, at least the following:
2. The implementation of proactive chat on top of reactive chat should increase a site’s engagement rate by ~388%. 3. Chatters ought to convert at ~7.5 times the rate of a regular website visitor. 4. Chatters via proactive chat ought to convert at ~8.5 times the rate of a regular website visitor. Customers with more traffic can expect higher proactive acceptance rates. 5. The use of a pre-chat form ought to cause ~47% of those who are presented with it to abandon, but should increase the survivability of the chat request. 6. Sites that use an unavailable email form on average capture contact data from 17% of those who are presented with it. 7. Speed of answer is more import than in our previous study. The 2009 benchmark indicated that answering a chat within 10 seconds would yield interactions with greater than 80% of visitors who initiated a chat. To achieve that same engagement today, you’d need to answer the chat in less than 5 seconds.
1. Depending on a website’s traffic volume the live chat engagement percentage should be between ~1% and ~15%.
BOLDCHAT BoldChat is a market-leading live chat solution enabling businesses to quickly and effectively engage visitors on their websites. BoldChat is offered in different editions and includes other integrated communications technologies like click-to-call, co-browsing, email management, SMS management, and Twitter management. Organizations of all sizes – from small proprietorships to large ecommerce enterprises – can drive more conversions and higher customer satisfaction by using BoldChat.
For more information: Phone: (866)753-9933 Email: [email protected]
Chat with us, start a trial or download more resources like this one at: www.BoldChat.com BoldChat is owned by LogMeIn, Inc. For more information, please visit www.LogMeIn.com 10