Three Generations of Chat Technology
The ability for companies to interact with online visitors has progressed quickly. While chat technology has evolved through three distinct stages, many companies still use first-generation technology and miss the business benefits that can only be achieved at the third-generation level. Reviewing the evolution of chat technology illuminates why offering a click-to-chat button or implementing pure rules-based technology level is not enough to produce significant business results.
The First Generation: Unintelligent Chat
In the early days of e-commerce companies strove to make their online channel as friendly and effective as their traditional sales channels. Offering a chat capability seemed the answer.
Service providers charged on a per-chat basis, regardless of outcome, so all chats--whether instigated by someone just researching or by a customer ready to place an order--were essentially equal. Furthermore, since the decision to instigate a chat was entirely up to the visitor, there was no way for the company to control the quality or volume of chat interactions, which came with a significant cost.
The first generation of chat technology was essentially unintelligent; no decision-making was brought to bear on qualifying visitors and the visitors, not the company, were in control of how and when company resources were deployed.
The Second Generation: Rules-based Chat
Service providers attempted to build intelligence into the next generation of chat technology to allow companies to selectively offer chat. By implementing the ability to track visitors' onsite behavior, providers could then define rules to govern which visitors would be invited to chat.
Second-generation technology may reduce the time sales reps spend chatting with people who are not going to transact. And it does introduce for the first time the critical proactive component to successful online interaction. However, this technology has a basic, inherent weakness: Rules are made and evaluated based on human intuition. One might guess that someone who views five pages is likely to transact--but what if the magic number is three pages...or 10 pages?
The truth is that any rule--unless somehow statistically verified--is based on an assumption about human behavior. And rules, no matter how finely tuned, always operate at this level of generalization.
To a limited extent, trial and error can ameliorate this problem. Sales reps might observe that visitors who go straight to the order form aren't likely to welcome a chat unless they've had exposure to the company first. The company may therefore modify a rule to specify that a "chat worthy" visitor must have visited the site twice in the past month and spent at least five minutes on the product page and/or reviewed the company overview page. But even if this kind of complex, multilayered rule results in more productive chats, the question remains: "Which element of the complex rule was key?" And more important: "Are there unconsidered factors that might better identify which visitors are most likely to transact?"
While rules-based chat is still the status quo for some providers, companies should consider whether the "analytics" involved in rules-creation have been tested and verified, either before or after their implementation.
Clearly, rules-based chat represents a huge leap forward from unintelligent chat, and introduces some level of analysis and targeting. However, companies should be aware of the drawbacks, chief among them the high hidden costs. Humans must interpret and hard-code the goals into their Web site and continuously measure their performance. Anytime there is a change in visitor behavior, they have to scrap everything and start all over. Generally speaking, rules become very high level, very simple and nonoptimal because they are very expensive to maintain.
Given that rules are driven by assumptions, are manual, are unable to contemplate unforeseen factors, and are very expensive to maintain, the need for a third generation of chat technology becomes apparent.
The Third Generation: Intelligence-driven Interactions
The problem with first-generation technology was that it made no attempt to qualify site visitors for a chat. The problem with second-generation technology was that it qualified visitors using untested assumptions. Third-generation chat technology resolves both these problems by qualifying site visitors using rules and ranking visitors based on statistically meaningful evidence.
This is how the technology works: During a training period, online visitor behavior and interaction results are tracked and recorded. That information is analyzed (via linear regression and proprietary algorithms) to see which actual visitor activities and session attributes are correlated with successful transactions. This analysis reveals which visitors were statistically most likely to engage in a successful transaction when provided with live assistance. New online visitors are compared against that model in real-time and assigned a score, reflecting how closely their current activities match those of the ideal visitor. Only visitors with the highest scores are automatically offered the chat option.
When organizations have statistical evidence, they no longer have to intuit who may be a good candidate for a chat. They no longer have to wonder whether they're targeting visitors with the greatest potential for transacting. Nor do their business analysts have to spend time and money manually creating and endlessly tweaking business rules.
This is not to say that third-generation technology makes business rules obsolete. Rules are still very useful when used for exclusionary filtering, visitor classification, and agent routing, prior to visitor scoring. For example, a rule could specify that visitors who view the careers page be excluded from the scoring pool. By removing noncandidates from this pool, business rules increase the efficiency of the scoring engine and the productivity of agents.
The Advantages of Third-generation Technology
Third-generation chat technology has several major advantages over second-generation technology:
Human guesswork and intuition no longer determine which site visitors will be given the opportunity to chat with a sales rep.
Visitors are targeted for a chat based on meaningful statistical data gathered on the site.
Agents spend more time interacting with people who are really interested in their product or service.
The company's staff or outside consultants don't have to manually write and update business rules.
The scoring engine is self-training and self-executing, which means companies don't need resident gurus to keep the technology working.
Site visitors are automatically scored and (if appropriate) directed to available sales reps.
The ongoing effectiveness of the scoring engine can be predicted and measured based on the mathematical relationship between a certain visitor action like visiting page 10 and its coefficient completing an order.
Third-generation chat technology has fundamentally changed the landscape of online interactions. With the e-commerce channel more important than ever, companies must take full advantage of the benefits offered by the latest in interactive technology. When considering solutions for their online enterprises, companies should look for one that includes regression-based propensity modeling and real-time scoring--and take the guesswork out of which visitors they should target for a chat to achieve optimal business results.
About the Author
Gregg Freishtat is the chairman and CEO of Proficient Systems. An innovator in online interactions, Freishtat has founded and built a series of companies throughout his career. He received a BA from Boston University and was awarded his JD from the University of Maryland School of Law. Please visit Proficient Systems
Contact Centers Chatting to Success
01 Nov 2008
No longer simply a forum for teens, chat is becoming an integral part of the cross-sell, upsell, and support strategies run by customer service.