One definition of insanity is to see patterns where there are none. So it is a little frightening to find an immediate exception to last month’s observation that the best-selling interaction management systems all provide integrated statistical modeling. This observation was the basis for many profound conclusions regarding the real uses for such products, and indeed the direction of the marketing software industry as a whole. Now a pesky little fact calls the entire structure into question.
Handshake Engine (Million Handshakes, 408/352-3400, www.millionhandshakes.com) is the anomaly: a successful interaction manager without statistical modeling functions. The system, launched in 1995, has more than 30 installations. This places it among the top vendors (as measured by number of customers) in this field of a dozen or so competitors. The other leaders – E.piphany Real Time, Norkom Alchemist and Data Distilleries DD Series – all provide integrated modeling and account for about 60 percent of total installations. Without Million Handshakes, their share would be about 75 percent.
What Handshake Engine does is perform the task originally envisioned for interaction managers but rarely implemented in practice: coordinate customer contacts across multiple touch point systems. The system does this through dialogues that lay out the process for a single interaction – say, handling a customer service problem – or for a master sequence of interactions throughout the customer’s relationship with a business.
The interactions are executed through a company’s customer contact systems, such as telephone call centers, Web sites, e-mail or direct mail. Handshake Engine links to these systems either by scanning their internal databases for relevant events or by reacting to custom-written calls to the Million Handshakes Application Program Interface, or API. The scanning approach lets Million Handshakes generate batch outputs such as mailing lists or react in nearly real time to produce e-mails or schedule outbound telephone calls. The API-based approach lets the system respond in real time as an interaction is under way.
Handshake dialogues are built as a flowchart by connecting icons in a graphical user interface. Each dialogue starts with a group definition. This can be a specific set of individuals selected with a query against a customer database, or people presented to the system by an external event such as a Web site visit. The dialogue itself can split groups into segments that follow different paths, generate personalized messages, wait for replies and react when a reply is received.
Splits can be based on external customer data, answers to system-generated surveys, the customer’s participation in past or other current dialogues, or random selection for test/control programs. A dialogue also can call an external program via XML, JavaScript or VBScript. This would be the approach used to integrate an external scoring system.
Dialogues can loop back within themselves to repeat a cycle or can call other dialogues. The latter feature allows a master dialogue to control the sequence of interactions, while the interactions themselves are executed by subsidiary dialogues. This improves usability by breaking a huge customer management process into manageable pieces and letting different master dialogues share the same subsidiary components.
System outputs can include personalized e-mail or direct mail messages. These are generated from user-defined templates using data pulled from the underlying databases. Results can be fed to an external system such as an e-mail server for immediate delivery or can be accumulated in a file for later batch output. The templates are built with third-party word processing or e-mail software that is integrated with Handshake.
Outputs to other touch point systems, such as call centers and Web sites, can either refer to specific pieces of content or provide attributes that the touch point will use to make its own selection. The system relies primarily on XML for such communications. Handshake also can generate nonpersonalized mailing lists that are fed to external sources on a regular schedule.
Although Handshake relies on external databases for most customer information, it does keep track of each customer’s progress through the different dialogues, as well as category assignments and responses to system-administered questionnaires. This internal database also stores administrative information about the dialogues, group and segment definitions, available messages, communication channels and other entities.
For each dialogue, the system stores active or inactive status, start and stop dates, and the cost and revenue associated with each step. Later this year, the vendor plans to let users place dialogues into groups and to limit user access to specified sets of dialogues. Both features will help manage dialogues in large organizations where many dialogues are used for different functions.
The vendor also plans to simplify API-based integration by releasing adapters for specific external systems, such as call center and Web site systems. These will let users embed Handshake dialogues within the external systems’ processes without writing directly to the Handshake API.
Handshake generally does not allow users to extend the standard administrative tables to capture nonstandard information. But users can store such data in external tables and read them like other external sources when necessary.
Reports in the system are written against the internal database using the Business Objects query and reporting tool. Standard reports show how often the different dialogues have executed and how many customers have passed through each step. The system can also draw on external data to generate profiles of the customers reaching each step. While the data underlying these reports are continually refreshed, the system does not have standard functions to generate alerts when specified conditions arise – such as a particular dialogue firing unexpectedly often or not often enough. This could be accomplished using the Handshake process monitor or third-party reporting tools.
Handshake Engine runs on a Windows NT/2000 server and Windows workstations. Pricing starts at $85,000 for the server plus 25 cents to $1 per active customer, depending on the total.
Million Handshakes is headquartered in the Netherlands. Most of its clients are European firms in financial services, communications, utilities, travel and retail, and most use it primarily for outbound e-mail and direct mail (which may explain its anomalous success). The company recently entered the U.S. market.
