Home > Archive > July 22, 2001 > FAQ
Partial page image followed by the entire text. Additional links are Waller, Niels - Quantitative Psychology, Donoho, Dave - Statistics (wavelets)
Xamplify Technologies is a next-generation CRM analytics software developer focusing on applications that support cross-selling and customer retention. Our objective is to power CRM applications by bridging the gap between analytical CRM and operational CRM through real-time customer models. We focus on three primary areas:
We build software that combines psychometric information with relevant transactional and demographic information. This creates predictive and actionable models of what is driving individual customer behavior (why they behave as they do).
The software updates and manages these customer models in real-time, integrating information from all existing customer touch-points.
Our software operationalizes these customer models in real-time, anticipating how to identify and maximize the value of your best customers. By integrating with existing rules-based systems, the software bridges the gap between analytical CRM and operational CRM, maximizing the value of existing enterprise infrastructure.
The primary target of this technology is financial service companies looking to enhance their cross-selling capabilities.
An industry-wide trend of commoditization and margin pressures are forcing a shift from product focus to customer focus. A critical mass of financial services companies realize the best way to survive is through value-driven customer management.
The entire CRM industry has focused on interpreting customer behavior and subsequently segmenting customers based on observed transactional or demographic data. Because the data being used is inherently historical, there is a built in limit to its predictive ability, in addition, this methodology generates so much data it is unwieldy to manage and act on.
By integrating transactional/demographic behavior to a psychometric-based model, and linking that to an operational CRM implementation, we not only answer the industry's need for real-time customer-centric information, we also eliminate the need to store terabytes of historical data.
From a sales perspective, our medium term trajectory is to anchor in and rapidly expand into the financial services market. Our longer term trajectory could take us into the telco and retail markets.
We are targeting the financial services market, specifically banks, credit unions, brokerage firms, credit card companies, mutual fund companies, and secondarily, insurance companies.
Xamplify Technologies' target markets are companies characterized by the following:
Companies driven primarily by increases in profitability flowing from the deployment of sophisticated technology that manages customers and focuses on retention and associated margin growth
Companies that offer product differentiation based on price, brand, and/or service, and are looking to segment customers by value
Companies that are relationship-driven, with a large volume of high-margin customers with varying needs
Companies with a need for real-time integration of customer touchpoints systems.
Companies whose CRM initiatives have increased the volume of incoming data.
Companies that are branching into new markets and need to become rapidly familiar with their new customer base
The underlying technology is very sophisticated, but has been embedded in the software with the understanding that a non-technical audience would be the ultimate user of the system. Because of this, the tools we have created are designed with the client's business environment in mind, and should therefore be a familiar paradigm to most users.
You don't need an in-depth understanding of psychometrics in order to know how to use them (similar to not needing to know how a car works in order to drive it).
In terms of software support, we offer a multi-tiered technical support service with our product that meets the needs of a wide range of clients.
We also offer Professional Services tailored to the specific requirements of each installation and address the customization needs of our clients. However, unlike other vendors we stress the client's internal involvement during this professional services engagement to fully transfer the administration and usability knowledge upon the rapid conclusion of the engagement.
It depends on the number of systems we need to integrate with, and the available client resources. Usually, a typical installation can take from one to three months.
It's up to the requirements of the customer. Our integration efforts can be extensive and comprehensive, or the system can operate in a stand-alone mode, polling data off a single touchpoint, such as the web. Even at its simplest configuration, you will begin to receive actionable results almost immediately.
We assume there is a rules-based outbound marketing or customer service application in place for the touch-point that you want to act through (email, web, call-center, etc.). As part of the integration, we provide you with a customized adapter for the specific touch-point system (with a rules platform). If you are not using any 3rd-party rules platform, we can develop and implement a rules-based editor for you as part of a Professional Services engagement.
We integrate our intelligence into the rules scenarios in your existing CRM, SFA, ERP, etc. software applications. Our intelligence is automatically ported into your existing infrastructure, thereby requiring little work from you.
However, in order to build this custom adapter for your existing system, we will need the APIs for this software system that would integrate with our Profiler.
Software scalability design limits the need for heavy staffing
We will provide knowledge transfer to the internal customer staff
We can outsource needs via an ASP model
We have relationships with SIs for long term support/integration needs
We leverage existing aggregation infrastructure by selecting and associating transactional data with psychometric profiles to construct customer models.
The software has an open API using J2EE-based technology that allows various touchpoints and analytics engines to access the real-time customer models.
The software has adaptors that integrate with third party SW at various touchpoints to allow rules-based operationalization of these customer models.
The infrastructure that we enhance has only recently become the industry norm.
The Xamplify X-Seller is a state-of-the-art CRM application. We are first to market in applying this technology to the needs of customer driven business, and the market leaders in commercializing psychometrics into the enterprise/e-business space.
The applications we develop and sell do allow our customers to maintain highly detailed and actionable information on their customers, but the disposition of that information is subject to the privacy policies and procedures of our customers.
The discipline concerned with the quantification and analysis of human differences. It involves two main activities: the construction of procedures for measuring psychological constructs; and the analysis of data consisting of the measurements made. Psychometrics can be defined as the systematic use of tests to quantify qualities, abilities, and problems, and to make predictions about behavior and performance.
Psychometrics emerged as a formal discipline with the formation of the Psychometric Society in 1935. Led by psychologists from the University of Chicago, the Society focused on "developing psychology as a quantitative rational science." Since then, psychometrics has developed into a sophisticated, mathematically oriented discipline aimed at providing methodology for handling the particularities of psychological measurement. To accomplish this, psychometrics has become strongly computer and algorithm-based.
Although it originated in universities, and is still studied and refined there, psychometrics has been successfully applied elsewhere. Its methodology is deployed in psychological testing of intelligence, personality, achievement, aptitudes, interests, and proficiency, in schools, organizations, business, government, hospitals, and the military. Psychometrics has also been used with great success in industrial and organizational settings for job analyses, consumer surveys, personnel selection procedures, and market research.
Yes. For Example:
In the earliest days of psychological marketing, motivational researchers led by Ernest Dichter introduced psychology to market research. For Chrysler, Dichter found that wives had a significant role in the car-buying decision. Dichter's methods helped equate convertibles with mistresses. He also put the tiger in Exxon's tank.
General Electric successfully used psychometric research to revamp its brand identity (Cf. American Demographics, December 2000, "Measuring Minds in The 1990s.")
The American Cancer Society (ACS) and the Centers For Disease Control and Prevention (CDC) use psychometrics to identify those segments of the American population receptive to health promotions or ad campaigns that attempt to improve quality of life
Saatchi & Saatchi Business Communications uses techniques based on psychological and anthropological study to understand the emotional factors driving farmers' business decisions.
There are numerous psychological segmentation tools available to marketers today and widely applied in commercial settings. A number of market research firms conduct yearly surveys that gather data from thousands of consumers and divide respondents into distinct segments. Findings are sold primarily to advertising agencies.
Like demographics, psychometrics is rooted in a multi-disciplined, scientific approach, but goes beyond demographic information to discover what makes consumers tick. This enables businesses to successfully zero in on the customer's hot buttons. While demographic segmentation has some predictive power, it is usually not sufficient, (especially because inhabitants of the same zip code are much less homogeneous than they once were). That said, the most effective psychometric segmentation also factors in demographic information.
Yes. Transactional information is inherently historic and outdated, although generally somewhat predictive if you have a lot of it. Psychometrics allows a very high confidence of prediction of future activity with a very small amount of information when combined with demographic and transactional behavior. The bottom line: the most effective consumer segmentation factors in psychological AND demographic AND transactional data.
Research studies have consistently found that money attitudes are essentially independent of a person's income (Yamanchi and Templer, 1982); and that attitudes toward money have clear links to personality, but are largely unrelated to demographic differences of gender, education, or occupation (McClure, 1984). To understand and predict behavior in financial matters, it's essential to get to grips with the deep psychological roots of individual differences in how people view and handle money.
There is a vast-and growing-body of research on the psychological roots of attitudes and behavior in regard to money. To take a recent example: "Behavioral Economics" has attracted a lot of notice this year. As reported in the New York Times, "Behavioral economics had finally arrived: a discipline that for a half-century had built its theories on the rigid assumption that people acted with rational, unemotional self-interest had formally recognized that human beings had another, feisty, side to them." (February 11, 2001, "Following the Money, but Also the Mind," a story by Louis Uchitelle on the front page of the Business and Money section); that same Sunday, the New York Times Magazine featured an article on the dean of behavioral economics.
Xamplify's software represents the first integration of psychometric principles into a commercial CRM technology. The Xamplify Profiler improves the accuracy of prediction of CRM customer models through efficiently and quickly constructing real-time models from dual data sets: (i) transactional and/or other behavioral data and/or demographic data, associated with (ii) empirical or extrapolated psychometric profiles. The combination of these data sets places customers' behavior (what they do) in the context of their fundamental motivations (why they do it). This provides businesses with deeper, more actionable, more opportunistic and more confident predictions for selling opportunities.
Because parallel to the processes of data collection and analysis, the Xamplify Profiler delivers (in real-time) a set of actionable recommendations based on a client's business requirements. These recommendations are written as rules in the native language of any content management systems that are already in place. In this way, analysis and execution are melded seamlessly in one functional whole. Further, the system relies on a "hypothesis testing" schema: once the extrapolations are finished and the system deployed, subsequent end-user transactions are continuously observed to "correct" any deviations from predicted behavior by adapting the hypothesis of the models. In this way, the system self-corrects over time. No one else is applying the principles of psychometrics in this way.
Xamplify segments the entire installed base through advanced extrapolation techniques. The software extrapolates models of the non-profiled portion of the customer base from those of the profiled portion. Our software creates highly focused associations between users via their demographic-plus-transactional history and their psychometric profiles. The extrapolation algorithms and transfer equations we use have been field tested repeatedly and become more accurate with additional data. We are currently operating at an accuracy rate of over 90%. In addition, the system's "hypothesis testing" schema ensures that the customer models self-correct over time. Even prior to self-correction, however, our sampling and extrapolation technique gives a far superior performance, since the hypothesis relies on a large, independent set of intelligence (demographics, psychometrics and transactional information).
How accurate is it going to be? With every response we get we gather a wealth of information allowing us to eliminate multiple subsequent questions. This technique (known as Item Response Theory, or ITR) has been thoroughly tested and validated in the US in other industries, is used to administer the SAT and GRE examinations, and by Human Resources departments in the screening of job applicants. It allows a computer to draw from a large bank of questions the ones that yield the most differentiating and conclusive information about a person. Our methods were developed in house by the world's leading psychometricians (and IRT experts), Dr. Niels Waller of Vanderbilt University, and Dr. Steven Reise of UCLA.
The questions are designed by our psychometric experts to be incisive; by asking pointed, value-oriented questions, we gain a much deeper understanding much more rapidly than the more conventional approaches used by our competitors. Also, the questions are designed through a semantic ontology that determines several layers of information besides the explicitly asked information in the "lingo" and relevant metrics of our client's domain.
Immensely improved effectiveness of cross-selling products (improved metrics)
More revenue from existing customers
Higher profits and improved operating efficiencies through maximizing existing resources (IT, CRM, Reporting)
Highly focused and up-to-date (real-time) intelligence on customer base Generation of targeted products/features/price points/messages
Flexibility to adapt to changing customer needs and expectations
Ability to react immediately to warning signals and head off customer defections
Highly actionable, automated and intelligent operations
Minimum integration relieves IT burdens, while the overall systems becomes highly responsive and intelligent
High-transaction, performance tuned design provides excellent usability and buy-in from operations
Single platform for strategic and tactical operations of the enterprise
Xamplify's proprietary brand of psychometric assessment and analysis draws on the most advanced statistical methods and on our own private research as well as over 50 years of research into human personality traits. We have adopted, refined and extensively tested elements from those empirically derived personality typologies that have gained broad mainstream acceptance. They include the following:
The Five-factor model (FFM) of personality: Studies among various language groups show that the factoring of large sets of personality descriptive adjectives culled from dictionaries produce five broad replicable factors; Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Openness to Experience. The FFM became a popular research framework in meta-analytic surveys of the validity of traits for predicting human behavior.
The Holland RIASEC theory of vocational personalities: has been successfully applied to vocational life in recent decades (Holland, 1997). People can be characterized by their resemblance to each of six personality types: Realistic (R), Investigative (I), Artistic (A), Social (S), Enterprising (E), and Conventional (C). The same typology is used to characterize environments such as educational majors or vocations. The key assumptions of Holland's theory are that people search for environments resembling their personality profiles and that behavior is determined by the interaction between personality and environment. The theory, developed by psychologist Dr. John Holland, is the basis for most of the career inventories used today. It has been used by over 22 million persons worldwide for over 30 years, is supported by over 500 research studies, and has been translated into over 25 languages.
The terms are often used interchangeably. Strictly speaking, psychometrics is the measurement of psychological attributes (such as introversion-extroversion, openness to experience, conscientiousness, agreeableness etc.). Psychographics emerged in the 1960s as an application of the core principle of psychometrics-which holds that intangible human qualities can be quantified-in the commercial sphere. Psychographics adds to the mix non-psychological variables like demographics, product preferences, and purchasing history and divides consumers into distinct functional segments. By including product-specific questions about past purchasing habits or future intention to purchase, marketers can identify those groups most likely to buy specific products and services.
Although Xamplify's software correlates purely psychological information with consumer data (as in psychographics above), we prefer to use the term "psychometrics" to distinguish the rigor of our psychological profiling from traditional psychographics, which is comparatively unsystematic, has not been effectively incorporated into software applications, and has no mechanism for the testing and refinement of customer models.
There are many professional journals devoted to psychometrics: Psychometrika; Educational and Psychological Measurement; Journal of Educational Measurement; Journal of Educational and Behavioral Statistics; Applied Psychological Measurement Journal of Classification; Psychological Methods; Journal of Mathematical Psychology British Journal of Mathematical and Statistical Psychology; Behaviormetrika (Japanese).
This is the basis of Item Response Theory, with every response we get we gather a wealth of information allowing us to eliminate multiple subsequent questions. This technique, also thoroughly tested and validated in the US in other industries (SAT, HR, etc.), allows a computer to draw from a large bank of questions, the ones that yield the most differentiating and conclusive information about a person. We have several academic and applied references to this effect that demonstrate its effectiveness, its accuracy and highly predictive power and its advantages
The questions are designed by experts to be incisive; by asking pointed value-oriented questions, we gain a much deeper understanding much more rapidly than the more conventional approaches used by our competitors. Also, the questions are designed through a semantic ontology that determines several layers of information besides the explicitly asked information in the "lingo" and relevant metrics of our client's domain.
We know the information we gain is accurate and highly predictive of future behavior (in terms of demonstrating actionable preferences) of end-users, based on extensive field testing and validating decades of primary research. What's the connection between someone's psychometric profile and their demographic and/or transactional profile?
Demographics are by nature an arbitrary and independent measure and have some predictive power, but is usually not sufficient, especially in modern, mobile times. Transactional information is inherently historic and outdated, although generally somewhat predictive, assuming you have vast amounts of it. Psychometrics allows a very high confidence of prediction of future activity with a very small amount of information when combined with demographic and transactional behavior.
The connection occurs at an individual level by the construction of holistic models of end-users (using Xamplify's engines) that are not only highly predictive (for analytics and strategic insight), but also real-time and extremely portable for immediate operational deployment.
These real-time models can then be grouped or clustered based on a multitude of attributes including psychometric parameters, rather than just demographic or transactional parameters.
Approximately 900 to 1000 users will provide a statistically valid sample of the core psychometric information through the dialogue collection mechanism. This sample is used by Xamplify's extrapolation algorithms to build models for the entire customer-base of a client. For a company like Charles Schwab, this sampling size would be approximately 0.02% of their customers.
The extrapolation algorithms rely on a 3-stage association/correlation technology that is well established and used in other disciplines in the US (e.g., presidential election predictions-notwithstanding the most recent one). These techniques were architected and audited by our Chief Statistician (and Stanford University Professor), Dr. Dave Donoho.
Clusters are derived empirically and are based on configurable settings. Based on extensive testing, 18 clusters have emerged as being the most consistently representative grouping of consumer behavior across businesses.
Not all of these 18 clusters are usually used by any one client. The system automatically detects the subset of clusters that our relevant and, based on configurable settings, those clusters emerge as the relevant ones in the strategic analytics and the operational rules.
The system also allows new clusters to emerge at different business with different customer populations.
It is inferred from the transaction stream in terms of shifts of user preferences for specific products and services, for example, moving from an aggressive investment stance to a more conservative stance.
These real-time changes are detected via a pre-programmed observation agent that updates the user-models in real-time.
Technically, no. The psychometric core of the user model is independent of the underlying demographic data.
However, the overall customer model will update to reflect a newer demographic and subsequently, the hypothesis of the model might also change, if any rule of any system relies on the model's demographic intelligence to make an automated decision
The initial dialogue-based sampling necessary is usually completed at install time (1-2 weeks). Extrapolation is then run within a day to generate user models for the entire user base.
Subsequently, the models refine over time through transactions. So, there is no further dialogue (unless the client explicitly requests it) for the system to "finish its profiling". Profiling is completed at install time and then automatically refined.
It depends on what one wants to do with the models. This is usually determined through the assessment phase of the product installation. If the models are going to be used to make highly pointed and narrow predictions, then more data is needed than if they are used to gauge broader responses. Eventually, both approaches yield great success, but need to be used where and when it makes the most sense for the client.
Also, since the transaction information is not aggregated (or collected) but merely updated, these customer models have almost no upper limit on the depth of prediction for which they might be needed.