Xamplify was a Series A start up offering Predictive Analytics applications to the financial services industry. When I began they had an alpha product, no web site, no collateral, and no customers. Within the first two quarters we created the web site, developed a complete collateral system, pushed the product through to commercial launch, and closed our first customer. When 9/11 hit, we tanked along with everybody else. We repositioned the company as a defense play, and applied the predictive model to anticipate the likelihood of someone committing a terrorist act. The product was surprisingly accurate, and Xamplify was quickly recapitalized and taken off the market.
I was responsible for Product Marketing, Business Development, MarCom, Marketing Programs, and Product Management.
Typical of a Series A company, there was very little Marketing infrastructure when I started. I had to define Messaging and Positioning, create Packaging and Pricing schemas, built out the website from scratch, and created an entire collateral and sales support infrastructure.
Given the stage of the company, what Xamplify needed most was commercial validation. The company went through a number of VPs of Sale very quickly with little results. We could not flip the switch on marketing until we had our first sale, so I took over the Sales function long enough to close the first sale, with a large credit union in San Francisco called Patelco. This was basically a cold call into the CTO’s office, he (surprisingly) answered his own phone, he was willing to take a meeting after two minutes of conversation, and we had our first contract signed two weeks after that. At this point we brought in a full-time, experienced VP of Sales and began executing.
The company required a complete marketing makeover. We built out a branding and design schema, created a complete collateral set and web site that included an ASP version of the application for demonstration purposes, and developed and launched multiple marketing programs targeting the financial services industry. Our early MarCom focus was on the analyst community, the ones covering Predictive Analytics and Financial Services Technology. Our value proposition was unusual enough that we peaked a lot of interest. See the next section for details.
In terms of communicating to potential end-users, we discovered early on in the market development cycle that our message resonated most consistently with very wonky end-users (similar to Metacode). We targeted business executives, but the gatekeeper was someone with a PhD in statistical analysis, who worked in a Fortune 500 Financial Services Company. An example of this is when we were working with Washington Mutual to determine feature requirements, we would be in a meeting with fifteen people (from their end), all of whom had PhDs, and we were able to hold their complete attention for as long as we needed it.
Xamplify’s predictive algorithm was built around psychometric analysis. The algorithm was applied to very large datasets (e.g. something on the scale of Citibank’s Customer Information File). Customers were asked to take a survey; the psychometric analysis was based on a hierarchical tree structure questionnaire (that is, the answer to the first question determined which of eight second questions would be seen next, and so on through a total of twelve questions). The questions were value-based, rather than being specific and intrusive (for example “Do you find people coming to you for advice, and if so, what sort of advice?, as opposed to “what is your current salary?).
We consistently received over a four percent response when we asked Patelco’s customers to fill out the questionnaire. The people who answered the questionnaire had profiles created that correlated their answers to the psychometric profile survey with detailed financial and demographic information (for example, the type of information normally found in a mortgage application), so for that four percent, we had very detailed profiling information. A four percent response rate is statistically significant, and we were able to extrapolate the results across the entire customer base using Hidden Markov Chains, with a resulting accuracy of nearly ninety percent.
One trait we were able to predict with high accuracy was the likelihood of someone committing credit card fraud, this was one of the key adoption drivers within the financial services market. It turned out the algorithm was also good at predicting other types of crime, such as the likelihood of flying an airplane into a building. This capability was what ultimately drove our exit strategy.