INSIDER: Evolutionary Sales Success
By Jacob Christie, MBA Candidate
Have you ever wondered how Pandora processes your musical preferences and recommends other music you might like? Or how Amazon finds and suggests the exact products that you would want to purchase? These (and many other) popular websites use evolutionary algorithms to analyze your decisions - in these instances, which songs you "like" and what purchases you make - to help the organization understand your decision preferences and to drive future decision-making opportunities. For real estate agents, the use of evolutionary algorithms has particularly interesting implications in terms of understanding and refining marketing and sales growth opportunities. Which marketing messages will my customers respond most favorably to? Which branding options help (or hurt) my ability to reach prospective clients in my market? These and other important questions can more effectively be answered by tapping directly into the preferences of the market through methods that employ evolutionary algorithms.
THINK POINT #1: How Do Evolutionary Algorithms Work?
Algorithms are mathematical processes that utilize a series of defined conditional statements to evaluate a set of information given key parameters. Evolutionary algorithms are mathematical processes made to resemble the "survival-of-the-fittest" paradigm by mimicking the way Mother Nature selects for specific traits. In the case of Pandora or Amazon, the song/product "recommendation" process works by gathering together opinions from participants to help make selections/suggestions on behalf of the user. In industry, paid services such as Affinnova employ evolutionary algorithms to simulate the recommendation process on behalf of an organization. By connecting virtually with a pool of willing subjects, an organization can test brand concepts, marketing messages, and product ideas before they go to market. Through a web-based interface, respondents are presented with two, three, or even four messages/concepts/images at a time and are instructed to select which one they like "best." The choice the respondent makes stays active on the screen and the others disappear, replaced by new permutations. By gathering enough opinions, it gradually becomes clear which message/concept will have the broadest appeal for a given target audience out of, perhaps, many thousands of possibilities. This is particularly helpful for an organization, as consumer insights are drawn from the target audience themselves, rather than from internally-based testing or opinions.
THINK POINT #2: Evolutionary Algorithms At Work - A Sweet Example
Consider a simple example to illustrate the idea of evolutionary algorithms. Assume you work for a major producer of candy bars that wants to develop a new chocolate product. The organization knows that the candy bar market can be hard to break into with a new product, even for an established brand. To give the new product the best chance to succeed, your leadership team has been brainstorming branding and marketing concepts for the past week. You are familiar with existing production methods and know that there are, perhaps, six physical forms the candy can take (long, round bar; small pieces; wide, flat bar; etc.). Furthermore, your team has come up with five possible packaging options, four catchy names, and ten advertising messages for this new chocolate product. When all is said-and-done, 1,200 possible combinations can be created from these individual elements. Would you order 1,200 unique mockups for testing? Would you identify 1,200 test markets to try these out in? Not likely. By analyzing which combinations (physical form, packaging, name, and ad message) receive the greatest number of "up-votes" through web-based evolutionary algorithm testing, though, the organization can make a decision that will likely have the most impact on the success of the product. Evolutionary algorithms do not help sell specific products or target specific customers, per se. What they do offer, though, is the ability to test and interpret which features of a proposed product or which aspects of a brand concept or message will resonate with the greatest number of people.
THINK POINT #3: What Does This Mean for the Real Estate Industry?
Evolutionary algorithms optimize outcomes over several iterations (or generations) and require several trials. Far fewer houses are sold each day than candy bars, so the issue of informational efficiency presents itself for real estate agents looking to capitalize on the opinions of the market. A web-based test pool is an efficient way for agents to gain additional consumer opinions to help drive decision-making processes. An agent looking to employ evolutionary algorithms in daily decisions might first take stock of his current messaging - all of the marketing, branding, and methods he uses to drive sales. Which aspects of these are the primary drivers of making a sale? Ideally, the common denominators in these elements will align with an overall marketing strategy (increase brand awareness in a specific market, advertise for a particular listing, etc.). Once the agent identifies the primary elements of his messaging, he might begin to brainstorm possible options for a forthcoming campaign. For example, the agent will likely want to include some form of graphic in his promotion: a company logo for branding purposes; a headshot to convey trustworthiness; a photo of a large house ("you could live here!"); and/or hands shaking to communicate the idea of a successful negotiation. Assume he has also developed four basic ad messages to accompany the campaign: "We sell more homes than anyone else in the San Fernando Valley" to convey his historical sales volume; "Anderson Partners - the Real Estate Team You Trust!" to convey trustworthiness; "Your Dream Home Awaits" to appeal to the desire of prospective buyers; and/or "No One Knows San Fernando Like Anderson Partners" to communicate local market knowledge. The agent could conduct controlled experiments with different combinations in different areas of town and still might not come away with a unified messaging strategy to appeal to his audience. In the meantime, he will have spent substantial time and money on advertising in the name of the scientific method. Instead, the agent could retain the services of a firm (like Affinnova<-- -->) that specializes in fusing evolutionary algorithms and input from a user base to gather insights from a specific market. Around 450 survey respondents would login to a web interface and vote on permutations of the agent's prospective campaign. As respondents progress through the survey, they gradually select which combination of graphics and messages appeal to them most; at the same time, the underlying algorithmic process tracks which elements are selected most often across the sample.
THINK POINT #4: What is the Significance of the Results?
An agent could spend his time whittling-down permutations himself, testing messages and making assumptions; however, the end result will still be entirely developed from internal and/or biased opinions. There is a natural inefficiency of information between an agent and his marketplace. Even purchasing all the best market research and conducting thorough surveys and focus groups can still leave an agent with a blind spot. Why spend time and money trying to uncover the desires of your target audience when you can have them tell you instead? Because different permutations appeal to respondents for different reasons, the evolutionary algorithm process optimizes two factors: which messaging element is most important to each respondent and which combination of elements is most important to him/her. In harmony with a computerized survey method, the "human element" in the decision-making process helps eliminate combinations that users feel "don't match" in a way that a computer cannot simulate. In the real estate example, testing will ultimately yield recommendations and provide data to help drive the agent's marketing decisions.
Conclusion
There is a narrow space that a successful marketing campaign occupies as it balances the promotion of the features of the product/service, its price-point, and the methods to get the message to the customer. Many additional factors go into making a sale, of course; however, we know that these represent a few of the principal vectors at play in any home sale. As an optimization method, evolutionary algorithms begin from the broadest number of combinations of elements and gradually work towards an optimal set of characteristics best adapted to the parameters of its environment and the pressures applied by that environment. Like any other modeling methodology, though, the tool is only as good as the person wielding it and the accuracy of the inputs they provide. Still, the outcomes of employing evolutionary algorithms can have significant value for an agent willing to tap into the informational efficiencies offered through a virtual, representative sample. Look forward to the March 2013 issue of the Keller Center Research Report for another look at evolutionary algorithms and their application to the real estate industry.
. . . . . . . . . . . . . . . . . . .
About the Author
Jacob Christie, MBA Candidate, May 2013, Baylor University
Graduate Assistant, Keller Center for Research
Jacob is a graduate student from Portland, OR. He earned his bachelors degree in Linguistics and Music from Tulane and is currently pursuing an MBA with plans of transitioning into technology consulting.
