1. Title of the paper CONVERSION RATE OPTIMIZATION THROUGH EVOLUTIONARY COMPUTATION 2. Authors Risto Miikkulainen, Neil Iscoe, Aaron Shagrin, Ron Cordell, Sam Nazari, Cory Schoolland, Myles Brundage, Jonathan Epstein, Randy Dean, Gurmeet Lamba. Sentient Technologies, Inc., One California St., Suite 2300, San Francisco, CA 94111, +1 415-422-9886, firstname.fastname@sentient.ai. Risto Miikkulainen also: The University of Texas at Austin, Austin TX 78712, +1 512 471 9571, risto@cs.utexas.edu. 3. Corresponding author Risto Miikkulainen; risto@cs.utexas.edu 4. Abstract Conversion optimization means designing a web interface so that as many users as possible take a desired action on it, such as register or purchase. Such design is usually done by hand, testing one change at a time through A/B testing, or a limited number of combinations through multivariate testing, making it possible to evaluate only a small fraction of designs in a vast design space. This paper describes Sentient Ascend, an automatic conversion optimization system that uses evolutionary optimization to create effective web interface designs. Ascend makes it possible to discover and utilize interactions between the design elements that are difficult to identify otherwise. Moreover, evaluation of design candidates is done in parallel online, i.e.\ with a large number of real users interacting with the system. A case study on an existing media site shows that significant improvements (i.e.\ over 43\%) are possible beyond human design. Ascend can therefore be seen as an approach to massively multivariate conversion optimization, based on a massively parallel interactive evolution. 5. Criteria satisfied E, F, G 6. Justification of criteria satisfied Many e-commerce companies employ teams of "conversion scientists" whose task is to design web interfaces that convert as well as possible. The human design process typically includes creating and testing new and improved solutions until no improvement can be found. The solution in this paper improves 43% over the design that was created by such a team. The solution was verified by the human team itself, thus providing a conservative arms-length evaluation. In terms of the criteria: (E) The human design process is principled and iterative. (F) The human design is the best the experts can create. (G) The design problem is difficult enough so that an entire field of expertise has formed around it. 7. Citation Miikkulainen, R., Iscoe, N., Shagrin, A., Cordell, R., Nazari, S., Schoolland, C., Brundage, M., Epstein, J., Dean, R. and Lamba, G. (2017). Conversion Rate Optimization through Evolutionary Computation. In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO-2017, Berlin, Germany). 8. Prize money, if any, will be divided equally among the co-authors. 9. Why this entry is the best This entry expands the scope of human competitive results in two ways: 1. It solves an important problem in the real world. Optimizing conversions is extremely valuable for e-commerce companies, and they spend significant resources on it. The humans whose performance Ascend in exceeding have been trained as experts in the field, and in many cases have years of experience. In effect, in Ascend, EA is being used to transform an entire industry. (In more detail: In 2016, e-commerce spent $72 billion to drive customers to their websites, with conversion rates of 2-4%. In 2014, 18% of the top sites did conversion optimization, and in 1/2017, 30%. There are several companies that offer conversion optimization tools, such as Optimizely, VWO, Mixpanel, Adobe. None of them are automated design tools.) 2. Ascend did not generate just one result that exceeds human performance---instead, in Ascend the ability to do so has been industrialized. It is an automated design system that *routinely* does a better job than humans, over and over again, in each new design problem where it is applied. (In more detail: below is a sample of web interfaces optimized by Ascend, the number of values and elements varied, the size of the search space, the length of the trial, and the percent improvement in conversion rate achieved over the best human design: #of #of #of #of weeks conv.rate values #elems #combos tested improvement Leading Euro Travel Site 18 9 512 8 43% Digital Commerce Payments 20 9 1152 3 9% Intimacy Apparel Retailer 15 4 160 8 38% Top AU Beauty Retailer 28 8 6912 8 6% Classic Car Retailer 30 8 28800 3 434% Leading Mobile Network 42 9 1296600 6 75% Comparison Shopping 40 8 241920 9 31% Annuities 11 3 48 12 24% Flower Retailer 16 8 256 8 35%) 10. Type of EC used Genetic Algorithms