The Digital Transformation Playbook: Rethink Your Business for the Digital Age
Download 1.53 Mb. Pdf ko'rish
|
The-Digital-Transformation-Playbook-Rethink-Your-Business-for-the-Digital-Age-PDFDrive.com-
Divergent experiments are best suited for learning that explores options,
generates insights, asks multiple questions at the same time, and, when done right, generates new questions to explore in the next iterative stage. (See table 5.2.) Both types of experiments increase our knowledge and test our assumptions. Both involve looking outside the organization for answers, and both require willingness to learn versus just planning and deciding. But the approach of each type is quite different. Let’s look at them in detail. Table 5.2 Two Types of Experiments Convergent Experiments Divergent Experiments Example: A/B feature testing or a pricing test Example: putting a prototype in the hands of customers Formal (scientific) experimental design Informal experimental design Asks a precise question or finite set of questions Poses an unknown set of questions Seeks to provide an answer May provide an answer or raise more questions Needs a representative customer sample (test and control groups) Needs the right customers (who might not be average customers) Needs a statistically valid sample Sample size may vary Focused on direct causality Focused on gestalt effects and meaning Goal is to test the thing itself Goal is to test as rough a prototype as possible for the question ( “good enough”) Confirmatory Exploratory Useful for optimization Useful for idea generation Common in late stages of an innovation Common in early stages of an innovation IN COMMON Increases knowledge Tests assumptions Looks outside for answers Requires willingness to learn versus decide 130 I N N O V A T E B Y R A P I D E X P E R I M E N T A T I O N Convergent Experiments The key element of every convergent experiment is the initial causal hypothesis: “If I add this feature, customers will spend more time on my site.” Or “If I change this interaction, customers will spend more money in my store.” Convergent experiments are critical for cases where it is not enough to know the correlation between two events; you also need to verify which event is causing the other. Convergent experimentation is applicable in a variety of contexts. It can be used with any digital product or service (website, mobile app, soft- ware, etc.) to test and improve any element of the customer experience. This is why not only Google but also every major Internet service, such as Amazon or Facebook, is constantly running A/B tests, in which two sets of customers see the same webpage (or the same e-mail) with one difference in design and the company measures any difference in customer behavior or response. Facebook is famous for experimenting with the News Feed of its users to find the right balance of photos versus text posts versus videos, the friends a user is more interested in hearing from, and the kind of con- tent that is interesting only in the short term or meaningful to a friend who only logs in to Facebook several days later. However, convergent experimentation can be applied in nondigital environments as well. These kinds of experiments are at the heart of data- driven strategies to optimize the guest experience and loyalty rewards given to customers of hotels, airlines, and resorts. When convenience store chain Wawa is planning changes to the food menu, it will run experiments to measure not just if customers buy the new item but also if there’s an impact on the overall profitability of customer visits. 7 Convergent experimentation is often used in communications and direct marketing. In both presidential campaigns of Barack Obama, contin- uous, rapid experiments on e-mail subject lines and website page designs helped to dramatically increase their effectiveness in signing up new sup- porters and garnering more donation dollars. Starting in the pre-Internet era, Capital One bank used convergent experiments to test the right pro- motional offer, the right target audience, and even the right color of enve- lopes as it mailed out credit card invitations. By running tens of thousands of experiments each year that focused on customer acquisition and lifetime value, it grew from a small division of another bank into an independent company worth $42 billion. 8 I N N O V A T E B Y R A P I D E X P E R I M E N T A T I O N 131 A convergent experiment can be as expensive as testing two different store layouts for a retail chain or as cheap as sending two versions of an e-mail promotion, each to a different group of randomly selected custom- ers, and comparing the responses. Because convergent experimentation needs to measure causality, it needs to adhere to the key principles of formal scientific experiments: r Causal hypothesis—so that you have an independent variable (the cause) and one or more dependent variables (the effect) r Test and control groups—so that you can see the difference between those who are exposed to your stimulus and those who aren’t r Randomly assigned participants—so that an external factor doesn’t influence the outcome of your test group r Statistically valid sample size—so that the differences you measure can rise above the noise of random fluctuations r Blind testing—so that you avoid the Hawthorne effect, where those involved in the experiment unintentionally influence its outcome Common mistakes in convergent testing mostly center on improperly assigning participants to the test and control groups. For example, a retailer might select a set of participants (its top customers or its better-performing stores) for a new treatment and erroneously assume that “everyone else” (all its other customers or stores) can serve as an equivalent control group. Some of the key writers on convergent experimentation for business include Stefan Thomke, Thomas Manzi, Eric T. Anderson, and Duncan Simester. 9 Download 1.53 Mb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling