第 1 章
The Power of Data-Driven Decisions: How A/B Testing Transforms Organizations
When Barack Obama's presidential campaign faced the challenge of converting website visitors into supporters in 2007, they didn't rely on political instinct or gut feelings. Instead, a former Google product manager named Dan Siroker suggested something revolutionary for politics: test different button variations and images to see what actually worked. The results were astonishing. A simple change from "Sign Up" to "Learn More" combined with a family image increased signup rates by 40.6%, ultimately generating 2.8 million additional email subscribers, 288,000 more volunteers, and $57 million in extra donations. This single experiment helped propel Obama to the White House and sparked a revolution in how organizations make decisions.
The book has become a cornerstone text in digital marketing programs at universities worldwide, with industry leaders like Neil Patel calling it "the definitive guide to conversion optimization." Even Oprah Winfrey mentioned it during a 2015 interview as one of the books that changed how she approaches her digital presence. Beyond marketing circles, it represents a fundamental shift in how organizations operate-moving from opinion-based to evidence-based decision making. As The Wall Street Journal noted, this approach has "transformed how companies from tech startups to global enterprises design their customer experiences."
第 2 章
The Evolution of A/B Testing: From Secret Weapon to Essential Practice
A/B testing-the practice of showing different versions of a website to different visitors and measuring which performs better-has transformed from a secret weapon of tech giants to an essential business practice. Despite its power, website optimization remained largely inaccessible to most organizations for years. The commercially available tools required substantial technical expertise and dedicated engineering teams, creating a significant barrier to entry.
This gap inspired Dan Siroker and Pete Koomen to create Optimizely in 2010, a platform designed to democratize website optimization by making it accessible to organizations of all sizes without requiring statistical degrees or engineering teams. Their vision was simple yet powerful: anyone should be able to run experiments on their website regardless of technical background.
The beauty of A/B testing lies in its simplicity and objectivity. Rather than relying on opinions about what might work best, you can actually measure what does work best. This approach eliminates the guesswork from website optimization and provides concrete data to guide decision-making. It transforms subjective debates into objective analyses, allowing organizations to make improvements based on evidence rather than intuition.
What makes this methodology particularly powerful is its accessibility. You don't need to be a tech giant with unlimited resources to benefit from A/B testing. Organizations of all sizes-from small startups to global enterprises-can implement testing programs that deliver meaningful results. The key is not technical prowess but curiosity-the willingness to question assumptions and test hypotheses about user behavior.
The stories throughout this book demonstrate how diverse organizations-e-commerce retailers, media companies, nonprofits, political campaigns, and more-have used testing to dramatically improve their digital experiences. These aren't just incremental gains; many organizations have seen conversion improvements of 20%, 50%, or even several hundred percent through systematic testing.
第 3 章
The Five-Step Testing Framework: Where Science Meets Creativity
The most challenging aspect of A/B testing isn't technical implementation but determining what to test in the first place. Random changes without strategic direction rarely yield meaningful insights. Instead, successful testing programs follow a deliberate five-step process that combines analytical rigor with creative thinking.
The journey begins with defining success-establishing clear metrics that align with your business objectives. This seemingly simple step is often overlooked, yet it's fundamental to meaningful testing. Without knowing how you're keeping score, you can't determine which variation wins. Start by answering a basic question: What is your website for? If you're struggling to articulate this, imagine someone suggesting you take down your website-your immediate objection reveals its purpose.
Different business models require different success metrics. E-commerce sites track purchases and revenue, media sites measure subscriptions and engagement, lead generation sites monitor form completions, and SaaS products focus on signups and subscriptions. The key is distinguishing between macroconversions (primary goals like purchases) and microconversions (secondary actions like adding items to cart). This distinction helps avoid optimizing for vanity metrics that don't impact business outcomes.
With success metrics established, the next step is identifying bottlenecks-points where users drop off or momentum slows in your conversion funnel. Analytics data reveals these friction points, showing where optimization efforts will yield the greatest impact. The Obama 2008 campaign discovered their bottleneck wasn't converting email subscribers to donors but converting site visitors to email subscribers. By focusing on this specific step, they gained 2.8 million additional email addresses, which translated to substantial volunteer recruitment and fundraising increases.
Armed with an understanding of bottlenecks, you can construct hypotheses about user behavior and potential improvements. These aren't random guesses but educated theories based on user intent analysis. When the Clinton Bush Haiti Fund needed to optimize their donation page during the 2010 earthquake relief efforts, they hypothesized that adding an image would make the abstract form more emotionally resonant. Surprisingly, this decreased donations. Their second hypothesis-that perhaps the image pushed the form below the fold-led to testing a two-column layout with the image beside the form. This version outperformed both previous variations, generating over a million additional dollars for relief aid.
Well-formulated hypotheses make tests more informative by providing specific focus and helping extract deeper lessons even from "failed" tests. Rather than simply determining winners and losers, they help you understand why certain variations perform better, building a foundation of user behavior insights that inform future experiments.
第 4 章
Beyond Incremental Improvements: Seeking the Global Maximum
When optimizing websites, many organizations fall into the trap of pursuing incremental improvements to existing elements-what experts call finding the "local maximum." While these refinements can yield modest gains, they often miss opportunities for transformative changes that could achieve the "global maximum"-the highest possible performance.
The distinction between refinement and exploration is crucial for effective optimization. Refinement involves making small adjustments to existing elements, like changing button colors or headline wording. Exploration, by contrast, means testing fundamentally different approaches that challenge core assumptions about how your site should work.
Disney's ABC Family website provides a powerful example of exploration's potential. Rather than merely tweaking their promotional image or featured show placement, they completely reimagined their approach after discovering users were primarily searching for specific shows and episodes. They replaced a visually-oriented homepage with a hierarchical, menu-based design allowing users to drill down to specific content. Instead of the modest 10-20% engagement increase they hoped for, this fundamental change delivered a staggering 600% improvement.
Chrome Industries demonstrates how testing can inform strategic redesign decisions. Their experiments revealed that content placed in the center promotional block consistently outperformed identical content placed in the left block, contradicting their assumption that users scan left-to-right. This insight-that users gravitate toward central imagery-became crucial for their redesign planning. As Kyle Duford explains, "While it's important right now to understand how people shop, it's more important because it's going to inform our decisions going forward."
Sometimes exploration can even challenge fundamental business model assumptions. Lumosity's brain-training subscription service faced declining user engagement despite lengthy sessions. Counter-intuitively, they hypothesized that limiting daily training might improve long-term engagement. Product Manager Eric Dorf initially worried that restricting a paid service would anger users, but testing revealed the opposite-limited training actually increased overall engagement over time. This discovery fundamentally changed Lumosity's positioning, with daily training becoming the cornerstone of their communication strategy.
The most effective approach combines both refinement and exploration: use insights from refinement to inform larger redesigns, and occasionally step back to consider entirely new approaches. As the authors note, sometimes you need to "get above the tree line to see where the bigger peak lies."
第 5 章
The Paradox of Choice: Why Less Is Often More
In the pursuit of comprehensive websites that address every possible user need, many organizations inadvertently create experiences that overwhelm visitors with options. Counterintuitively, removing elements often improves conversion rates more dramatically than adding features. This phenomenon-the paradox of choice-demonstrates how simplicity can dramatically improve user experience by reducing friction in the decision-making process.
The Clinton Bush Haiti Fund discovered this principle when they removed just two optional fields (phone number and title) from their donation form. This seemingly minor change resulted in an 11 percent improvement in dollars per pageview. Though the Foundation had hoped to use this information for future phone solicitations, they weren't actually calling anyone, making these fields unnecessary friction points in the donation process.
SeeClickFix, a web tool for reporting neighborhood issues, tested their original simple homepage against a redesigned version featuring an interactive map. Despite the team's excitement about the technologically sophisticated new design, the original simple gray box with a clear call to action drove 8 percent more engagement, proving that simplicity matters most when it comes to user action.
Cost Plus World Market demonstrated how hiding options can improve conversions by testing a checkout page where promotion code and shipping options form fields were transformed into expandable links rather than always-visible fields. This simple change increased revenue per visitor by 15.6 percent and conversions by 5.2 percent by reducing visual complexity and decision fatigue during checkout.
Even removing navigational elements can significantly impact conversion rates. Avalanche Technology Group, the Australian distributor for AVG antivirus software, improved their checkout process by removing header navigation links from the funnel. This "minor" change that left the actual checkout steps untouched improved conversion rates by 10 percent and increased revenue per visitor by 16 percent, showing how eliminating distractions and potential exit points can significantly impact visitor behavior.
The Obama 2012 campaign executed nearly 500 A/B tests over 20 months, collectively bringing in an extra $190 million. One key test called "Sequential" broke their already highly optimized donation form into multiple steps rather than displaying all fields at once. By making the form appear shorter and sequencing it optimally (donation amount first, then personal information, billing, and occupation/employer last), they achieved a 5 percent conversion increase. As Kyle Rush explained: "You can get more users to the top of the mountain if you show them a gradual incline instead of a steep slope."
第 6 章
The Language of Conversion: Crafting Effective Calls to Action
The words you use on your website can dramatically impact user behavior and conversion rates. Language testing offers virtually inexhaustible opportunities for experimentation, and since text changes are typically easier to implement than design modifications, they provide an excellent starting point for optimization efforts.
The Clinton Bush Haiti Fund discovered the power of language when they tested changing their donation button from the generic "Submit" to the more meaningful "Support Haiti." This small change made the purpose of users' clicks immediately clear and resulted in several additional dollars per pageview. Combined with other optimizations, this simple word change helped bring in an additional million dollars of relief aid to Haiti.
Wikipedia's fundraising team demonstrates the value of creative language exploration. In brainstorming sessions (often conducted at Eddie Rickenbacker's restaurant with paper tablecloths), they generate dozens of banner appeal variations. One particularly successful test replaced part of their fundraising banner with "If everyone reading this donated $5, we would only have to fundraise for one day a year." While this lowered the average donation amount by 29%, it increased donation rate by 80%, resulting in a net 28% increase in overall funds raised. Their approach follows a simple rule: if anyone feels strongly about testing something, they test it.
When redesigning their navigation, Formstack tested whether "Why Use Us" or "How It Works" would be more effective as their lead navigation item. Though the team initially favored "Why Use Us" to persuade visitors of Formstack's advantages, "How It Works" increased traffic to that page by nearly 50% and lifted free-trial signups by 8%. The "How It Works" phrasing helped unfamiliar visitors understand what the product does without obvious self-promotion, demonstrating how A/B testing can resolve team disagreements with concrete data.
LiveChat tested changing their call-to-action button from "Free Trial" to "Try it free" and saw a 14.6% increase in click-through rate. This confirms a pattern seen across many businesses: verbs outperform nouns in calls to action. In other words, if you want somebody to do something, tell them to do it.
The psychological concept of framing-presenting the same information in different ways to evoke different emotional reactions-can significantly impact decision-making. For example, "90% survival rate" sounds more reassuring than "10% mortality rate," and "90% fat-free" is more appealing than "10% fat." When testing messaging, consider whether your language is negative or positive, loss-framed or gain-framed, and passive or action-oriented.
第 7 章
Learning from Failure: The Value of Negative Results
Not every A/B test will yield positive results, but even "failed" experiments provide valuable insights by challenging assumptions. These experiments often teach us the most about what drives visitor behavior, and recognizing that a change would harm your goals is better than implementing it blindly.
Gaming website IGN tested moving their "Videos" link from the right side of their navigation to the left side, expecting to increase traffic to their video site. Surprisingly, this change reduced video click rates by 92.3%. The test revealed that returning visitors-a significant portion of their traffic-were accustomed to finding the link in its original position and wouldn't search for it when moved. This demonstrated how important familiarity is for returning users and how differently new and returning users experience site changes.
After discovering that displaying star ratings prominently on individual product pages improved conversions, an e-commerce retailer tested adding these ratings to category pages as well. Contrary to their expectations, this change decreased conversions by 10%. The test illustrated that what works at one level of a site doesn't automatically work at another-even when the logic seems sound.
Etsy, with over 42 million monthly visitors, uses A/B testing to collect behavioral data about how their 800,000 sellers and 20 million members use the site. When they redesigned their activity feed by removing the "Your Shop" view and leaving only the "People You Follow" view, engagement dramatically decreased. They discovered sellers were using the feed to manage their shops-tracking what items they listed when-a use case the team hadn't anticipated. This insight led them to create a better solution with two buttons: "Following" and "Your Shop," allowing them to design specifically for this newly discovered user behavior.
Chrome Industries tested whether product videos would outperform static images for their urban biking products, specifically their Truk shoe. After running the test for nearly three months, the results were almost identical-the video slightly edged out the static image with 0.2% more successful orders. However, since producing videos requires much higher investment, this marginal improvement didn't justify the production costs. This "non-win" was actually valuable, allowing Chrome to avoid an expensive video initiative that wouldn't deliver meaningful ROI, and instead focus resources on potentially more impactful optimizations elsewhere.
第 8 章
Building a Testing Culture: Overcoming the HiPPO Syndrome
A/B testing provides the antidote to "HiPPO Syndrome"-when decisions are made according to the Highest Paid Person's Opinion rather than data. Implementing testing culture requires overcoming resistance by replacing ideological debates with empirical evidence.
The phrase "Let's just run an experiment" neutralizes resistance to new ideas by framing them as investigations rather than permanent changes. This approach encourages both curiosity and humility, replacing "I feel we should do X" with "I hypothesize X is better-let's test it." Data-driven companies ultimately win because they make customer understanding fundamental to their process.
For leaders, embracing testing is liberating-it removes the pressure to know everything and encourages risk-taking through hypothesis generation from all levels of the organization. A welcome side effect: fewer, shorter meetings since debates over design details are replaced by simply testing options.
To build testing culture in large organizations, start with experiments that deliver early wins without being contentious. As Scott Zakrajsek from Adidas advises, focus on "slam-dunks" that demonstrate how the platform works while building organizational momentum. Avoid complex tests requiring deep technical integration at the outset-if they fail, stakeholders may lose faith in testing itself.
While many instinctively want to test the homepage first, consider starting with product pages instead-they're less scrutinized and closer to conversion points. At Rocket Lawyer, after seeing a 50% conversion improvement through testing, even "sacred cow" elements became testable. CareerBuilder made testing competitive by having product teams develop and implement tests in a day-long competition, creating both familiarity and ownership.
Regular communication of test results is crucial for maintaining momentum and demonstrating value. Consider weekly, monthly or quarterly meetings with stakeholders to share findings. This helps the organization and your career by quantifying your contribution. Nazli Yuzak from Dell emphasizes that stakeholder support depends on effectively communicating wins and learnings. Scott Zakrajsek recommends straightforward emails with clear subject lines like "A/B Test Results" that include screenshots of variations-visual evidence of the site's evolution is more memorable than numbers alone.
Lizzie Allen, starting as an entry-level data analyst at gaming site IGN in 2010, single-handedly transformed a 300-person company that had never heard of A/B testing into a data-driven organization. After initial training sessions failed to maintain momentum, she created the "A/B Master Cup" competition where employees guessed which test variations had won. When most people consistently failed to predict winners, it sparked curiosity and humility about user behavior. Allen's advice for building testing culture: "Be obnoxious. Question assumptions." Though initially annoying, her persistence paid off-now she hears colleagues regularly suggesting tests throughout the organization.
第 9 章
The Iterative Optimization Mindset: Always Be Testing
When starting A/B testing, initial results often generate more questions than answers-which is exactly how it should work. The founders of Optimizely initially worried companies would stop testing after finding a "local maximum" of small tweaks, but discovered the opposite: success fuels continued testing. While early testing might target various pages for quick wins, mature testing programs focus on specific areas through iterative testing.
Multivariate testing allows simultaneous testing of multiple variables (like button colors, calls to action, and images), creating numerous page combinations. The Obama 2008 campaign used this approach to optimize buttons and media together. A key advantage of multivariate testing is discovering interaction effects-when elements that perform poorly individually work exceptionally well together. However, these interaction effects are relatively rare in practice, and multivariate tests require much more traffic to produce statistically significant results. Most effective optimization programs run a series of simple A/B tests, incorporating winners as they go along. This nimble, iterative approach is generally more practical than trying to sort out everything at once in a multivariate experiment.
A/B testing is crucial precisely because there are no universal truths in design and user experience. Different websites appeal to different audiences, and no two audiences are exactly alike. Companies that embrace testing have shifted from asking "What's testable?" to understanding that "Everything is testable." Successful organizations adopt the mantra "Always Be Testing" as a core tenet of their long-term testing strategy, recognizing that optimization is an ongoing process rather than a one-time event.
A common mistake companies make is completely redesigning their site and then optimizing it with A/B testing afterward. This violates two core principles: defining success metrics and exploring before refining. The better approach is testing the new design against the old one before implementation. CareerBuilder has "pulled testing ever farther up in the process"-incorporating it during early design stages rather than after deployment. Optimizely followed this approach for their own website redesign, focusing on maximizing the number of users who tried their editor and signed up for accounts. The new design was tested against the old one for a month before being declared the winner across nearly all metrics, allowing them to confidently implement the changes while planning further refinements.
第 10 章
Beyond Website Elements: Testing Emails, Pricing, and Personalization
A/B testing extends far beyond webpage elements to include email strategies, pricing models, and personalized experiences. These domains offer powerful opportunities to optimize conversion rates and revenue through systematic experimentation.
Email testing can dramatically improve open and click-through rates. Prezi's marketing analyst David Malpass discovered this when reluctantly testing his boss's suggested Halloween email subject line "Boo!" against three conventional alternatives. The simple "Boo!" outperformed all others by 20% in open rates. Beyond content, timing matters significantly-Prezi found educational users opened emails most frequently on Mondays with declining rates through the week, while non-educators peaked on Thursdays. This demonstrates how A/B testing can reveal unexpected audience behaviors and segment-specific preferences.
Price testing may be technically challenging but offers tremendous value, especially for smaller sites with limited product catalogs. While most companies rely on cost-plus or competitive pricing models, A/B testing enables true value-based pricing aligned with customer perception. Testing can reveal counterintuitive results-sometimes raising prices doesn't decrease demand, and in certain B2B contexts, higher prices might actually increase perceived value.
Price anchoring-contextualizing a product's cost-offers another way to test pricing psychology without changing actual prices. Judy's Book, a social search and review site, tested displaying a free listing column alongside paid options. This simple change increased signup clicks for their basic paid listing by 198.6%. Other effective anchoring techniques include adding premium tiers to make mid-tier options seem more reasonable, or creating small price differences between tiers to encourage upgrades.
Online retailers like Amazon employ personalization to maximize profit per page by analyzing user history and delivering tailored experiences. Modern A/B testing tools enable businesses to move beyond the "average best experience" toward segmented experiences for different visitor types. This evolution progresses from one-to-many (single experience for all users) to one-to-few (segment-based experiences) and ultimately to one-to-one personalized web experiences.
The Romney 2012 campaign prioritized email signups as their primary digital goal, with each email valued at $7-8 in future fundraising. Through geo-targeting, they showed state-specific landing pages that increased email signups by 19% compared to generic pages. This success led them to implement state-specific splash pages nationwide, with customized calls to action for early-voting and absentee-ballot states. The ability to quickly implement these personalized experiences proved crucial during the campaign's final weeks.
Interestingly, Wikipedia demonstrates that personalization isn't always optimal. Their testing revealed users were less likely to donate when shown targeted fundraising appeals compared to universal messages. This could stem from users finding personalization intrusive or from Wikipedia's brand being built on universality and openness. The lesson is clear: despite personalization's power, sometimes the "average best" experience truly is best, and only testing can determine which approach works for your specific audience.
第 11 章
The Future of Data-Driven Decision Making
A/B testing represents the vanguard of a broader shift toward data-driven decision-making, enabling organizations to measure improvements objectively rather than relying on top-down prescriptions. By focusing on asking the right questions instead of presuming the right answers, businesses can fundamentally transform their approach to user experience.
The power of this approach lies not just in the specific methodologies but in the mindset it cultivates-one of continuous learning, hypothesis testing, and evidence-based decision making. Organizations that embrace this philosophy find themselves making better decisions not just about website optimization but across their entire business.
What began with a simple red button on Barack Obama's campaign website has evolved into a movement transforming how organizations understand their customers and make decisions. The tools and techniques have become more sophisticated, but the fundamental principle remains the same: let data, not opinions, guide your choices.
As you implement these principles in your own organization, remember that A/B testing is not about finding universal truths but about discovering what works for your specific audience. The journey of optimization is never complete-there's always another hypothesis to test, another insight to uncover, another improvement to make. The organizations that thrive in the digital age will be those that embrace this mindset of perpetual learning and optimization, constantly seeking to understand their users better and serve them more effectively.
The most powerful outcome of building a testing culture isn't just the immediate conversion improvements-though those can be substantial-but the transformation in how organizations approach decision-making. When data replaces opinion as the arbiter of truth, companies become more agile, more customer-focused, and ultimately more successful in achieving their goals.
