Nội dung

CHAPTER 10: Measuring Success: A “What’s Changed” Report And there are even more nuanced goal setups that may require you to work around problems with your site. For example, if your pages always include dynamic ID strings, you can instruct Analytics to use a match type that disregards the changing strings, so that it recognizes the same basic page as your success page. If you have Ajax presentation in a checkout or multistage formcompletion process, your “goal funnel analysis” may not work, so you can’t set goal pages for somewhere in the middle of the process, and you may not be able to correctly assess abandonment rates. I’ll leave it to you, your development team, and a bottle of ibuprofen to figure out. The Substance of Goals in Analytics I discuss this elsewhere as well, but just a quick reminder of what you might do with the information you see. You’re going to be trolling through reports to look at the performance (dollar or conversion rate wise) of particular keywords or segments such as ad groups. You can also do this for ads. One example involved an “arbitrary” goal for a financial site that had deferred revenue and a third-party page that we didn’t have access to. For Analytics to help us assess user behavior day to day, we considered “reached the application page” as a goal. We had the arbitrary goal value as a few dollars, and we decided that any keyword that had a value of over $1.00 was worth further investigation. Those below that were suspect, and those below $0.50 were red flags and might be paused. Two issues arise from this example. First, the arbitrary goal value itself proved highly arbitrary; a poor predictor of real revenues. Users merely reaching the application page helped us with quick assessments of which traffic was at least sentient and reasonably targeted, to be sure. Very low values were associated with confused users misinterpreting ad copy, weakly targeted keywords, and so forth. But many of the high-value words did not ultimately translate into backend revenues. In some financial areas, many users will show brief interest but ultimately not qualify for the product. Bounce Rates, Time Spent, and Pages Viewed Are Core Metrics in Analytics Bounce rates, by the way, were always strongly negatively correlated with the dollar values of the keywords. A variety of “you done good” metrics tend to show you about the same thing, just expressed in different ways, but this depends on your sales process. The bounce rate for a keyword (or any segment, such as a page, a campaign, a geo-segment, of your site in the aggregate) is the percentage of users who get to your website and do not click even one more time. A bounce is not the same as a “very short visit,” though it usually is that, too. Some users who “bounce” may stay on the site for several minutes, just on the first page, without clicking a link. In some cases, websites with Ajax presentations that do not take interacting users to a new page on a click may be seeing very misleading bounce rates. Time spent and number of pages viewed speak for themselves, and based on the preceding comment, it would seem that even more detailed user tracking might be beneficial to websites that want to go beyond these sometimes uninformative metrics. Check out analytics tools like GetClicky, CrazyEgg, and Robot Replay; they provide visual information about mouse click patterns and even replays of whole user sessions so you can literally get a better picture of what 273 274 Winning Results with Google AdWords is going on. Do people always go for the search box? Do they ever click the links below the fold? Are too many people clicking on an insignificant link? High bounce rates are red flags that your traffic is untargeted or unsatisfied. Unfortunately, on broad terms in paid search, you’re going to see a lot of bounce rates in the 50% range. Don’t panic at this level, although 15–30% should be your comfort level. Above 70%, it’s a bright red flag that something is amiss. Bounce rates are doubly important because they are likely being used in Google’s assessment of website quality for Quality Score, and in predictive quality assessments in the early going. Google might even look at how quickly users hit the Back button to return to Google Search; 3–6 seconds might be a huge red flag for Quality Score (and yes, even organic search algorithms are looking at user clickstream data), whereas just a little bit longer, like 20 seconds, might be treated much more leniently. For the purposes of evaluating you as an advertiser, there is likely no substantial difference between a user time spent on site of 20 seconds and 5 minutes. More may be better for you, but a bit less isn’t getting you in hot water with AdWords. But Don’t Forget to Install Google Conversion Tracker It’s helpful and fun to “troll through” these reports, as I put it above. But Google Conversion Tracker and other numbers right inside AdWords are what really put actionable analysis into high gear. Because Google Analytics is getting so much ink, newer advertisers put a lot of stock in it and actually aren’t aware of how plodding their routine is. The available reporting is comprehensive, but in fact, the paid search specific reports are not always as handy as they could be. To really motor through your routine, you want to be working right inside the AdWords interface. To do this, you’ll need to install Google Conversion Tracker JavaScript code as well. Conversion Tracker, as discussed earlier, allows you to adjust bids on keywords, adopt winning ads from an ROI or goal standpoint, and determine the relative effectiveness of segments such as content targeting. When the numbers look really bad, you’ll need to pause keywords or even re-evaluate entire campaigns, including your targeting and website effectiveness. What’s so handy is that you get to look at these numbers and adjust your bid strategy right there in the AdWords interface. AdWords Conversion Optimizer So how, exactly, do you adjust bids to ensure you’re staying within range of your target ROAS or CPA goals? There’s no universal method, believe it or not. As mentioned in previous chapters, you might want to tweak your bids on individual keywords, but on mid- to lower-volume keywords, you’ll probably need to measure ROI and manage bids for the entire group as a whole; in other words, consider keywords in “buckets.” How convenient that Google designed AdWords’ powerful logic around the ad group level, not solely at the keyword level. Without having the space to prove it here, I’ll assert that some makers of third-party bid management software are working with flawed assumptions or flawed technology. At best, it is fair to say that the primary sweet spot for such solutions is high-volume e-commerce that CHAPTER 10: Measuring Success: A “What’s Changed” Report absolutely requires intraday bid changes to avoid overpaying for product sales with razor-thin margins. Bear in mind, as well, that high-volume, frequent bid changes cost money if they’re done through a third-party application that accesses AdWords through the API. That’s why those solutions cost a lot, or pass costs directly onto you. If you have specific automation needs, you could build your own app and maybe get a break from Google on API token costs, but unless your application is really one-dimensional, you’re talking $40,000 to $200,000 in development costs. Automation that costs 3–5% of your monthly spend leaves very little in your budget for human analysis. That could be a costly budget mistake. A low-tech way to automate some of your bidding—or at least, to guide your bids to slot your ads into a specific range of ad positions on the page—is to enable Position Preferences in Campaign Settings. If you always want your ads to show up in positions 3 through 7, and to not show up if Google’s automation deems it impossible to show up in that range, you can set that range of favored ad positions. I’ve never found this to be particularly advantageous as I have limited proof that I’ll make less money if I occasionally spike up to (say) position 2 or down to 8. This setting may put you where you want to be, but generally speaking, your own bid strategy can do the same, without cutting into volume as the Position Preferences setting typically does. I am still hearing many marketers on a regular basis making assumptions that bid adjustments are needed frequently or that human analysts are inferior to “sophisticated” computers. Certainly, computers are better than us humans at handling rote tasks on high volumes. However, many of those who adopt third-party bid management solutions are wasting money and possibly handing over too much of their process (sometimes requiring more code, privacy intrusions, upsells to expensive services, and so forth) for account work that could be accomplished by (and which, indeed, may require) a qualified human analyst. Many accounts do well with bid adjustments no more frequent than a “round” of bid adjustments weekly. Some bids remain valid for months at a time. In other cases, bid gamesmanship may be warranted, but in the new complex auction, most legacy software is obsolete when it comes to the bid environment. Still, there is a certain attraction to “set and forget.” Similar to the benefit of Google Analytics, Google’s free feature, AdWords Conversion Optimizer, will adjust your bids in an attempt to keep you within a target CPA range that you set. You can set this feature at the campaign level; 100 monthly conversions are required by a campaign to qualify for this feature. My experience is that it works fairly well for certain kinds of accounts, but you may suffer from decreasing volume. The availability of this free, integrated tool (that Google engineers are working to perfect) prompts me to further question the benefit of third-party bid management. It’s notable that Google sets a hard minimum of 100 conversions per campaign to qualify for this intelligent bid automation. To be blunt, those who seem to believe that bid management solutions are going to make an appreciable difference in performance on low-volume campaigns— or worse, on poorly performing campaigns with low conversion rates—either deeply misunderstand the math involved or don’t understand the factors that combine to achieve high e-commerce performance. You should not adjust a bid until performance on a keyword or segment achieves statistical significance, and a “smart” bid does very little to rescue a dismal campaign; in many cases, the smartest bid may be zero, and you shouldn’t need a computer to tell you that. 275 276 Winning Results with Google AdWords I like alert-based bid solutions that assist human analysis by letting you know about the bids that are farthest out of your target range, or identifying keywords that are at zero ROI for a significant period of time. This way, you get the best of both worlds: the advantages of automation with the control and judgment of a human analyst. Two comprehensive campaign management solutions that take this approach are Clickable and Adapt. You might say I’m a fan.3 Quality-Based Bidding and Instability There are enough variables in the Quality Score formula, and they are sufficiently dynamic (gathering significance and certainty over time), that our ability to “read and react” is affected. Sometimes, our ad positions may rise or fall significantly as we gain or lose favor with the algorithm, but this process is even less predictable than it was in the past. In the past, the system was semitransparent at least: we could see our bids and our CTRs and know that some assessment of these was pretty much the only thing affecting our ad rank. The system is also sophisticated enough that it may be moving you up in ad position and giving you more search volume as a result because you’re more likely to convert to a sale. Don’t take the “because” literally here; conversion rates aren’t included in Quality Score calculations. But many of the signals that give Google a strong indicator of quality are the exact same factors that go into creating strong user experiences and thus higher conversion rates. The takeaway? Don’t be biased against certain ad positions. Keep reevaluating where you stand—be open to more volume, and don’t short-change yourself. Marketers Are Using Analytics to Test Sophisticated Theories I alluded earlier in the chapter to the usefulness of using statistics like bounce rates, pages viewed, and time spent to diagnose problems with your campaign. These are the deeper forms of analysis that take you beyond simply bidding to ROI, into a more thoughtful analysis of targeting and technical issues. There’s plenty more you can do in Analytics. A favorite of mine is the Ad Position report. If your Analytics account is properly linked to AdWords, you can pull up an AdWords ad position report with multiple views that you can toggle. One view just shows you how many clicks you got on any given keyword, by ad position. This is shown visually so you can see exactly how many clicks (over a given date range) occurred in the top premium spots or in the right-side margin spots. From there, if you have Goals enabled, you can go so far as to examine your conversion rates in various ad positions. If conversion rates are not significantly higher, or are in fact lower, on higher ad positions, then you’ll want to consider whether it’s better value to go for lower volume in a less visible ad position. More than anything, this excellent report is a fine way to overcome myths in the industry perpetrated by similar reports that aggregate data across the entire industry. Who cares what someone else’s account did in various ad positions, for keywords unknown to you? This is your conversion behavior, for your account, for the keywords you select. This is extremely powerful information, and Google has seen to it that you get it for free with nearly no customization effort or expense. It’s fool’s gold, of course, if you don’t look at a large enough chunk of data to get statistical significance. Also, if you haven’t experimented with different ad positions (we sometimes call CHAPTER 10: Measuring Success: A “What’s Changed” Report this a “volume test,” when we bid much higher for a week or so to secure an average ad position of 2.5 or higher), you don’t learn as much from the report. You can test dozens of other kinds of theories using Analytics data—you can, for example, assess user sophistication levels and behavior by looking at screen sizes or browser choices. You can inform usability debates, again by looking at typical monitor resolutions. You can look at conversion rates varying by geography, and decide to exclude certain areas in the geo-location settings in AdWords campaign settings. There’s not much you can get on user demographics yet; for this, you might want to check out Microsoft’s rival analytics product, code-named Gatineau. I have no doubt that if you study hard, you’ll gain enough insight to not only become conversant with the metrics shared by presenters at E-Metrics Summit—you could very well be up there on the podium yourself. Marketers Understand That “Analytics” (Relevant Statistics) Live Right Inside AdWords People who frequent the big ol’ Professional Analytics world sometimes seem to have an awfully narrow view of where to get relevant statistics. For particularly unimaginative members of this crew, analytics is something that happens within Google Analytics, or preferably, within WebTrends or Omniture, or some other third-party vendor product that they can get reseller revenue (or subsidized travel to industry powwows) out of. Analytics? How ’bout Ad Positions, CPCs, CTRs, and ROAS? Friends, you can get the majority of the numbers you need right within the AdWords interface. As we reviewed in the first part of this chapter, all of your keyword-specific, group-specific, and ad-specific data are pretty much right there in front of you inside AdWords. And yes, usually that conversion data (CPA, ROAS) is what really rules, though clients and bosses also like to hear peripheral numbers about spend, CPC, and CTR trends. Quality Score Is a Stat Did you consider that Quality Score is actually a statistic? When you see a keyword Quality Score of 3 or 1, as opposed to 9 or 10, that should be speaking volumes about your status as a good/bad guy in Google’s eyes, and as to whether you’ve managed to deal well with Google’s obsession with tight targeting for new accounts, and leniency on looser targeting once strong account history has been established. High minimum bids on mature accounts likely mean that your CTR relative to competitors is still lagging, which means you’ll need to redouble your efforts to test ads. Segmentation, Segmentation, Segmentation Marketing analysts use generic terminology to discuss segments of the target audience. In our world, the most important segments are keywords and ad groups; in some cases geography is also key. 277 278 Winning Results with Google AdWords Over on the content-targeting side, the AdWords interface offers increasingly sophisticated choices. You can choose specific publications you want to show up in, or you can let the system match content around the Web for you. Of particular interest is the excellent reports available in the Reporting tab in AdWords. You can run a report on the performance of ad placements broken down by domain, to identify poor converters. You can respond to these by entering site exclusions at the campaign level (or “negativing out”) those sites. Again, in content targeting, AdWords now breaks out content types, such as parked domains, or tragedy & destruction content. You can opt out of these, or at least see that they are performing up to par, in order to combat industry stereotypes about performance. CPCs Have Increased and Competition Has Intensified Going over the past four or five years, it’s not uncommon that a typical CPA number has tripled or quadrupled for any given company using AdWords. Multiple factors are at play here. First, Google has developed increasingly sophisticated means of monetization to bump up CPCs, including making it harder to bid at the absolute minimum on “long tail” terms. Second, many of the individual “micro-auctions” going on in the overall AdWords marketplace have reached a tipping point. With sufficient advertisers in the race for the top five or so ad positions, bid prices simply go up; sometimes, dramatically. Third, other advertisers are also testing and improving their copy and other aspects of their targeting, so they can afford to bid more. Fourth, other advertisers are improving their conversion rates, improving their offerings, offering discounts or free shipping, and so forth, so again, they achieve relatively attractive offerings at the expense of others’ conversion rates. Fifth, opaquely buried in Quality Score algorithms is a punitive minimum bid regime on clever keywords we used to thrive on, including brand names and people’s names. Google admits that ceteris parabus, “certain” keywords such as large company words that are potential trademark issues, and people’s names, are going to give you Quality Score headaches more often, and cost you a lot more. My experience has been that this essentially cripples the better-performing 5–10% of some accounts. The end result of all these factors? It’s a Darwinian environment, but also one that is tilted to favor the intelligent designer of that environment. If you stood still in the past four years, you probably saw your ROAS fall sharply, or your CPA number quadruple. If you improved steadily, it’s not uncommon that your CPA might have doubled. Everyone else feels the same pain. The big winner in all of this? Look at the annual financial statements! Look at Google’s revenues and profit margins. The winner is Google. The consumer also wins, as advertisers are forced to become more relevant and as Google actually removes ads from users’ field of view while making more money from the remaining ads. How does this affect your reading of trends? It’s tough to say, but I’m willing to go out on a limb and suggest that it’s unfair to expect CPAs to improve in absolute terms over any given four-year period. Prices on targeted clicks in our industry began at rock bottom in 1998–2002, and rose from there. Sound process, relentless testing efforts, and localized improvements need to be given due credit, and businesses need to look hard at strategy. In other words, judging a long-term testing CHAPTER 10: Measuring Success: A “What’s Changed” Report effort solely on the CPA/ROAS statistic is misleading. The environment has deteriorated for all advertisers. So, unfortunately for those who like to think in absolutes, we are left to speculate on “this is how bad our campaign would have looked had we sat still and done nothing.” Remember, many AdWords success stories are that way because a certain company was virtually the only serious one in its industry segment. The frontier is no more; it’s bound to get harder to compete as others climb whatever barriers to entry may exist, and begin going after your customers. What Hasn’t Changed (But There Is Always Hope) Given how specialized the AdWords task is, it’s unsurprising that many myths and clumsy practices still prevail among companies in the first stages of coming to grips with it. Specifically with reference to data and measurement, I’ve only gained secure beliefs and insights after years of double-checking assumptions and watching real-world performance. It’s not hard to see why someone would come to the table with certain off-the-cuff assumptions. But unfortunately many of these assumptions are wrong. Here’s a list of some of the things advertisers (I may broaden the term to call them “marketers” in this section) are still messing up.4 Marketers Must Set Objectives, but Fail to Do So Some advertisers come forth with a media “buy” mentality only: can we get x number of clicks this month for 20 cents a click? Sure. But then eventually the recriminations about the quality of the traffic, and the ROI on that spend, kick in. Even where such advertisers are just promoting something in general, management needs to engage in some kind of planning process—such as setting arbitrary goal values, placing values on impressions that may be accruing from quasi-arbitrage activities such as ad sales, or supporting existing ad listing clients—in the beginning. Asking the lower-downs in the organization to buy a lot of traffic and then questioning the value of the traffic out of the blue is, unfortunately, commonplace management practice to this day. It takes some time to defend or define ROI contributions on campaigns of this nature. This is ideally done early in the process, not after spending money for two years. Marketers Fail to Appreciate Complex System Math Let’s just take one example. CEOs of e-commerce companies are notorious for making wild claims about the “free” or inexpensive forms of online exposure they can achieve based on their “brand,” or “direct navigation,” at “no cost” or “low cost.” A manager at a major consumer electronics company once told me they get most of their “web traffic” from “direct navigation,” and implied that this was without cost. (That direct navigation, of course, spikes after TV spots and full-page color weekend newspaper ads and flyers.) In other words: in planning online campaigns, many organizations act as if their lavish offline (print, television, and so on) campaigns are done at zero cost, while assigning full cost or more to their web divisions in their efforts to gain direct response online. One organization we worked with cut its web marketing budget abruptly because a production crew for their lavish television ad shoot in Africa exceeded its budget. (Keep in mind, this company sells tents and barbecues back here in North America.) I can only assume that some of this type of behavior is more tied to internal politics and inefficient nest-feathering than to a true bottom-line concern. 279 280 Winning Results with Google AdWords The investment in offline advertising has long been largely a “faith-based initiative.”5 Online has been moving in the opposite direction. But in a world of complex customer behavior, a little faith is not a bad idea. Pure direct response isn’t feasible at every moment of every campaign. Marketers Fail to Appreciate Randomness As I’ve mentioned previously in deference to Nassim Taleb’s excellent work in the field, not every event can or should be explained or narrated. A brief spike upwards in campaign performance does not constitute a new benchmark for expected performance that you can deposit in the bank. Nor does an unexplained pause in sales flow mean much if patterns are looked at with a zoom lens. The explanation may simply be that randomness and, occasionally, highly improbable events are built-in features of the natural and social worlds. Marketers Only Pay Lip Service to Volume and Statistical Significance Issues but Do Not Understand the Math Try this exercise. Design a multivariate landing page test with 16 page permutations, with 50% of the impressions going to the test pages and 50% going to the control. Jot down your current conversion rate—say, 2%. And project your expected improvement in that rate as a result of a test—say, 25%. Now jot down how many clicks you’ll be getting to that page, per day (let’s say 500). This is a test you could run using Google Website Optimizer. The projection of how long it will take this test to achieve statistical significance is something you can figure out with a projection tool Google provides in their Google Analytics help files (at https://www.google .com/analytics/siteopt/siteopt/help/calculator.html). OK, now it’s time to guess how long the preceding test would need to run to achieve statistical significance. Done?6 And yet a quick guess might produce a “let’s rock and roll” type of prediction, such as that it would take about a month, or “we should be seeing solid improvements in a couple of weeks.” It simply doesn’t work that way when the math is this complex and your volume is low. There is hope, however. Change the test up a bit. Get to driving 1,000 clicks per day to that page. Send 85% of the traffic to the test and only 15% to the control. Chop the permutations down to 12. And hope to do better: project a conversion rate improvement of 40% over your baseline of 2%. Now you’re down to 87 days. That’s more than 12 weeks, but at least you won’t be racing against your own life expectancy getting to a completed test. This is why I’d recommend an A/B/C test for this particular example. Putting the clicks back down to 500 a day, we can still project the A/B/C test will pick a significant winner within 23 days. Nice. There Are Still Data Discrepancies Analytics experts agree that total accuracy in tracking conversions is impossible. Different tools differ even on the definition of a click. Conversions are “attributed” to the source that caused a conversion, but what does causation mean? Typically, that means the last click before the sale. Sophisticated industry research indicates that prior research may influence final sales, and choices of search terms may be narrowed as users get closer to certainty. With short cookie lengths, latent conversions may not be credited. These nuances are important to take into account, not to nitpick, but to realize that many of our points of analysis are intended to be relative and heuristic, rather than absolute and “true.” We’re Still Poor at Measuring Offline Conversions While it is rare to find a company that has mastered a “silver bullet” solution to attributing offline and phone conversions back to their CHAPTER 10: Measuring Success: A “What’s Changed” Report AdWords campaign, one thing anyone could do would be to assign a dedicated 800 number or numbers to landing pages created solely for the AdWords campaign. This and other “broad brush” attempts to give AdWords credit for sales that do not take place online are helpful in assessing the value of the investment. Yet few companies take the trouble. Some business owners prefer to hide high-ticket special order sales from marketing staff, perhaps in an attempt to downplay the website’s contribution, as a motivation for said staff to work harder. Again, the definition of a profitable campaign is up for debate, and a little faith helps lubricate operations for most any growth-oriented company. Marketers Don’t Keep Tasks in Proportion What’s probably more important to the bottom line: a killer ad headline, or many across an account, or adding 20 more negative keywords to each campaign, or 10 more negative keywords to each ad group? I’m thinking the former. Ad positions, conversion improvement, changing business contexts, and a host of other drivers are important to watch and execute on, as well. It’s easy to dwell on seemingly “overlooked” tasks, and it’s pretty tempting to believe that flashy “look busy” marketer or Google staff member is helping your business with the appearance of hustle. Sure, I think it’s important to “look alive out there,” as your Little League coach might have said. But let’s be clear: some account actions nibble at the margins whereas others are fundamental to the bottom line. Marketers Try to Sell People What They Don’t Want Perhaps this is most fundamental. Most of the mistakes in AdWords are made by trying to put your ads in front of people who just aren’t interested. So much so that Google has taken away your ability to do that by setting low initial Quality Scores on advertisers who seem like they’re probably bothering uninterested searchers. Before concluding that there is a problem with a bid, some kind of discrepancy in Analytics, a change in the “trend,” or the need for a “more timely” report, have you asked yourself whether your performance woes may be related to a real lack of demand for a particular product or service, by the people at whom you’ve targeted your ads in an attempt to attract them to your website? If your definition of “analysis” also includes “brutal honesty,” you’re ahead of the pack. Somewhere in between avid consumer interest and a complete lack of demand usually lies a middle ground. You’ll be led to a number of opportunities for additional segmentation and refinement to filter out uninterested prospects. Moving from analytics to action in paid search is often about such careful refinement over time. Endnotes 1. Google’s Matt Cutts, in his generous (and extremely accurate) review of the first edition of this book, wrote: “The cover of this book looks like your typical ‘How to use Adobe Elements/Microsoft Word/3D Studio Max’ book. I normally hate those books. Do I really need a book to walk me through what the File menu does in an application? Do I need a chapter devoted to simple things that anyone could figure out in five minutes just using the software application? No! Personally, I hate that 60% of the bookstore space is filled with books that tell you how to use a software application. Where are the books such as Hackers and Painters or The Mythical Man-Month or Tufte?” Thus exempted by the dude from Google, I’ll continue on with a view to giving you insight into the landscape 281 282 Winning Results with Google AdWords and concepts, rather than explaining how to drill into menus which may have changed by press time. 2. This book’s technical editor suggested several alternative names, such as Urchinoficles, Urchinedes the Great, or Urchinias. Sticking with Urchinator, I must point out that in this chapter as with the rest of the book, responsibility for all errors, omissions, and stubborn refusal to take suggestions, lies solely with the author. 3. Disclosure: I have been an early beta tester and adviser to Clickable. I have no stake in the company. 4. Due to space limitations, I make little effort to defend my assertions in this section. If you’re interested in deeper discussion, I’ll buy you a venti green tea and we can have it out. 5. Thanks to Avinash Kaushik for the turn of phrase. 6. The answer is 952.64 days, or getting on for three years!!! Chapter 11 Increasing Online Conversion Rates O nline conversion science, which attempts to define those factors most critical to turning a casual click into a profitable action, is one of the hottest trends in Internet marketing today. But it hasn’t been an overnight success. Earlier generations of web marketers often threw together websites hastily, optimistic that they’d get rich as long as they could find visitors. Such a strategy might have worked then, but certainly not in today’s competitive marketplace. Competition aside, the cost of paid clicks continues to rise. Achieving higher conversion rates is a must if you expect sustained profitability from paid search campaigns. Online conversion science is an emerging field, so some terminology used in this chapter will be new. I am coining the terms conversion science and conversion scientist in an attempt to describe a particular group of Internet marketers—those who specialize in increasing clients’ online conversion rates. I consider online conversion science as a subset of the broader field of web analytics. Conversion scientists are not only interested in measurement issues per se; they are engaged in “doing something about it.” There are few famous conversion scientists yet. Many have toiled behind the scenes in marketing positions, improving the ROI on online marketing campaigns for their employers by testing various online buying processes. Much of the work has been experimental and based on trial and error. Testing response online started primarily with email marketing, which is why the first wave of conversion scientists cut their teeth on email. Many marketing publications still devote a large (often disproportionate) amount of space to the subject of testing response to so-called email “blasts.” But email-offer testing protocols, often inherited from direct mail methodologies, don’t begin to scratch the surface of user interface testing that’s made possible on the Web. For our purposes, users will be arriving at some page on your site from a search—but a very specific kind of search: a paid search click whose messaging and path you have some control over. 284 Winning Results with Google AdWords A typical unsung hero in the early going of online conversion science is Marc Stockman, an email marketing consultant who was formerly a marketing VP with TheStreet.com. As a publicly traded online content company in the financial field, TheStreet.com was suffering from regular quarterly losses. In 2001, TheStreet.com only generated $15.2 million in revenues for the entire year, posting an operating loss of $31 million. To come closer to breakeven, the company wanted to dramatically increase the conversion rate of its free email newsletter subscribers to paying subscribers. Under Stockman’s tutelage, the layout and sales copy in emails, as well as the layout, copy, and checkout process on the website, were all tested and improved. This resulted in a surge in paid subscriptions and a material improvement in TheStreet.com’s profit picture. The company generated $9.5 million in revenues, and a slight profit, in the final quarter of 2004. By 2007, the company’s annual revenues were $65 million; net income was a very healthy $17 million (excluding a one-time tax benefit). Subscription revenue makes up about half the company’s total revenues. The profit picture is now so consistent that TheStreet.com pays a small quarterly dividend. There are few better examples of the power of conversion science to make or break a company. Did Stockman’s work literally mean the difference between delisting and bankruptcy, and the $235 million market capitalization TSCM enjoys today? We’ll never know for sure. But it’s safe to say that testing online response towards higher subscription rates played a large role in the company’s survival, and then, profitability. Testing key conversion pages for a highvolume marketing program may cost $5,000, $10,000, or $100,000. But the expense pays for itself many times over. The added revenue, especially for digital products, comes with negligible additional cost or overhead. And confidence rises that it’s possible to expand marketing efforts so that profitability comes with growth. Stories like TheStreet.com’s are not uncommon. Many online content companies needed to replace lost ad revenues that had put a serious hole in their business model. Many promised investors that they’d make a transition to paying subscribers, but many did a poor job of converting free readerships into paying subscribers. The rate at which prospects convert into paying customers can make or break companies of all sizes. The low conversion rate pinch has been acute of late among the proliferation of Web 2.0 companies who relied on “freemium” business models without ever figuring out whether they could convert users to paid services. If only more of them would try, rather than deferring the inevitable! Web 2.0 companies that refuse to test for profitable conversions all but resign themselves to the need for quick acquisition. They also weaken their position vis-à-vis subsequent rounds of investment; by contrast, hot companies with some provable revenue streams can afford to take less money at a better valuation. Why do some venture capitalists salivate over the Web 2.0 “free services” model? Because hot companies crash and burn without revenue, and the early-round VCs can swoop in and vastly increase their ownership stakes before figuring out whether to flip the companies or inject better business practices to turn them around. But that’s neither here nor there. In this chapter I’ll focus on the impact of website and landing page design, copywriting, and layout on conversion rates. Experience has taught us that conversion science has a number of elements, each of which plays an important part in moving an online visitor from curiosity to action. These elements include, in addition to your actual offering, the aesthetics of the user interface, ease of use, content, and credibility of your website. CHAPTER 11: Increasing Online Conversion Rates I’ll review a few elements in the overall recipe of what to test, before turning to more textured discussions of how case study information has bubbled up to create useful principles that are becoming more widely adopted across the online marketing professions today. Even Google support staff, who once were only versed in the language of clicks and traffic, are keenly aware of user experiences all the way through to goal completion. This goes hand in hand with widespread adoption of Google Analytics and Google Conversion Tracker, and the general interest Google has developed in the user experience. Conversion Science Isn’t a Beauty Contest Before we delve into the meat of conversion science, I want to cover a couple of general principles. To begin with, conversion science isn’t about turning water into wine, or putting “lipstick on a pig.” Trying to sell a blah product or a poorly conceived service on a great website is generally a prescription for failure. Therefore, make sure your offering is solid before assuming that problems converting sales are the fault of the usability specialists, data analysts, or web developers who are advising you. Great user experiences aren’t purely a function of the attractiveness of your website. You might have long-term image and revenue goals that require a consistent, contemporary design. If so, great. But many plain (Craigslist) or even ugly (PlentyOfFish.com) sites are notorious for making their founders boatloads of money while eschewing fancy designers or even usability studies! According to legend, Google’s own plain home page was conceived because the company founders “weren’t designers and didn’t do HTML.” A lucky accident indeed. Today, Google is possibly the best living example of a company that merely pretends to be laid-back about user testing and interface design, while in reality placing enormous weight on certain key factors. User response testing at a massive site like Google played an important role in the company’s overall growth. Of course, the quality of search results and the clean interface would be right up there among the key factors that have kept users coming back. Certain fanatics have held Google to its original (accidental) principles; one Google supporter is famous for sending in the number of bytes in the weight of Google’s home page, as a reminder not to “bloat” that all-important page. Google has also set a gold standard for page load times in the most complex of large database searches in the world, putting other, slower websites on the defensive. Google is also a reminder, though, of the power of the familiar. Arguably, many large companies’ sites have many “imperfect” elements if you go by the book. So, some of the current links and layout elements on any Google page are not “by the book” perfect. Usability experts can easily point to principles and practices that, on the average website, would improve the “flow.” But how about the cost-benefit of making further tweaks to a familiar experience that is making tons of money for a leading company? As part of the cultural experience expected by millions of users, leaving “imperfect” but beloved elements unchanged isn’t always rational, but it may be wise. The task facing a conversion improvement exercise, then, varies widely by the stage of business and the type of business. Unless you’re a real kamikaze visionary with very deep pockets, you’ll probably be looking to build design, copy, and interface elements that “reference” 285 286 Winning Results with Google AdWords or “look like” something an Internet user already knows about. Jakob Nielsen once humorously referred to “Jakob’s Law of the Web User Experience”; namely: “users spend most of their time on other sites.”1 If you monkey with web standards and conventions, tinkering with the meanings of known icons, design cues, word choices, and so forth, you’ll make people’s brains hurt. Don’t make folks drive through hedges and over parking barriers to get to the entrance of the drivethru. And by all means show them “conventional” photos of burgers, salads, and drinks when they get there. They’re hungry. Recently, Bryan Eisenberg solicited his blog readers to comment and advise on usability tests for the online presence of Canadian housewares store Home Outfitters. Amidst the wide variety of voluntary comments proffered by expert observers, the one that came up the most often was “please test whether replacing ‘Add to Bag’ with ‘Add to Cart’ has an impact.” The site designers had just decided to rename “Add to Cart” to “Add to Bag” in an attempt to reflect something slightly cutesy about the brand, or to place their own stamp on the exercise. I sympathize with the ennui they may be feeling in building yet another site, but it’s also our professional responsibility to generate profit unless instructed otherwise. Heck, “Add to Bag” might work out fine. On another site I’ve visited recently, the Zegari laptop bag site, I notice “Add to Shopping Bag” and don’t feel as leery about it. That’s probably because of the super-clean design of that site, and the limited choices that make it pretty clear what the user would do in this case. At the time of this writing, Zegari offers only two products: two different models of laptop bag; one of them, fittingly, is called The Minimalist. It’s OK to test innovations occasionally—Seth Godin might say that introducing a meme mutation (or mDNA) into your company’s processes gives you a better chance of evolving ahead of competitors.2 But let’s not confuse system conventions with fashion and product innovation. Sure, you need to innovate, but if you want to make money this year with a better checkout interface, my money’s on intelligent design, and sticking with Add to Cart—until proven otherwise. These principles apply to user experiences at all levels, including, of course, the design of everyday things and a variety of technology products, online or offline.3 To minimize the cost and cognitive strain of rediscovering useful conventions, we generally want the milk carton to pour in one or two standard ways. We want the Back button on our web browser to keep working like a Back button should (and we’ll hit it really quickly if we think we’ve landed on the wrong web page!). What and How to Test Depends on Business Type Debates on what is the “best” home page, category page, offer page, product page, and overall interface design will never be definitively solved. Nor will the question of how much brute force testing is warranted, as opposed to how much “borrowing from leaders” and “best practices” will need to be built into your assumptions from the start. There are probably several hundred significantly different business types when it comes to conversion improvement. Just breaking down into categories such as B2B vs. B2C, small volume vs. high volume, several demographic segments, degree of content or community possible or required, and so forth, we could arrive at a few hundred types of businesses that elicit different user responses in general. CHAPTER 11: Increasing Online Conversion Rates To get to a correct overview of your audience and probable user responses, a full-service approach might be to begin with “persona research,” and to construct offers on your site based on the known psychology of the typical visitor. That’s going to require a significant amount of organizational continuity, and a significant investment. AdWords advertisers working with modest budgets, though, can do a kind of “skinny research” on their own without as much fuss. They can test responses to a variety of landing pages, connected to relevant keywords, very rapidly. More importantly, the process of testing ads and landing pages will—I bet you—lead to a confirmation of at least one high-level principle that can simplify most of the rest of what you do. Just a friendly reminder: when testing landing pages, be careful not to “orphan” them. Google’s landing page and website quality guidelines want you to make it easy for users to figure out where they are and who you are, so include at least some elements of your global site navigation on any page or you may risk a credibility gap and lower Quality Scores. The Discovery of Scent: God’s Gift to Interface Designers? That high-level principle is now widely known in the industry as “information scent.” I won’t waste our time attempting to sort out exactly who coined the concept; whether it was Jared Spool, whether it has descended from research findings originally shared by Jakob Nielsen, or whether Bryan Eisenberg has been its leading popularizer, although all of the above are no doubt true.4 “I like a man who knows what he wants,” so say a surprising percentage of personal ads. The folks who run search engines and usability labs today often see confused people stumbling around, but still assume “that man” is buried in there somewhere, and that it is the fault of marketers, interface designers, and so on for not giving “that man” what he “wants.” True enough, the subset of individuals who really do know what they’re looking for online is significant, and our effort as conversion scientists is to make sure we don’t do anything silly that prevents “that man” from getting to what he (or she) wants. Researchers noticed early on (putting aside, for the moment, that users in these studies are usually supplied with tasks and strong intent that fuels “scent”) that a common pattern holds for users seeking relevant information and products. They keep going forward, drilling down from a search result or web referral link, into the conversion funnel or conversation interface of a website, until such time as they either complete a desired action or encounter something off topic, bewildering, or contrary to the exact thing they were seeking. With the latter, they typically leave immediately, hitting the Back button. Web users are fickle and impatient. In the inimitable words of web analytics expert Avinash Kaushik, users who leave your site immediately without clicking on a single link (known as “bounces”) are basically saying to you: “I came, I saw, I puked.” (Or maybe “that man” wasn’t in the chair that day, and the weak-willed, bored, or ambiguous person stumbled on your site. So maybe the reaction isn’t always as severe as Kaushik lets on. Maybe some users—rather than puking—just lose the scent, and wander off, like suddenly uninterested or confused bloodhounds. Or maybe some of them aren’t bloodhounds at all, but are, rather, butterflies.) 287 288 Winning Results with Google AdWords As an AdWords guy, I can summarize this scent concept in a tiny nutshell. In large part, this take on it is shared by the designers of the ad ranking formulas at Google, Yahoo, and Microsoft. Let’s say you run an ad that shows up whenever a user types bag of hockey pucks into Google. Ideally, your ad would reference the availability of a bag of hockey pucks at your store. The user would be taken to a page showing a bag of hockey pucks, and that page would allow them to easily add the bag of hockey pucks to their shopping cart. They then have no problem making that purchase. From there, you rapidly ship them said bag of hockey pucks. That’s a successful transaction based on relevance and continuity of purpose throughout the path to purchase—in other words, a strong and persistent scent. There are many complications to this oversimplification, of course. A user searching for a 2004 Reserve Cabernet from Chile might be far enough down the path of consideration that it makes sense to think of the Web as a simple store shelf. But we also know that users are often in research mode. In other cases, their queries are ambiguous—maybe even to themselves. If someone types wine, or even a more specific query about wine, it’s pretty presumptuous to show an ad like: Wine Get wine here! Free shipping on wine. Buy now. www.vino.com/yeah-baby-yeah From the standpoint of connoisseurs, wine is something you research, savor, and consider carefully—it’s not a commodity. And yet how do you develop the expertise to know enough to type in a more advanced query? That’s something that rule-bound folks who have newly discovered the concept of information scent don’t have a very good answer for. Web search cannot, surely, be merely an ordering mechanism for predetermined tastes.5 It also needs to be a discovery mechanism for the cultivation of tastes, framing of debates, and asking and answering of questions. So search engines’ role in leading users from general interest into more refined or specific needs and wants is pivotal. And corporate websites may need to play a dual role, enlightening consumers through a forum approach, building content and community while somewhere else on the site running a tight catalog. Public relations firm Thornley Fallis, for example, has focused much of its efforts for client Cuisinart Canada on creating a more vibrant corporate website that brings customer voices and brand ambassadors to the fore. Scent orthodoxy and the building of an accurate catalog-type mechanism will come into play later, when the company invests more in the e-commerce division. Both are valid objectives. It is safe to say that on broader queries, Google may want to show users fewer paid results (because they’re less likely to be what users are looking for), and may attempt to rank pages high in the organic search results that could be considered definitive, or at least very useful starting points, for users looking for information on general topics. They do this in a number of ways: by suggesting popular or related queries to broad queries, by ranking highly definitive content sites in a category, and more. In this process it is pedantic to expect designers, marketers, or anyone else to adhere to “the” principles of “the” scent, because in many cases users are not “informavores.” CHAPTER 11: Increasing Online Conversion Rates In some cases, they have yet to discover themselves, and don’t know exactly what they want. Any number of human “guides” who offer overviews of “where to begin” on a topic might be at least as helpful as machine-driven search results pointing to particularly “granular” or “accurate” pages on a particular site. (Hence the role of human-driven quasi-search services such as Squidoo, About, Mahalo, and more, as alternatives to literal-minded scent-obsessed search paradigms.) All told, then, the concept of scent is helpful, but not to the point where it’s gospel; not to the point where we should accept one-dimensional lectures or examples (like my bag of hockey pucks one) as valid for every possible user’s intent. As we’ve seen time and again, also implicit in many users’ information queries is an unspoken plea: I’m looking for someone who will help me to figure out what I really think! (Anyone who is a highly followed Twitter member can attest to that.) Common Errors That Kill Conversions These may become obvious as we study the subject in more detail, but at this point it may be useful to catalog some of the key errors made in online navigation and design that make it difficult to achieve a positive ROI on a paid search campaign. Error #1: Not Understanding What a Landing Page Is A landing page, of course, is the destination URL that users “land on” after they’ve clicked on your paid search listing. Landing pages are a fundamental variable that can make or break your numbers, but so many businesses neglect them. Here are some things to keep in mind concerning landing pages, especially if you subscribe to more literal versions of the scent theory: ■ Virtually the entire goal of the landing page must be to induce users to take a particular action, such as making a purchase or requesting information. ■ Some landing pages are better than others at achieving that goal. (Cha-ching!) ■ Designing your landing page with an understanding of what is likely to work will get you reasonably close to optimal performance from the start. ■ Testing in real time, using a valid testing protocol, is ultimately the only way to determine what kind of landing page converts best. While it is true that both your site navigation and design as a whole, and the layout and message of a particular landing page, influence user behavior, it pays to focus primarily on the landing page. Yes, your “site” is important. But excessive focus on “the site” does not necessarily help you understand how to create results. Isolating the user’s experience with the page she is actually on is always helpful to understanding your task at hand. 289 290 Winning Results with Google AdWords Error #2: Overloading the Landing Page with Information One of the most common mistakes site designers make is to try to cram too much information on the landing page. The impulse to get everything in front of the potential customer while you’ve got his or her attention can be irresistible. However, when you’re paying for traffic, you need to be very clearly focused on what you’d like the user to do, and avoid wavering from that goal. A good approach is to limit your landing page to one primary goal, plus a secondary, fallback goal. The possibility of the user finding another part of your site through an easy-to-understand navigation interface might be the third priority. Keep your options open without overwhelming the user. Landing pages that present the user with too many options rarely perform well. For example, if you’re thinking that flag animation or a “site counter” somehow enhances your credibility, think again. Eliminate what isn’t necessary. If users are particularly interested in researching your company, they can and will navigate to your About Us page. Most businesses don’t put cheesy stuff like that in the user’s face anymore, though. A more common error, at least from the standpoint of a paid click from a particular keyword, would be to add vaguely related, but still distracting, offers. So a page that is supposed to be about a particular camera accessory might also show users the featured offers that top management has decided to “unload” to consumers, for some reason making it a priority to add (say) refurbished printers to the template for the entire site. So somebody (the guy who kept too much refurbished printer inventory on hand?) gets to save face, while the clicks expressly purchased to convert buyers of a lens accessory “don’t perform.” The only solution is to empower those running paid search campaigns to test different versions of product pages, including versions with less clutter. Often, all that is accomplished by asking users to pay attention to “something else” is that you squander or divide the attention you’ve paid 20 cents, 50 cents, or $4.00 to purchase. Why ask them to pay attention to something else (see Figures 11-1 and 11-2), especially if they barely know who you are or what your main offer is? In the case of Urban Challenge Online, you can see that all of the ideas competing for attention on the page shown in Figure 11-2 are good ideas. It’s certainly not a bad idea to indicate who your charitable and corporate partners are, for example. As a page whose sole goal should be to convert visitors to registered players of a game, however, it’s cluttered both in a visual sense and in the sense of priorities facing the user. Most users have no idea what this is all about, and most have no intention yet of registering as players, but the page seems intent on changing the subject by referring to distracting issues such as an entirely different contest they could enter. At the very bottom of the page, the user is asked to “send this page to a friend.” How about converting that user to a registered contestant first? (By the way, it’s worth noting that Urban Challenge has since revamped their site significantly, and many of the early problems are now corrected.) Offer pages need to be thought of differently from content pages that users may navigate through in high volumes. These latter might contain conventional sharing icons and a variety of navigational options in a dashboard-like setting. But when you’re paying to get someone to respond to an offer, they’re arriving fresh. Narrowing their options is likely the best approach. CHAPTER 11: FIGURE 11-1 Increasing Online Conversion Rates The user’s precious attention is divided three ways. Error #3: Assuming That the Best Landing Page Is the Home Page Many site owners have the attitude that “the site has all kinds of stuff on it, so we’d better let them come to our home page and discover what we have to offer.” This works sometimes. Most times it does not. Typically, this leads to users browsing around indecisively, or worse, leads to their simply not caring that they are on “your company page” and promptly leaving. If you got them there to look at an offer, especially if that offer is related to a niche keyword, give them an offer. Depending on the type of business you’re in, you’ll be better off bringing the searcher to a tailored lead-generation-oriented landing page, a product description page, or a product category page. Don’t be coy. Requiring just one additional click at the beginning of your sales process could result in a significant drop in goal conversion rates: 30–40% lower conversion rates may be typical when the thought process towards attempting that first click on the site is particularly annoying or confusing. Don’t lose persuasive momentum by taking users to an introductory page (see Figure 11-3) that makes them decide, for example, which division of your company they want to deal with.6 If you 291 292 Winning Results with Google AdWords FIGURE 11-2 A fourth contest detail competes with six corporate logos. have a carefully targeted AdWords campaign, you should already know who you’re targeting, so you should be able to lead users to exactly the page that contains the information you want to show them. Don’t risk even a single extra click if you don’t have to. (Segmenting buyers into types is another matter: a clean process that sorts out small business buyers from corporate buyers may be absolutely appropriate.) Failure to prioritize basic information scent and navigational conventions can be particularly damaging in some situations, such as large retail organizations juggling regional priorities, infighting from franchisees, and other internal company politics. Requiring users to specify languages or postal codes prior to arriving at the desired landing page will absolutely murder conversion rates. Do whatever you can to send users to the correct page for them, even if this means making some assumptions about users or providing uniform web pricing instead of “regional” pricing. (On the language issue, you can set languages inside AdWords, and run separate campaigns for Google users searching in any language, so if you have a lot of Spanishspeaking customers and separate campaigns for Spanish language inside AdWords, there is no reason not to show that user a Spanish-language page immediately.) In high-complexity fields like insurance, users may expect to fill out forms that ask them a lot of information, and they CHAPTER 11: FIGURE 11-3 Increasing Online Conversion Rates Why start here when you can start the user on the relevant page? understand that pricing depends on many factors, including location. In retail, consumers figure the store is uniform—it’s on “the Web,” and want you to show them the product, with the price, immediately. Don’t make them puke with a blank-looking splash screen or rude information request too early in the process—not if you can help it. Error #4: Assuming That the Best Landing Page Is a Bare Contact Form If you’re in the lead-generation business, don’t expect users to be so excited by your offer that they immediately give you their contact info. For example, I clicked on a banner ad on Yahoo for a specialized service by Gomez Advisors, a consulting firm. The ad took me to a raw contact page with no accompanying copy. I had absolutely no incentive to fill out the form. A better approach would be to offer a few paragraphs of information, followed by the contact form. Another example I ran across is in the competitive credit counseling industry. Although the landing page (see Figure 11-4) has some info in the form, a few bullets, and looks like it’s been 293 294 Winning Results with Google AdWords FIGURE 11-4 While not completely barren, this landing page still doesn’t offer enough information. optimized for visual appeal and credibility, it still doesn’t offer enough to encourage the casual visitor to take action. In a highly competitive industry, your visitor has plenty of alternatives if your landing page doesn’t do the job. Your goal is to grab each prospect when you have the chance—you may not get another. Error #5: Assuming That the Best Copy Is Brief Copy If clutter is bad, some may reason that six paragraphs should be chopped to three. But that is not always the case. The more abbreviated some companies seem to get with their copy, the more they appear to be talking in industry gibberish that only makes sense at internal company meetings. If you’re Apple and there have been 40,000 news stories written about the iPod, you can get away with any length of sales copy. But if you’re a relative unknown, it often helps to tell a story. Check out the big car companies’ sites. While sometimes difficult to navigate, they generally contain a lot of material describing every facet of their vehicles, right down to the characteristics CHAPTER 11: FIGURE 11-5 Increasing Online Conversion Rates This product page provides rich detail for shoppers to sink their teeth into. of their new All-Wheel-Drive system, braking, airbags, and so on. Imagine that you made or sold the airbags, or the brakes, since you probably don’t make cars. It might help to have more than a couple of words on your site to describe the features and benefits of the product, or the reason other customers like to buy it from you. As you can see in Figure 11-5, this telecommunications equipment vendor has proper descriptions on some of its product pages, but not on others (see Figure 11-6). An early version of the site had many such omissions. Conversion rates were low. Costly clicks were not translating into buyer interest because the site seemed inhospitable and uninformative. Having proper, full-length product descriptions on your website will help you attract more “stumble-in” search engine traffic as well (often at no cost to you). Without words, search engines have nothing to index. 295 296 Winning Results with Google AdWords FIGURE 11-6 Inconsistency in adding product copy to each page has a tangible result: poor conversion rates and wasted ad dollars. Insights Leading to Principles: How Case Study Data Leads Us to Conversion “Schools of Thought” There are some pretty complex formulas floating around to help the average decision-maker make sense of the task facing them as a marketer. A few take things to an extreme of simplicity, such as the title of Steve Krug’s classic usability text Don’t Make Me Think. Others are very simple devices to help us remember key principles of user interface design, implicitly backed by much research—the concept of information scent is foremost among these. Unlike most of us, the biggest websites in the world have a lot of data to back up their assumptions about how users act online. But we aren’t always in a position to copy them. What we need to do is to combine our own user data and specific audience research with broader laws of behavior that we can learn about—and sometimes learn from—as gleaned from broader casedriven principles, plus the conventions and imperatives that the world’s largest sites (such as Amazon and Google) force us to consider implementing because, as Nielsen argued, your users spend most of their time on “other websites.” CHAPTER 11: Increasing Online Conversion Rates Case Approaches: Tinkering for Dollars—It Worked! Even a large, high-traffic website can make fatal errors that drop sales conversions by more than half, losing a company millions in the process. Bryan Eisenberg refers in presentations to a change his firm recommended to client Overstock.com. A promotional banner on a key page referred to a great new deal on DVDs for kids. Unfortunately, for the great many users who were arriving on that page looking for other things, this eye-catching promotion made it look like the whole site was about that, so they left immediately. A change to the page removed the singleissue promotion and focused on the huge DVD selection. Based on this single change, the ROI on this high-traffic page skyrocketed, with a revenue improvement impact well into six figures. We can relate little case studies like this all day long. What is at least as interesting, though, is the high-level thinking that has emerged out of a large volume of such studies. By combining data and research from many sources, great thinkers in the business have started to refer to some processes with consistent terminology. This emerging lexicon of conversion science can help us figure out why something happened, and provide us with an agenda for testing. Are We Plumbers or Persuaders? In the popular film Michael Clayton, the lead character, an underachieving but “useful” attorney played by George Clooney, wishes to dispel any assumptions about his magical powers to make problems disappear. A wealthy client facing hit-and-run charges refuses to cooperate with Clayton’s initial efforts at damage control. Clayton responds: “I’m not a miracle worker, I’m a janitor. The math on this is simple. The smaller the mess, the easier it is for me to clean up.” As an online marketer, you’re probably not a miracle worker, either. So it might be instructive (if not inspirational) to view yourself as more “plumber” than “persuader.” The more relevant your offer, the fewer miracles will be required to complete a sale. The proprietary term Persuasion Architecture—used in particular consulting contexts by Eisenberg’s firm, FutureNow—may be a bit of a misnomer. My take on it is that there is less “persuasion” going on than the term would have us believe. But yes, the path itself—the website architecture—is persuasive. I’m pretty sure a lot of beginners will misinterpret the concept of persuasive architecture as meaning talking people into buying something they otherwise wouldn’t buy. It’s far from that, in most cases. In the first edition of this book, I tried the following terminology out to classify major schools of thought in conversion science: ■ Economists These are the conversion experts who promote removal of undue barriers to commerce as the way to increase conversion rates. ■ Ideologues Conversion gurus in this camp are convinced that you need to convince or persuade users to buy your product through sales copy, pricing, psychological triggers, limited-time offers, comparisons, and other emotional elements. But a more informal-sounding way of classifying our major tasks is to substitute plumbers for economists and persuaders for ideologues. No matter which terminology you care to use, I’d like to make the counterintuitive point that it’s mostly by getting the website’s “plumbing” right 297 298 Winning Results with Google AdWords that we become persuasive. The lowest-hanging fruit—the big mistakes most companies fail to correct—is generally related to making the path to purchase smooth and uncomplicated. It’s been that way for years. Seth Godin, in The Big Red Fez: How to Make Any Website Better, falls into the “economist” or “plumbing” camp. At one point in this short beginner’s guide to usability, Godin compares a website to a Japanese game of chance, pachinko. In pachinko, disks are dropped into the top of a wooden box that contains pegs and “scoring holes” at the bottom. Whether or not you win depends on whether the disks bounce crazily off the pegs in such a way that they land in the right scoring area. It’s an interesting analogy and, along with other examples given in the book, falls into the economist camp. Godin rightly reminds readers that the business owner needs to be working with a clear endpoint in mind when designing a site and particularly when designing a landing page. Any site may require users (disks) to “bounce off a few pegs” (navigate around a bit), but the amount of bouncing around should be minimized to heighten the probability that the user winds up in the scoring area (buys your stuff or signs up for something). Is an effective “pachinko machine offer page” a “persuasive” type of architecture? Yes, but mainly due to the architecture itself, as opposed to some kind of mind control. The most extreme ideologues (or persuaders) are the ones who want you to believe that if you just adopt the right design elements and, in particular, choose the right words to “hypnotize” prospects, you’ll create an audience of automatons willing to lap up your wisdom and buy your stuff. Some even delve into neurolinguistic programming to add scientific cachet to their advice and training. But if you can achieve your goals without consulting experts at that extreme, I’m not sure these theories are much more than interesting curiosities. At the very least, they’re not appropriate for everyone. Eisenberg (along with his team at FutureNow) is one of the best-known conversion scientists in the world today, and he seems to marry elements from both camps. As anyone gets more experience in the web marketing field, and as we develop a larger database of results, we tend to gain a more holistic point of view. On one hand, Persuasion Architecture principles (or, for that matter, the “radical decluttering” philosophy from Godin’s Red Fez book) can easily lead to common-sense testing points to help us fix what is glaringly broken or unduly distracting on a web page. But on the other hand, a long-haul perspective on the data we see can lead to an ongoing exercise in theory construction and debate among a variety of experts. In that sense, there are no right or wrong answers in a given usability debate, but a lot of great food for thought. Ever thought of the issue of shopping cart “abandonment,” for example? Your knee-jerk response to people frequently abandoning the process at stage 3 or 4 might be that there is something wrong with stage 3 or 4. Why? Poorly targeted traffic, rampant comparison shopping, a weak brand, and any number of reasons might mean that an abandonment is predetermined by the user in a high percentage of cases, as they enter the process. They are the quintessential tire-kickers. Or, they may well be interested, but are just budget conscious. They become a latent conversion a year later, smoothly sailing through a checkout process that you wrongly assume is “broken.” As Eisenberg brilliantly argues: if only data mining and web analytics were as easy as eyeballing a predigested stat like “cart abandonment rates.” It isn’t always so.7 CHAPTER 11: Increasing Online Conversion Rates As an antidote to uninformed hand-wringing about a “bad cart experience” (presumably, you will fix such problems, too, without expecting a smooth cart to be a panacea), if you “focus on providing relevant and persuasive content based on understanding visitor intent, easily inferred from keywords (or ad copy, or the e-mail they arrived from), you’ll have a much higher overall conversion rate.”8 Plumbing aside, your most targeted prospects will be looking for more, not less, content. So lowering the rates of abandonment is a good start, but the next step will be to satisfy the information requirements of those who want to stick around your site awhile. (This is where a well-intentioned interest in creating a “smooth” and “uncluttered” page can lead site developers to fail to test long copy, and lead companies to undercommunicate in general.) Eisenberg wants us to understand who’s coming to our site and to create content tailored for them. That’s where plumbing stops, and persuasion takes center stage. It’s logical to work harder on providing this relevant content to visitors, considering that it’s often you, with your carefully written ad triggered by carefully chosen keywords, who induced the user to visit the site in the first place! Eisenberg’s theories are easy to corroborate by looking at your own web stats. Some time ago, by looking at data generated by web analytics service ClickTracks Optimizer, I learned something significant about the users on one of my sites. Those who bought something or filled out a quote request form stayed on the site for an average of seven minutes. Many of them not only read the long copy on one landing page, but also looked at the About Us and other background pages to ensure that they felt comfortable with me and my business. By contrast, those who didn’t take action—those who stumbled in from search engines, for example, but quickly found they weren’t interested—lasted only an average of 15 seconds on the site. Those users who spent seven precious minutes of their time reading through my material are the most interesting to me. Experiences like these bear out Eisenberg’s theories about persuasion. Although I contend that there is more “architecture” in Persuasion Architecture than people realize, Eisenberg brings a truly holistic view to the task, reminding us that clean and simple doesn’t mean empty or vacant. Remove Barriers to Conversion (Unclog the Plumbing) Think of all the reasons you didn’t buy something online this year because it was difficult to do so: ■ ■ ■ ■ ■ ■ ■ The site was too slow to load. The page was broken somehow. The site wasn’t optimized to properly display in your browser. The site wanted you to install a plug-in. The checkout process was lengthy. You had to become a member before you could buy (or read) anything. There were so many options, you got confused and left. 299