Announcing a New Free Primer on Web Analytics

kickstartguide2When I was preparing to write my first book, Web Analytics Action Hero, I decided to focus on material that I felt was underemphasized or missing from the existing literature on web analytics. I wanted to help transform the industry’s reporting-centered mindset to one focused on analysis and driving action from digital data. At the time, I didn’t feel the need to re-hash what other web analytics authors had already covered. Essentially, I focused on Web Analytics 201, not Web Analytics 101.

However, since publishing my book, I’ve found there’s still a healthy appetite for basic information on web analytics–not necessarily from analytics practitioners but from the growing number of executives, digital marketers, online merchandisers, journalists, creatives, and other professionals who now rely on digital data in their roles. Even though web analytics has become table stakes for most businesses with any kind of online presence, it’s still a marketing technology that is vastly underutilized, often misused, and frequently misunderstood. I felt it was time to return and take a fresh look at the fundamentals of digital analytics—which have evolved significantly over the years—to help close this persisting knowledge gap.

I’m happy to announce that I’ve finished a new 150+ page ebook, Web Analytics Kick Start Guide: A Primer on the Fundamentals of Digital Analytics, which is now available as a free download from Adobe Press (PDF or ePUB formats). Yes, you read that right – 100% gratis.


While my first book presumed its readers were somewhat familiar with web analytics, I make no such assumption in this new ebook. It actually serves as a prequel or supplement to my Web Analytics Action Hero book. It is intended for individuals who want to gain a better understanding of the technology that supplies most of the online metrics they see in their dashboards, weekly reports, and internal presentations. With most organizations investing more and more in digital, the audience for this 101-level content will continue to expand.

In Web Analytics Kick Start Guide, I start by exploring the evolution of the web analytics industry and then I dive into the business, technical, and process essentials that all aspiring data-driven professionals should know. To prepare you for what’s covered in this new primer, I thought it would be helpful to share a brief overview of the four main sections:

1. The Definition and Evolution of Web Analytics

  • What web analytics is and what it can do for your business
  • The origins of web analytics and how it has matured as a technology over time

2. The Business Essentials of Web Analytics

  • How your online business strategy and goals define what should be measured
  • Definitions and gotchas for commonly-used web metrics
  • Business model-specific KPIs

3. The Technical Essentials of Web Analytics (for the Non-Technical)

  • Data collection overview for page tagging
  • Overview of cookies and reporting architecture
  • Deep dives into key areas such as interaction/event tracking, campaign tracking, mobile/cross-device measurement, data enrichment, and tag management

4. The Process of Digital Measurement

  • The steps involved in an effective digital measurement process (data collection through data usage)
  • Organizational maturity levels for digital measurement

I want to highlight this book is not intended to be a product guide or manual for any particular product. In the technical section, I do map technical concepts to actual product capabilities so readers have a contextual reference to help them grasp the information. While I did this explicitly for Adobe Analytics and Google Analytics, the book’s core concepts should be readily transferable to similar features across most other web analytics tools.

metrics_explained2For analytics practitioners, this primer will be a helpful tool for getting your internal customers or stakeholders to up to speed on the field of web analytics. Obviously it doesn’t focus solely on your unique business model or the details of your particular implementation, but it can lay a useful foundation of knowledge to build upon. Even seasoned digital analysts might learn something new from this book. I know I learned during the process of writing it as I identified and filled gaps in my own knowledge.

I want to thank everyone who contributed to this primer—in particular those who shared their valuable insights and feedback with me. I look forward to hearing from you after you’ve had an opportunity to read this new book too. My expectation is that this primer will help you to better understand this exciting area of analytics and help improve the return you receive from it. Kick start your digital analytics proficiency by downloading my free ebook today!

Posted in Career development | Tagged , , | Leave a comment

Five Pitfalls that Will Derail Your Data Storytelling

derail2Analysts and marketers are recognizing the importance of telling stories with data. For too long we’ve watched data-intensive presentations fail to connect with internal stakeholders and have little impact on decision making. Data storytelling represents a powerful way of bridging facts with emotion to make your insights more engaging, compelling, and memorable for business audiences. However, as you use more storytelling techniques in your data presentations, it’s important to consider five ways you might be inadvertently undermining your effectiveness as a data storyteller.

 1. Not knowing your audience

unknowns2Being familiar with your audience sounds so basic and simple. However, too often analysts presume to “know” their audience and then—surprise, surprise—end up completely missing the mark. If you don’t clearly know what’s important to your audience and what their priorities are, you might as well be spinning a roulette wheel with your presentation. The simple fact is the less you know about your audience, the more likely you’ll fail. You could have a great data story, but it could be the wrong one for your specific audience.

2. Using unfamiliar analytics jargon

analytics_jargonMost people don’t live in the analytics tools and aren’t necessarily immersed in the numbers like analysts are. Rather than expecting your audience to understand our language, we need to speak theirs—typically that means putting things in business terms. You must make a conscious effort to translate what you’re going to share into something that your audience will comprehend. In most cases, that means not overwhelming them with references to statistical terms (e.g., correlation coefficients or R-squared values) or analytics tool features such as eVars (SiteCatalyst) or Regex formulas.

3. Providing too much detail

An opportunity to present your analysis findings to internal stakeholders shouldn’t become an excuse for a data dump (some call it data puking). Your audience can only absorb so much information in one sitting. You need to be selective about what you share with them. Analysts often make two key mistakes when deciding what to include in their presentations.


First, they feel obligated to substantiate or defend every insight. While you should be prepared to answer questions, most audiences are going to trust your expertise. If you need a safety blanket of supporting details, move it to the appendix of your presentation and reference it only as needed.

Second, analysts often want to show the steps or process they used during their analysis. Whether it’s an attempt to demonstrate how much effort was spent on the project or display their analytical ingenuity, most audiences won’t care about the steps or processes you used—just the insights you uncovered. To use an analogy, they are interested in sampling the delicious cake you’ve baked—not examining the ingredients you used or inspecting each step in how it was prepared.

4. Leaving out valuable context

contextIn contrast to the previous point, this is where an analyst ignores or leaves out relevant information for their analysis. There are two ways that a lack of context can ruin a data presentation.

First, when analysts don’t have adequate context into what they’re analyzing, they can go down misguided paths with their analyses. For example, if an analyst doesn’t know the promotional discount was recently reduced from 40% off to 10% off, he or she may unnecessarily jump through various hoops to explain why online sales are down this month. Eventually, any data set will surrender some kind of insight after it is sufficiently prodded, probed, or tortured. However, it won’t necessarily be accurate or useful without the right context. As an analyst you don’t want to waste cycles analyzing something that can be easily explained by a simple piece of information that lives outside of the data you’re examining. You want to secure as much context upfront to avoid this type of scenario.

Second, audience members also need sufficient context to properly comprehend the insights you share with them. As analysts, it can be difficult for us to NOT KNOW what we know. Chip and Dan Heath, authors of Made to Stick, referred to this as the Curse of Knowledge. Your audience hasn’t examined all the data forwards and backwards like you have, and therefore, they may not draw the same conclusions without the same context. You need to make sure you don’t overlook vital contextual information that’s in your brain but not in your slides. For example, you may know which types of video content performed better last year, but does your audience know this? You need to insert enough context into your data presentations to help frame your insights.

5. Talking too much and not allowing for discussion

discussion2Lastly, as an analyst it’s natural to get excited about the insights you’ve uncovered. However, you need to be careful that you don’t spend too much time presenting your findings and not allow adequate time for discussion. Ultimately, you want your audience to ACT on your findings, not just HEAR them. In most cases, stakeholders will have questions and may need to discuss what to do with your insights. Are you leaving enough time to accommodate questions and discussion at the end of your presentation?

A few years ago I watched a smart analyst deliver a great presentation to a number of senior executives. Unfortunately, he used the entire meeting to present his findings and didn’t reserve any time for discussion. I saw the panic on his face when all of the hard-to-schedule executives start packing up their laptops near the end of the meeting, and no conversation had occurred around what they would do based on his findings and recommendations. Opportunity missed. It took another couple of weeks before he could get everyone back in a room to determine next steps.

Storytelling with data is an essential skill for all digital analysts and data-driven marketers. It can make the difference between insights being adopted or ignored. Don’t let the aforementioned pitfalls prevent you from driving value from the nuggets you’ve discovered. To reiterate, it all starts with knowing your audience. With them in mind, you will want to avoid unnecessary jargon that they won’t comprehend. You’ll then want to determine how much detail and context are necessary so they can fully grasp what you’ve uncovered. Finally, you’ll want to leave ample time for discussion so together with them you can move the ideas forward. Good luck fellow data storytellers!

Posted in Data Storytelling | Tagged , , , , | Leave a comment

How Big Should Your Digital Analytics Team Be?

If you’re an NBA fan, you might know about the Los Angeles Lakers’ recent troubles with injuries (Kobe Bryant, Pau Gasol, Steve Nash, etc.). On Feb. 5th, the team actually ran out of eligible players while playing against the Cleveland Cavaliers. With only 3:32 left in the game, the Lakers’ already depleted eight-man roster shrank to only four eligible players after two additional players were injured and two more fouled out of the game. In case you’re not familiar with NBA basketball, each team needs to have five players on the court and can have up to 12 active players on its roster. Luckily, due to an obscure NBA rule the Lakers were able to keep one of its fouled out players on the floor and close out the game by beating the Cavaliers 119 to 108.


Normally, you’ll see six or seven players on the bench during a typical NBA game.

As humorous and unique as this situation was for the LA Lakers (and disappointing for Cavaliers fans), I wonder how many companies don’t have enough players on their digital analytics roster. On several occasions people have asked me how many individuals they should have on their analytics team. One of the biggest challenges that impedes many digital analytics programs from being successful is insufficient staffing. On a recent business trip to New York City, I met three companies that were struggling with inadequate analytics resources. One organization had no one owning its analytics program, another had one individual dedicated 15-20%, and the remaining billion-dollar company had only one dedicated analyst. Ouch.

Just like it’s hard to win basketball games with an incomplete roster, it’s also difficult to extract value from your data if you don’t have enough analysts, especially if you want to move beyond just basic measurement and reporting. While I wish I could give you a simple answer such as “you need at least X people”, it’s never that easy. The right number will be different for each organization, and the sweet spot can change over time as business needs evolve.

After interacting with lots of different analytics teams, I’ve identified several factors that contribute to how many people are needed on an analytics team. While I haven’t developed a magical formula or algorithm that will spit out the ideal number of analytics resources, I can share some team size considerations that I recently gave to a customer. They might help you to evaluate the resource needs of your specific digital analytics program.

First, I’ll lay out the baseline requirements and then review the different factors that will expand or contract the specific number of people you’ll need at your company.

Key Roles for a Digital Analytics Team

Before I get into how many people you’ll need on your analytics team, it’s important to understand the basic roles on a typical digital analytics team. Just like a basketball team has different positions such as power forward, point guard, or center, digital analytics teams also have different players:

  1. Program owner: Every company needs someone to own or manage the digital analytics function. Within smaller companies, a single individual may wear multiple hats and handle all aspects of the digital analytics program (requirements gathering, implementation, project management, reporting, analysis, etc.). They may even be tasked with other non-analytics responsibilities such as email marketing or paid search that pull them away from focusing on just digital analytics. Most mid-to-large companies have a manager or director who oversees the operations and performance of the digital analytics program. This role is essential to leading your data-driven efforts and gaining any sort of sustained momentum and success with analytics.
  2. Technical staff: At a minimum, you’ll need a web developer (JavaScript expertise) to deploy and update your analytics implementation. In many cases, this individual won’t live on your digital analytics team and will typically come from your IT group. For larger, more complex organizations, an embedded technical resource may lead major deployment projects as well as guide implementation standards and processes. While tag management solutions can lighten the burden placed on this type of resource, it won’t entirely remove the need for technical expertise.
  3. Analyst staff: Companies generally need at least one analyst to handle the reporting and analysis responsibilities. In some situations, the analytics manager might also be the analyst who distributes reports and insights throughout their organization. This analyst/owner combination will only work for small companies with limited analytics needs. Most organizations require a team of several analysts to glean insights from their digital data and use their recommendations to optimize the business. Analysts’ responsibilities also go beyond just slicing and dicing the data such as gathering business requirements, architecting new measurement strategies, tweaking existing tags, validating new deployments, training and supporting users, etc. One of the biggest mistakes that companies can make is to understaff in this critical function, which ultimately limits how much value they can extract from their web analytics tools.

Note: A fourth role may be needed at organizations that have a high volume of analytics projects. A project manager can prioritize, schedule, and manage different analytics projects. Project managers may also be essential in establishing and maintaining analytics-related processes.

These three key roles form the basic building blocks of every digital analytics program. As long as you have at least one person fulfilling one or more of these roles, you have the beginnings of a digital analytics practice. However, getting off the ground with digital analytics and really capturing value from it are two separate things.

I’ve found that organizations that struggle with digital analytics have often underinvested in analytics staffing. While consultants can bridge short-term staffing gaps, companies eventually discover they need in-house talent to take their analytics programs to the next level. When a digital analytics team is understaffed, it is susceptible to significant setbacks created by employee turnover. Whereas an adequately-staffed team can generally absorb losing a key individual, an understaffed team can’t. I’ve seen web analytics programs take two or three steps back simply because one key employee left, and the team couldn’t easily recover from losing their expertise and knowledge. You don’t want to avoid this potential situation at your company.

Whereas you only need one program owner and some ad-hoc technical coverage, you will most likely need multiple analysts. Primarily, most of the staffing concerns will center on this resource type as they are the foot soldiers (and potential action heroes) of your digital analytics team. The following factors will shape how many analysts you will need for your particular business:

  • Number of business units or groups that require analytics support. How many business units require data or information on their digital properties? How many internal teams or groups need help with reporting and analysis? Multinational organizations need to factor in localization needs and language coverage as well (i.e., a San Francisco-based analyst isn’t going to be as beneficial to your Nordics region as someone based in Stockholm).
  • Number of internal customers per business unit. How many active customers are there in each group? How many potential customers could you have within each business unit? A single large internal team may need more than one analyst. In other cases, a single analyst may be able to support multiple teams if each group only has a few internal customers. I’d estimate that a single analyst can support between 10-20 internal customers who need more than just basic measurement and reporting support (i.e., in-depth analysis and actionable recommendations).
  • Relative data-neediness of customers. How frequently are customers asking questions about the data? How sophisticated or complex are their questions? Are they simply looking for recurring reports or actual analysis with findings and recommendations? On one hand, an analyst who is only building and sharing reports and dashboards could potentially “support” hundreds or even thousands of internal users. However, the same analyst could be completely consumed by analytics projects generated by a single data-driven executive. The data-neediness factor will be shaped by several considerations:
    • Corporate culture: Data-intensive or gut-driven?
    • Tool accessibility: Ability to self-service, level of tool training, etc.
    • Importance of your digital initiatives: If most of the revenue comes from other non-digital channels, digital data won’t be as critical or strategic to your business.
    • Complexity of your online business: Number of digital properties, homogeneity of those properties, mix of digital channels, frequency of major campaigns/events, release cycle cadence, sophistication of the analytics implementation, etc.
  • Number of analytics tools used. What analytics tools are being used by your company? You may have different analysts that specialize in using different types of tools such as data visualization tools, statistical packages, A/B & multivariate testing, tag management systems, etc. It’s unreasonable to expect an average analyst to be an expert on multiple systems.

While the aforementioned factors help determine how many analysts are required, there are still some other considerations that may weigh into your analytics staffing decisions:

  • Level of expertise. How experienced do you need your analysts to be? While seasoned resources can accomplish more than junior analysts, they are also more expensive to hire and retain. Using senior analysts to perform menial reporting work isn’t a good use of their skills and experience (not cost effective either). Ideally, teams should have a blend of senior and junior-level analysts. Senior analysts can be valuable mentors for junior analysts and help to accelerate their development.
  • Level of centralization. How standardized or coordinated does your analytics program need to be? How much global or enterprise-wide reporting and analysis is required? Should analysts be embedded in the business units they serve? It’s common for large organizations to use a hub-and-spoke model where a centralized analytics team (hub) works with embedded analysts in different business groups or regions (spokes). Depending on your team strategy, you may require less or more headcount.
  • Level of oversight. How wide or narrow is a typical manager’s span of control at your organization? What’s appropriate for your analytics team? You may need to factor in some sort of management layer to ensure a large analytics team can function effectively. If everything is funneling through one analytics leader then it’s going to be difficult to scale and be agile as a team. However, too many chiefs and not enough Indians can also be a problem.
  • Level of outsourcing. How much do you depend on consultants for day-to-day tactical or strategic work? As a former consultant, I recognize the important role that consultants can play in filling temporary staffing gaps, shoring up expertise, and providing an outside strategic perspective. Sometimes headcount funding can be hard to obtain; whereas, consulting fees can come from a different budget. In addition, it can be difficult for companies to find experienced talent in a timely manner so hiring consultants may be the only immediate option.
  • Level of automation. How many repetitive, menial analytics tasks (e.g., reporting) have been automated or streamlined with technology? Wherever you can use technology to offset the need for human beings to perform low-value, labor-intensive tasks will free up analyst resources to focus on more strategic, high-value responsibilities (e.g., deep-dive analysis). With analytics talent being difficult or expensive to acquire, technology can offer some supplemental support.

Each organization’s analytics needs are unique, and each of these factors will be slightly or significantly different from company to company. In order to help you evaluate your own needs using these criteria, I’ll provide a couple of sample company scenarios to help illustrate how the different factors influence the digital analytics team size.

Team Size FactorCompany A (Financial Services)Company B (Retail)
Number of business units or groups that require analytics support5 (different product divisions and regions)1 (single ecommerce site)
Number of internal customers per business unit20 active users per group (100 total)5-10 active users
Relative data-neediness of customersHigh (data-driven culture, multiple web properties & campaigns)Low (design-focused culture, physical stores more important than website)
Number of analytics tools used4 (digital analytics, online surveys, data visualization, statistical package)2 (digital analytics, testing/optimization)
STAFFING ASSESSMENTMinimum: 1 program owner + 5 analysts

Rationale: With a user base of 100+ people company-wide, the program needs someone to own this important function. With each BU having 20+ active users, they would at least need one analyst per BU. In some cases, they might need more than one analyst due to data demands and tool expertise. Junior analysts might complement a senior analyst for each BU. With the BU’s being fairly autonomous with little cross-over, there’s no need for heavy centralization. However, if governance were an issue then you might add one or two individuals for just this purpose.
Minimum: 1 program owner/analyst + 1 testing expert

Right now the data demands are relatively low, but they do have both digital analytics and testing platforms. One person couldn’t effectively own both initiatives. While many organizations lump its testing team in with its analytics team, they represent complementary but different skill sets. This company may need to lean on consultants until the organization decides it wants to invest more in this area and grow this expertise in-house.

You might go through the different factors I’ve presented and realize you’re even more understaffed then you thought you were. Don’t despair! While it’s true that without a strong executive sponsor it will be hard to get the budget to hire several new analysts at one time, it doesn’t mean you can’t start small. If you can demonstrate the staffing gap and advocate for hiring at least more person, you can begin to climb out of the hole. That might mean hiring an intern, a consultant, or a college grad as a junior analyst to begin with.

If you document and evangelize the wins that are generated by the additional help, you might be able to hire a second team member and then a third. Analytics teams generally grow when they repeatedly demonstrate their success and positive ROI to key stakeholders. Eventually, you’ll reach a team size where adding one more resource wouldn’t add sufficient incremental value to warrant an additional hire (diminishing returns). However, I’d argue most digital analytics teams are nowhere near hitting this threshold and have plenty of capacity to grow.

An HBR study found that companies with above-average analytics talent had significantly higher marketing ROI than those with below-average analytics talent (+4.18% vs. +2.51%). If you think one action hero is dangerous, imagine a whole team of them. Companies with successful digital analytics programs have figured this out—hopefully your organization will too before your competitors do.

Posted in Hiring | Tagged , , | 4 Comments

Why Everyone Loses with Google’s Secure Search — Except Google

Back in October 2011, Google first announced it would introduce encrypted search and no longer provide search keywords for users who were logged into their Google account. Originally, Google indicated that the change would only have a single-digit-percent impact on Google’s organic search traffic.

However, since then I’ve watched the percent of organic search keywords that were not provided climb steadily above 60% for all my search terms on For a website owner and data geek, this was a very disappointing trend—especially given the misleading expectation that Google had set. I came to the same conclusion as many other analysts and SEO experts that the keyword data would never return to what it once was—but I felt as though I still had some limited insights into natural search traffic that I could use.

A couple of weeks ago, I was upset to learn of Google’s recent expansion of its search encryption to all organic search keywords. This time Google didn’t even attempt to notify website owners of its new approach. When you dominate two-thirds of the search engine market and are bigger than Facebook, Netflix, and Instagram combined, I guess you don’t need to explain what you’re doing—you just do what you want.

It’s difficult to lose data that you previously had, especially in a key area like keyword data that seems so elementary or basic to managing a website. In web analytics, I’ve become accustomed to gaining new insights, not losing them so it’s a bitter pill to swallow. Before this news surfaced, I had already come to conclusion that Google was being disingenuous with its desire to protect the privacy of its users with secure search. I believe nobody wins, except Google, when it comes to its move to search encryption.

When you evaluate the secure search topic, there are three key stakeholders: individual users, websites, and Google. I’d like to review what each party stands to benefit or lose from Google’s move to fully encrypted search.

Individual Users

With the increased focus on online privacy, it makes sense that Google positioned its move to fully encrypted search as a way of protecting the personal search queries of its users. As the public grows increasingly concerned about unlawful hackers and big brother (NSA), nobody is going to question why Google is emphasizing secure search.

A Google spokesperson affirmed their position with Search Engine Land’s Danny Sullivan: “We’re going to continue expanding our use of SSL in our services because we believe it’s a good thing for users…The motivation here is not to drive the ads side — it’s for our search users.” This is wonderful news, right? Who wants to have their searches eavesdropped over an unsecure WiFi hotspot? Not me and probably not you either.

In a separate CNET article, Sullivan makes an interesting argument for Google to push the entire Web to be more secure, just like it urged publishers to introduce faster sites for better rankings. For good measure, can we also throw in an SSL unicorn prancercising across a double HTTPS rainbow? It’s not going to happen—unless it was in Google’s best interests.

Privacy is just a vehicle for Google to drive its own strategic agenda. If its individual users’ privacy were the real concern then Google wouldn’t leave a gaping loophole for advertisers to still obtain paid search keywords. Simply because a searcher clicks on a paid ad instead of an organic search result, the search query data is still passed to the website or advertiser. Many Google searchers probably don’t realize their search queries are being shared with advertisers. When you think you’re safe from prying eyes, you behave differently and are more adventurous. Individual users may be more exposed now if they feel a false sense of privacy with Google’s implied secure search.

Ultimately, the real privacy concern would be how an individual’s stream of queries could be stitched together to create an interesting user profile—something only Google is in a position to do. Each website only sees the portion of organic search keywords that brought someone to their site. A publisher would have no idea what other queries individuals are performing and what other websites they are visiting (without the help of a third-party).

In most cases, I’d argue individual users actually benefit from sharing their keywords with the websites they visit. While you may not want your co-workers, neighbors, or parents to know what keywords you’re searching for, you want the relevant websites to understand the purpose of your visit and help you find the content or products you’re looking for. Without organic keyword data, websites have less ability to help visitors accomplish their goals. This gap will lead to less precise content, more assumptions about what visitors are trying to achieve, and a potentially inferior user experience. With incomplete privacy protection and less effective user experiences, individual users will lose.


From a website owner perspective, Google’s reluctance to share organic keywords means individuals and businesses can only see search keywords for non-Google queries and AdWords campaigns. While you still have organic keyword data from other search engines such as Yahoo! and Bing, the vast majority of search traffic comes from Google. It’s like losing one of your favorite TV channels from your satellite or cable package—it sucks.

The days of the symbiotic relationship between Google Search and search marketers are gone. The popular search engine benefited from having websites optimize and improve the quality and relevance of their organic search listings by leveraging its keyword data. Now it appears Google is prepared to go it alone by removing the organic search queries. The free ride is over–literally.

Without organic keyword data from Google, you now have a sizeable blind spot in your online marketing efforts that will impact your ability to convert organic traffic from Google into repeat visitors or customers. Google offers its Webmaster Tools reports as an alternative method for discovering popular organic search queries. If you’ve used these WMT reports in the past, you will have noticed significant rounding issues and discrepancies with your analytics reports. Unfortunately, the main problem with these reports is that they are disconnected from your engagement and conversion metrics. In essence, we’ve gone back to the mid-to-late 1990s with keyword data, when it was only about popularity and not about the downstream effectiveness or impact of the keywords.

Aside from trying to infer the organic search keyword traffic from the landing pages and other onsite behaviors, websites can only get Google keyword data by continuing to invest in paid search campaigns (hmmm). Previously, you might have been able to ramp up your organic search traffic and reduce your paid search spend, but now that strategy will be less clear cut and more risky without keyword data. With the growing dependency on AdWord campaigns and the added difficulty in understanding how organic search is performing, website owners—large and small—will lose as well.


If this change isn’t a good idea for individuals or companies, why does it make sense for Google? It all centers on Google’s AdWords program, which is the company’s main source of revenue and generated $45.2 billion in 2012. An interesting infographic from WordStream showed how 96% of Google’s revenue came from advertising in 2011. Clearly, advertising revenue is critical to Google’s continued success and future growth.

Now if I asked you what is the single biggest threat or competitor to Google’s AdWords program, you might have thought Bing Ads. However, in reality Google’s biggest competition comes from its own organic search results. Whenever individuals click on an organic search listing instead of a paid search ad, Google makes no money. If companies shift dollars from their paid search budget to funding SEO efforts, Google loses potential ad revenue. Google’s superior search results become its own Achilles heel when it comes to maximizing ad revenue.

While Google wouldn’t do anything to jeopardize its effectiveness in providing relevant results to its search users (no cooking the golden goose), it did find a way to force companies to continue spending on AdWords campaigns and reduce spending on SEO— by severing the keyword data pipeline. It basically ordered a hit on its chief competitor in order to strengthen its overall ad business. Despite what Google states publically, its move has everything to do with driving its ads side and nothing to do with its search users.

By removing an organization’s ability to understand how its web content acquires and converts non-paid Google traffic, the popular search engine will force more businesses to “pay to play” in order to leverage its valuable keyword data. While most large companies already invest in AdWords, Google will hope to expand their spend on paid search ads.

In terms of SEO, what was once clear is now muddy. The shroud of mystery placed over organic keywords will either further complicate SEO initiatives or make it difficult for organizations to justify their SEO efforts. SEO professionals will need to increase their value by focusing on content marketing, page-level (not only keyword-level) analytics, and on-page conversions.

Of course, Google could choose to monetize its organic keyword data by charging for it. However, I think this would only be chump change for Google in terms of its overall ad business. Besides, I think they would be entering into antitrust territory if Google decided to provide this data in only its own analytics solutions (full disclosure: I work for Adobe). Ultimately, the real rationale for no longer providing keywords will relate to Google’s core ad business such as forthcoming mobile AdWords functionality.

It’s unfortunate that in a digital world that is increasingly saturated with data that we’re actually losing valuable keyword insights. We should be moving forward, not backwards. As a smart, data-driven organization, Google used its treasure trove of search data to make this pivotal decision. Clearly, we live and die by the sword of data in analytics. This time everyone is on the receiving end, but as Picasso once said, “Success is dangerous.” As it continues to wield its market power, Google might expose an opening for other search engines or technology startups to capitalize on. Stronger competition will benefit either individual users or websites at Google’s expense.

While I’m disappointed by Google’s recent move as an analyst, website owner, and searcher, I’m confident we can adapt to this new environment. We’ll simply do more with less. I look forward to seeing the new approaches we develop for evaluating how organic keywords perform. Onward and upward, fellow action heroes.

Posted in Data Sources | 1 Comment

Web Analytics vs. Mobile Analytics: What’s the Difference?

People are increasingly using mobile devices to interact with organizations through mobile browsers and apps. A recent study indicated that mobile devices now represent 15% of Internet traffic. In December 2012, tablet devices for the first time surpassed desktop PC and notebook sales. By the end of 2013, it’s estimated that nearly two billion apps will be downloaded each week.

If you’re not paying attention to your mobile traffic today, you will be in the not too distant future. If you’re new to mobile analytics or just getting your feet wet, I thought it would be helpful to note some important differences between traditional web analytics and the emerging area of mobile analytics.

Mobile Analytics Spans Mobile Web and Mobile Apps

Mobile analytics is generally split between mobile web and mobile apps. Mobile web refers to when individuals use their smartphones or tablets to view online content via a mobile browser. Many companies redirect these users to a mobile-specific site (typically a different subdomain such as or use responsive design to adapt content to the screen size of the user’s device or computer. Some organizations are starting to build tablet-specific sites as they are discovering that neither their mobile-specific site nor their main website ideally serves the tablet segment.

In the beginning, many smartphones didn’t support JavaScript or cookies; however, today most popular mobile devices support both of these technologies. At its core, the page tagging method for measuring web properties is similar to measuring the mobile web—with just a few caveats.

    • Data connection speeds of mobile networks can vary dramatically by location and carrier technology (3G, 4G/LTE). Mobile sites need to be light and fast so that they load quickly for impatient mobile visitors. Because JavaScript can slow down mobile site performance, the JavaScript analytics tags should be optimized for mobile devices.
    • There are several mobile-specific dimensions/reports such as device name, device type, mobile browser, and carrier network that apply only to the mobile web. Many of these reports key off of the device’s user-agent string, which is much more diverse for mobile devices than desktop computers because the device’s make and model are also included. In contrast, with desktop computers you only have to worry about two operating systems and a handful of web browsers that are regularly updated. In the mobile ecosystem, there exists a wide range of operating systems (iOS, Android, Windows, Blackberry, etc.) and mobile browsers for each device type. Web analytics vendors often partner with device library services to help map the user-agent strings to up-to-date mobile device lists. Unfortunately, this approach is not foolproof if manufacturers decide to re-use the same user-agent string for different devices such as what recently happened with Apple’s iPad 2 and iPad Mini tablets.
    • Mobile sites that leverage HTML5 can tap into the GPS location of visitors if individuals grant permission to access this data. The ability to pinpoint where visitors are by specific GPS locations far exceeds the geographic precision that can be provided for desktop visitors based on IP address.
    • While screen size or resolution isn’t foreign to website analytics, its importance increases for devices with smaller, varied form factors. The orientation of the screen (portrait or landscape view) adds a new twist that isn’t a concern in the desktop world as well as the fact that users are not interacting with the mobile site by clicking but by touching and swiping.

Although the mobile web is primarily reliant on the JavaScript-based page tagging approach for data collection, mobile app tracking uses an entirely different client-side approach that is more conducive to capturing native app activity. Web analytics vendors have created software development kits (SDKs) for various mobile platforms such as iOS, Android, Windows, and Blackberry. Analytics SDKs provide a package of pre-written code that developers can insert into applications, which can be tailored to measure different app-related dimensions and metrics.

The SDKs help to streamline the measurement process because developers don’t need to write their own unique tracking code. For example, an iOS SDK will provide measurement code in the Objective C programming language that is used to build iPhone and iPad applications. Once a mobile application has been implemented with tracking code, it will send data directly to the data collection server whenever the mobile device is connected to a mobile network.

Besides using SDKs to deploy analytics code, mobile app measurement is different from both mobile web and website tracking in the following ways:

    • Say goodbye to page views and hello to screen views. Applications don’t have pages like websites, but users do interact with various screens. You also have sessions instead of visits. Despite these subtle differences, you’re essentially trying to understand the same thing—usage. Understanding the usage of specific screens is just as important as knowing which content is or isn’t being consumed on a website.
    • For mobile devices you can measure more than just what appears on the screen as mobile app analytics can access other built-in features such as the device’s accelerometer, gyroscope, GPS, and storage capabilities. Web measurement is limited to just the content that is seen in the web browser and some basic information about the computer, IP address, and referral source. Mobile app measurement offers the ability to track new types of user interactions that aren’t seen on the Web.
    • Unique users (not visitors) are identified via user IDs instead of cookies. Unique user IDs are more resilient than cookies, which are susceptible to being deleted. Due to mobile carrier contracts, people are often locked into using the same device for at least two years. In addition, due to the personal nature of mobile devices—especially smartphones—you’re more likely to understand behavior for a specific individual as opposed to a shared family computer that could have several users. User IDs can persist across version updates so that users are not lost when they upgrade. With user authentication the same unique user can be recognized across multiple apps and devices.
    • Mobile apps have a shorter session timeout than that of websites. In general, a session will end after 30 minutes of inactivity for websites. However, for mobile apps the session timeout may be as short as 30 seconds of inactivity due to a shorter perceived attention span. In addition, when users are multitasking and the app remains idle in the background for longer than the timeout duration, a new session will be triggered when the user returns to the app.
    • Depending on how the application was developed, a user may not need to be connected to a mobile network to use the mobile application. Analytics tools can store what offline interactions occurred, record when they happened with time stamps, and then upload the data to the collection server when the user re-connects to their mobile network.
    • App development teams are frequently rolling out updates and new versions. Unlike websites where all visitors receive essentially the same experience, the app users’ experience will depend on what version they’re using. When analyzing mobile app data, you can have users who are spread across different versions with potentially dissimilar app experiences.
    • There’s a greater emphasis on cohort analysis where distinct groups of users are measured over time to evaluate app retention or churn rates. By analyzing weekly cohort groups based on when they installed the application, you can evaluate the effect different app updates are having on retention as well as the app’s overall performance in terms of engagement and conversion.

This table summarizes the differences and similarities between web analytics and each of the two mobile analytics areas.

Aside from these key differences, mobile app analytics still inherits familiar measurement practices from web analytics. Web analytics and mobile analytics may not be siblings, but they’re still closely related (first cousins? Maeby and George Michael but not as dangerous?).

For example, measuring engagement is a key emphasis for mobile app analytics—something that has been frequently measured in web analytics. When measuring the effectiveness of mobile apps, downloading the app is only the first step for users. Organizations want to know how engaging their mobile apps are and whether users are using them on a regular basis. Event tracking is employed in apps to provide insights into how users are interacting with different features within each application. Similarly, measuring in-app conversions shows how successful each app is at driving specific key outcomes.

Campaign tracking is another mainstay of web analytics, and it has surfaced on the mobile apps side. For example, campaigns can be tied to the Google Play Store (Android apps) so that you can understand which campaign and traffic source led to an app download (Note: iTunes Store currently does not support campaign tracking). In addition, some organizations are direct linking from their mobile site into their mobile apps and tracking these links like campaigns.

As long as you’re aware of the subtle differences from web analytics, mobile analytics represents an exciting, new frontier for digital analysts. Eventually, cross-channel analysis (apps, mobile, and web) will be just another day at the office for most analysts.

Posted in Mobile Analytics | Tagged , , , , | 3 Comments
Page 1 of 41234