SemAngel

X Change - Content is King

It was a Shakespeare-heavy weekend in the Angel household beginning on Friday with my daughter's camp presentation of Much Ado About Nothing. On Saturday, we enjoyed Shakespeare under the stars with an opening performance of Midsummer Night's Dream. There was even a brief Kenneth Branagh reading during the Olympic's Opening Ceremony.

I had not my fill, for I would have liked a bit more of the spoken word at the Opening Ceremony. The transformation of the stadium from pastoral to industrial England? That was stunning. But somehow the show lost narrative focus after that and became, like Beijing, mostly circus and spectacle. We got the quintessential Englishness of great children's literature, of 4 decades of great pop music, of great winged dove cyclists...Okay, I'm lost here.

I see little hope (and less point) of topping Beijing for spectacle when it comes to these events. Daring as it might be, I'd love to see some future director seek resonance from narrative. A kind of vast stage play with the stadium for canvas might be a truly new art form: a participatory combination of theatre and the movies. Daring because a story is far harder (e.g. "Lost my phone") than circus; meaning more demanding that spectacle.

It's the challenge of meaning that faces any kind of event - Conferences like our own X Change included. Spectacle is easy (if expensive), meaning is hard.

Like every Conference, we have our share of spectacle. The setting, the hotel, the events. I won't deny we try hard to deliver on pure and enjoyable spectacle. Sometimes it even works. One of the highlights for me in Berlin was sitting atop the boat gliding under Berlin's many bridges and chatting amiably (daughter included) about accidentally "killing" phones, German cinema, and other such nonsense as the sky darkened above us and the city lights grew crisp and bright. I still remember the dinner at the Monterey Bay Aquarium from two years ago. I'm sure lovely Terranea has much to offer as well.

But like I say, that's the easy part. Getting meaning out of content - that's the real challenge.

It's a challenge X Change was uniquely created to meet. TED is proof that presentation based Conferences can deliver the goods by delivering extremely high-quality content. But the success of TED hinges on unusual (and unusually interesting) breadth not depth. The consumers of the content at TED aren't expert in the disciplines represented - indeed, that's the whole point. Inside an industry it's a little different. Effective learning for experts in an industry simply isn't a function of canned presentations however sound or rich. Experts learn from each other, in conversation. It's that simple. And that's the whole point behind X Change - bringing practitioners together to talk, to discuss, and to learn from each other.

Bringing practitioners together to talk is no guarantee of meaning. Over the years, I've sat in Huddle sessions that worked and ones that simply didn't. It's a never-ending battle to figure out ways to raise the bar on those conversations and find ways to consistently deliver meaning. We haven't found one magic bullet, but we have found one overwhelmingly important factor. No surprise, it's the participants. Even with the best participants, not every conversation will take off. But given the parameters at X Change - controlled deep-dives into topics of shared interest - it's a pretty good bet that if you have 10 to 15 participants with deep experience and interesting opinions, you're going to have one heck of a conversation.

Which is the long way around to saying that if you're doing interesting work in Web analytics, then like the U.S. Government in time of war, "I want you!"  In this case, I want as many of the most interesting, most opinionated, most experienced digital measurement folks in the U.S. to come to X Change. We're hitting that time when the Conference is filling up fast - when it's peak season for planning to attend. If you've thought about coming in years past. If the idea of Conference all about conversation with fellow professionals appeals to you. If you're doing genuinely innovative, interesting or challenging work in digital measurement. If you're dealing with particularly difficult challenges. If you know you've got a lot to add. Then you're the kind of person who will not only make the Conference better, but will get the most out of the format.

Shakespeare still sings to us, even after 400 years, because he found ways to deliver meaning within spectacle. Sure, Shakespeare is high-art as we think of it today. But it's high-art loaded with puns, innuendo, farce, physical humor and downright silliness. Perhaps it's just as true to say of a play like Midsummer Night's Dream or Much Ado About Nothing, that it's the meaning that takes us by surprise.  We are caught, off-guard amid the story and the fun, by the sudden revelation of meaning.

Good conversation is like that as well. It's the unexpected point of view, the sudden twist in topic, the challenge to something once held sure, that illuminates understanding and creates real meaning. Come to challenge and be challenged. To learn and to teach. To discover and reveal. Come to enjoy by all means - for the setting is lovely, the food ecstatic, the company pleasing - but come most of all for the conversation.

Click here to Register!

What Matters Most in Enterprise Analytics - X Change as a Mirror on our Community

This happens to be one of my favorite times of the year. It's partly for all the reasons you'd expect: long days, BBQ, having the kids around more, the fresh fruits (Grapefruit TEN for a DOLLAR in Monterey this weekend! My youngest says I should have bought a thousand and sold them in the city.). All great stuff. But what I really look forward to is my daughter's annual Shakespeare camp stint. It's one of their favorite camps, and it's certainly mine. Love Shakespeare...love performing it...provided I'm not actually on stage!

Since I'm too chicken to get up on stage myself, these annual camp practice sessions are a heaven sent opportunity for me to perform at the dinner table. I enjoy them so much I'd probably make them enroll even if they didn't want to. In this year's addition I get to channel Kenneth Branagh and Emma Thompson with Benedick and Beatrice's sparkling wit - and, I hope, exceed Micheal Keaton as Dogberry's master of malaprop - in Much Ado About Nothing.  Good fun.

Which brings me, in typical round-about fashion, to my other favorite time of year: the X Change Conference. With July well upon us, it's time to start thinking seriously about September's U.S. X Change. I have plenty of reasons for thinking this will be the best Conference we've ever done. We have the usual great line-up of Huddle Leaders, the attendees continue to represent the best and brightest in enterprise measurement, and we have the format nailed. The venue this year is the spectacular Terranea in Palos Verdes (Los Angeles) - a Conde Nast Top 100 Resort. It's an incredible setting with gorgeous, sun-filled meeting rooms (which should inspire us to great discussion), amazing settings for our events, and the likelihood of the same spectacular weather we enjoyed last year in San Diego and the year before in Monterey. I'm pretty sure people are going to love it.

Each year, about this time in the summer, I spend some time on X Change talking about the Huddles, the events, and what we have planned. This year, I thought I'd do a short series laying out some high-level thoughts around the the most interesting (to me) topics. Before I go there, however, I thought it would be interesting to take a high-level survey of the topics to get a sense of what's truly on the minds of our industry.

At X Change, Huddle Leaders mostly pick their own topics. I encourage everyone to pick topics they feel personally interested in and consider to be top-of-mind for them. Since all of the Huddle Leaders are enterprise analytics and measurement managers, the end result is a pretty good overview of what matters, right now, in our industry.

I took our 60+ break-out sessions and categorized them into six broad groupings:

• Analytics
• Organization
• Talent
• Big-Data
• Infrastructure
• Day-to-Day Tasks (automation, governance, etc.)

Almost half (46%) of the topics at this year's X Change I classified as Analytics focused. That's the highest percentage we've ever had. These analytic sessions are quite diverse - representing everything from classic site-focused analysis (Measuring B2B Website Effectiveness) to analysis of key site tools (Your Users are Talking to You - Site Search Analytics) to Testing and Personalization (Driving Personalization with Analytics Data) to a focus on methods (Social Analytics - What's Working and What Isn't or Incomplete Data - Finding the Right Balance).

There's a lot to like about this trend. First, I think it evinces a growing maturity in our industry and a focus on the actual USE of data. A couple of years ago one of my standard plaints was that nobody ever talked about analytics at Web analytics Conferences. X Change is proof that, at least at the high-end, that simply isn't true anymore. (So I guess I have to change my complaints). If you're organization is still stuck trying to get your infrastructure or organization setup to do analysis - I think that means you're behind.

Second thing to like about this trend - this is just more interesting stuff. One of the Huddle topics I classified as belonging to the "Talent" category is titled: "Where does all my time go: Or why can't I do the analysis that I love." I know this complain comes up a lot. It comes from analysts whose time is totally sucked up by ad hoc requests and report generation. And, of course, it comes from managers who've been promoted out of the trenches. There's probably no hope for managers, but at least for day-to-day analysts, the Huddle line-up at X Change suggests that a good number of enterprise measurement organizations ARE doing analysis.

The second most popular category was "big-data". If you've seen the steady growth in big-data topics at X Change over the past few years (and even more, the dramatic growth in the sophistication of the conversation), then this one is probably no surprise. What I find interesting and encouraging is the healthy skepticism evinced by a fair number of big data Huddle topics. Whether it's the "Enterprise Data Warehouse Chimera" or "Big Data, Big Deal? It's the next bandwagon, so what?", I get the feeling that not everyone is going to be hyping the hype.

Not only is that quintessential X Change - with discussions that are often quite surprisingly candid - it, too, is reflective of an significant increase in maturity with respect to the big data landscape. When organizations go beyond talk and start trying to implement big data analytics solutions, you'd expect a healthy skepticism to emerge. We all know that in the real-world, lots of stuff doesn't work quite the way we'd hoped or imagined. I, personally, am a firm believer (as must be clear from Semphonic's Whitepaper and Webinars) in this direction. But that doesn't mean I think the skeptics are wrong. Everything always turns out to be harder (and more expensive) than you expect.

Next on the list, organization and day-to-day concerns tied with about 12% each - together making up about a quarter of the Huddle topics. Organizational topics seemed geared toward increasing the profile and effectiveness of the analytics function with topics like "Web Analytics Leadership" and "An Analyst's Guide for Using Data to Influence your Website". Day to day concerns were all over the map from "Automation for (Data) Monkeys" to "Educating End Users" to the very intriguing "Applying Agile Development Methodologies to Measurement".

Talent - a growing topic in the past few years - came next. Talent accounts for a little less than 10% of the Huddles. Perhaps it ought to be more given the level of frustration and interest in finding and retaining analytics talent. But this may be one of these areas where there just isn't as much to say that isn't non-duplicative.

Infrastructure comes last (though, to be fair, topics like "Agile Development" might easily have ended up here. That still makes infrastructure a top tier concern for enterprise measurement leaders - with Tag Management being the most common topic on the board. I don't expect infrastructure concerns to go away anytime soon, but I do expect that at least some of those concerns are gradually getting subsumed into the big-data discussions.

Taken as a whole, I think the trend and distribution of topics evince a fairly consistent picture - one where analytics is increasingly coming into mature practice supported by big data and warehousing platforms, is focused on broadening its impact on the organization, and is, of course, desperately short of people!

Sound a little like your world? If so, I hope you'll join us at this year's X Change and share your experiences and conversation. Benedick's and Beatrice's most welcome. Dogberry's need not apply. Good fun to be had for all!

 

Big Data, Advanced Digital Segmentation, and Powerful Data Exploration

If you're a Seinfeld fan like me, you probably remember the classic episode in which George pursues his own personal holy grail: sex, pastrami sandwiches, and TV (simultaneously). Last week, Semphonic announced a partnership which made me think of George.

We've teamed up with InfoBright (provider of a powerful columnar database) and Pentaho (data exploration and flexible reporting) to deliver our digital segmentation and analysis in a complete package. InfoBright provides the horsepower. Pentaho provides the sleek chassis. We provide the professional driver.

Together, we've built a live demo that embodies Semphonic's ETL process plus basic dashboard with digital segmentation, InfoBright's lightening fast columnar database, and Pentaho's rich and flexible data exploration interface. We simulated almost 15MM records of data (based on actual Weblogs) to show you a real-world performance scenario and we've opened it up in the Cloud so you can play with it.

As I've said, our goal at Semphonic is to deliver our segmentation across a wide-range of enterprise ready big data platforms. Last week's announcement and whitepaper (and this week's upcoming webinar - I'll be showing off the demo!) are a major step forward.

Big data. Powerful Data Exploration. Advanced Digital Segmentation.

Trifecta!    

The Process of Enterprise Reporting

Segmentation is fundamental to enterprise digital reporting and in my last post, I showed how Semphonic is working to embed a two-tiered segmentation deep into the heart of our reports. Those reports are designed to immediately answer both who did something on the Web and what they were trying to do. These two questions - the who and the why - are critical to any deep understanding of web performance and drive the approach we call two-tiered segmentation:

 

But if a report like this - with deeply embedded segmentation - is the end goal of enterprise reporting, what's the beginning? Like everyone else, we have to go through a process of gathering business requirements and translating them into a report set. Having this type of report as the end goal fundamentally changes this process, making it at once crisper and more coherent.

Here's why.

In the old days, we'd come to the task of gathering business requirements with a fairly substantial questionnaire designed to explore what data a stakeholder looked at, what they wanted to look at, and what measures were important to them. At root, we wanted to understand how the business worked. The goal of this process was to help us isolate both Key Performance Indicators and supporting metrics that might serve as key performance drivers (KPDs).

The goal of this process (sitewide KPIs) was fundamentally mistaken, but that doesn't mean that the requirements gathering was equally flawed. In exploring  the way the business worked, we often learned a great deal that was useful and the contributed to potentially interesting reporting and analysis.

On the other hand, there was a significant disconnect between the type of questions we asked, the feedback we got, and the reports we produced. It's easy enough to see why this happened. In stakeholder meetings, the type of performance oriented metrics we heard about fell into one of two types: not Web specific and too high-level to be really interesting (Revenue) or Web specific and generally meaningless (traffic). Measures like revenue are, of course, fundamental to the business. But they aren't Web analytics metrics. Few organizations rely on their Web analytics tool for revenue reporting. To make revenue interesting at the digital level, you need to be able to show how Web-specific metrics translate into revenue. Unfortunately, the Web-specific metrics we elicited (like Traffic) are extremely poor for making that translation. So we often found that, coming out our stakeholder requirements process, we were still largely creating reports to our specifications and then having to explain how the reports emerged from the requirements. What's worse, the method we had for requirements gathering wasn't particularly effective at driving the discovery of these linkages - sometimes it worked but often enough it didn't.

As we've replaced our reporting systems, however, we've also found ourselves replacing the process of gathering business requirements with something much more specific, focused, and comprehensible.

To build the report above, we need to understand four basic things:

• The audiences to a Website
• The visit types for each of those audiences
• The business goals for each of those visit types of each of those audiences
• The way to measure the accomplishment of those business goals for a given visit type of a specific audience

So to accommodate our report design, we re-focused our process to answer these questions. That's way, these days, we do business requirements gathering by beginning with the audience. We explore with stakeholders how they think about their customers. What segmentation they use - explicit or implicit - becomes the first step in gathering business requirements. This step, as it happens, also tends to drive us into a better, deeper understanding of the business.

When we've explored the audience and segmentation, we turn to the purposes of the Website. Our goal here is to elucidate visit types, but we structure this as a conversation around how the stakeholder wants to serve the needs of each audience. By exploring these questions, we create a good model of the possible visit types on the Web. It's a model we'll have to flesh out behaviorally, but it often provides us with most of what we need to create an abstract visit-type segmentation.

In the next step, we explore with stakeholders the business goals for each type of audience and visit intent. If a "blank" comes to the website because they want to "x", what's our business goal? Sometimes, we'll get wonderfully specific answers to this type of question. Business goals might be to sell them a product, to sell them more products, to generate a lead, or to register so we can contact them going forward. But, of course, goals aren't always of that sort. Sometimes, the business goal is to deepen brand awareness or answer a customer support question. What we don't accept as an answer to this question is something like "View Page X." That's not a business goal, that's a Web metric.

When we have those business goals specific to each audience and visit type, the ball is largely in our court to find the right metrics to measure that goal. Those metrics may involve Web analytics measures like "viewed page X" or "purchased product of entry" or "showed increased satisfaction relative to control" or "exited without calling." The range of possible metrics is enormous and there are deep technical issues involved in choosing the right metric. But that's our job.

The beauty of this process is two-fold. By structuring the process in this way, we collect the information we need to help us effectively build a two-tiered segmentation using our analytic techniques. That's the putative purpose. But even better, stakeholders can directly see how their feedback drives the reporting.

Each part of their responses drives specific parts of the report they end up seeing. The audience information is, of course, embedded in the core of the top-level dashboard (1).

The visit-type information (2) is embedded at the next in the drill-downs we provide:

The business goals are embedded in the success metrics (3) that are at the heart of the reports calculations:

 

Each piece of stakeholder input flows directly into the reports and becomes a part of the organizational consensus about measuring success. This brings a completely new level of coherence and consistency to the process of building reports via requirements gathering.

Our part at Semphonic is to engineer all of this in the analytics by creating the segmentations that instantiate and identify both visitor and visit types and measure each of the businesses goals. That's what people pay us for. But it seems to me that as valuable as that nuts-and-bolts measurement expertise is, the biggest value in the process is the clarity it provides. It's a clarity that supports the creation of a consensus about what really matters in digital marketing: what audience are you trying to serve, how are you trying to support them, and what's your business goal. With that clarity, our task of tying these business goals to specific metrics becomes vastly easier and the work we provide in that realm far more compelling and useful.

Armed with a proper understanding of audience, visit-type and goals, stakeholders find it much easier to evaluate the degree to which specific metrics actually capture business success. This, in turn, deepens the usefulness of Web reporting and drives better understanding of the actual levers of change. Ultimately, it translates into the true holy grail of enterprise reporting: better marketing decisions.

The Role of Segmentation in Reporting

Over the past year, I've written repeatedly about the need for segmentation to drive effective digital reporting, site testing and data warehousing. At the same, time, I've argued that digital segmentation extends traditional segmentation by adding a new dimension that is visit specific and absolutely essential to understanding the role of a digital touch in the broader customer journey. The purpose of this visit-specific segmentation is to identify the purpose of the visit and, when cross tabulated with who the visitor is (classic customer segmentation), creates a matrix within which interesting reporting, testing, and analysis can take place.

I've spent a lot of time describing how this segmentation works and explaining why I think it's essential to digital measurement. I've spent almost as much time debunking traditional approaches to enterprise reporting including sitewide KPIs of every sort. What I haven't done, is shown how Semphonic is actually embedding our Two-Tiered Segmentation into effective reporting systems.

That's an important gap, because I think there's a widespread perception that while a two-tiered segmentation might be great, it would produce enterprise reports that are too complex to understand or too difficult for the organization to use. It's the old "crawl", "walk", "run" discussion (which I loathe) and which I never hear trotted out with the words: "And we're ready to run." I don't think organizations should be content to "walk" through life and I'm not convinced that it's a good competitive strategy. On the other hand, I think it's fair to be skeptical of our ability to embed segmentation in reports without maxxing out their complexity. That has indeed proven to be a significant challenge.

Creating a good segmentation is hard work. No doubt about that. But most organizations don't begrudge the work necessary to create the segmentation. It's the incorporation of the segmentation into a reporting system that seems formidable. It's not so much the work involved as the apparent complexity of representing multiple levels of segmentation inside a digestible and understandable enterprise report that seems nearly impossible - particularly for a generation of digital marketers raised on the "five actionable KPIs" school of measurement.

In the last six months, I think we've begun to find our way through this dilemma and create reporting systems that not only capture the essence of a two-tiered segmentation in an easily digestible form, but create a "systems" view of Web performance that has tremendous resonance with the organization.

Our approach started with a simple idea. We wanted to show how the Website as a system works. To do that, we wanted to use the Two-Tiered segmentation because that's the best general model we've found for actually modeling a Website.

One of the examples I've used repeatedly to illustrate what we're looking for in a report set is based on common real-world experience. If I walk into a digital marketing manager's office and announce that traffic to the Website is up 5%, the first question out of their mouth should be "with whom?". Because traffic outside of audience is meaningless. If those visitors aren't the right people, then the traffic is pointless. Even in mass media, there's an implicit assumption that all reach is relative to audience. No matter how large his audience, Elmo won't sell your smooth vodka. On the web, bad traffic is ubiquitous and cheap. Audience does matter.

But once I've answered the "who" question, I'm not quite done. Because the next, equally important question to ask is why they came to the site. If the answer to the "who" is "customers", it matters (very) greatly whether they visited to get product support, to buy a new product, or to register a product.  In fact, when it comes to measuring my success with that visit, it's pretty much all that matters. Success on the Website is almost entirely driven by customer goals. This doesn't mean you can't or shouldn't have business goals that are rather different from the customer's intent. A customer coming to your Website may be looking for an entertaining video or useful content. You may be hoping to get them to register or sell them a product. But their intent shapes which business goal you can reasonably choose, how you drive to that goal, and how you measure your relative success.

Of course, answering this "who" and "why" is the whole point our two-tiered segmentation.

The goal in building enterprise reporting systems is to help answer one seemingly very basic question: "Is our Website successful?" with the two-tiered approach. It should be clear that the answer to that question needs to be more complicated than a simple yes or no. A Website can be very successful with one type of visitor and visit and an abysmal failure with another.

 So in building a basic site report, we wanted a method that would capture exactly this thinking and instantly show a decision-maker how traffic in the Website is driven by audience:

 

This is a striking representation of the Website as a system. On the left is traffic flowing into the system. On the right is successes on the Website. In the middle are the audience types (the who) driving that traffic and success. The segmentation becomes the transformation layer between traffic and success.

This type of report instantly answers the key question: "Is our Website successful?" and it does so in a manner that forces a much more intelligent appreciation of the question. Now, I can look at the report and instantly see that the majority of success on the site is driven by Qualified and Upsell Customers and that the Qualified audience is now struggling.

By clicking on that specific visitor type, I can drill-down into the next tier of the segmentation and see exactly how that audience is struggling or succeeding:

The Visit Type flow tells the decision-maker instantly which types of visits are successful and which aren't for every audience coming to the site. And it does so in the same visually consistent and strikingly rich paradigm.

We've taken this same concept of systemic flows and extended it into enterprise campaign reporting. Here's a mock from a marketing traffic report showing how each significant digital channel is contribution to both traffic and success. The report allows for the middle, transformation section to include either segmentation or additional levels of marketing hierarchy:

 

We've also taken these same concepts and built more analyst oriented, Tableau-style dashboards around them. I'll show some examples of these in a later post.

To me, these reports are by far the best enterprise-level reports we've ever produced at Semphonic. They are the first reports we've been able to build that force report consumers to think productively about Website performance and digital marketing and do so in an almost painless fashion. They are beautiful, coherent, and powerful. They can include powerful models of system performance and they force an organizational consensus around both segmentation and success. They are not mere exercises in Excel, Pentaho or Tableau because all the meat is contained within the segmentation. But they deliver the segmentation in a fashion which makes consumption a pleasure not a chore.

I'm convinced that this type of reporting will be at the heart of what we deliver for years to come. In fact, we're creating a network of partnerships to deliver integrated segmentation and reporting across a wide-range of enterprise tools. Our goal is to support access to system-based reporting built on two-tiered segmentation regardless of the underlying technology stack.

We've already instantiated these reports in Omniture with V15 or Discover and ORB (and I'm going to be doing a whitepaper and webinar on that) and we're delivering segmentation and automated reporting projects right now for multiple clients in our traditional toolset. We've created a partnership with iJento to deliver even more advanced two-tiered segmentation into SQL-Server where it can be accessed seamlessly with Excel or Tableau (and we've created this type of reporting in both). We've just inked a partnership with Infobright (from my recent webinar) and Pentaho, and you can see the fruits of that labor in a full-fledged demo (rolling out the week after the 4th of July) that joins the power of data-driven segmentation and visualization with a blazing fast columnar database and a really rich reporting and data exploration tool. This is awesome stuff - true Web intelligence on 30MM+ rows done in real-time.

I'd like to make this capability ubiquitous on every platform from SiteCatalyst to Hadoop, and while we still have a ways to go to get there, I'm fairly confident that the paradigm will hold. I think these reports answer more than the question "Is my Website successful?" - they answer the question "Can real segmentation be baked into powerful enterprise reporting appropriate to non-analysts?" I believe the answer is, resoundingly, yes.