I’m going to be out at eMetrics the week after next and I have a couple of presentations on tap. Both, in a sense, reprise previous efforts. On the 21st, I’ll be presenting on the pitfalls of measurement and the Agency relationship. This is content based heavily on the webinar and work I’ve done with Bob Heyman – it focuses on the perils of letting Agencies measure themselves and how an enterprise can/should centralize measurement of channel performance. On the 19th, I’ll be reprising my San Francisco “panel” with Michael Healy. We’ll be joined this time by Christopher Berry (who I don’t actually know), so it might be a bit more panel-like than the conversation Michael and I enjoyed in SF (two-people no real moderation). I hope not, I really liked that session.
Both are topics that I have, fortunately, been continuously involved in over the intervening period. Our social media practice, in particular, is booming and some of the most interesting work we’ve been doing on our team has been in various facets of social media measurement and its integration into broader Web and Customer analytics efforts. Over the next couple of weeks, I hope to post on some of the most interesting work we’ve been doing in Social measurement and our evolving theory of the space and the corresponding role of measurement. It’s a little different than Web analytics.
Before I go there, I wanted to reflect on last Monday’s WAA Seattle Symposium on the “Future of Analytics.” This was my 2nd visit to the Microsoft Conference and Executive Briefing Center; it’s a beautiful facility and the turnout was pretty amazing. Events like these are drawing audiences almost as large as eMetrics, though, of course, they are heavily driven by highly-committed parent companies. I’ve always found Microsoft folks to be pretty cool, so it’s nice to hang out there a bit.
With a topic like “the future of analytics” you can imagine that the conversation was a bit varied. But there were a couple of interesting themes that emerged. First, I think it’s worth noting that the discussion started by explicitly dropping the “Web” from the front of analytics. That’s not exactly what you’d expect at a WAA event, and, to be honest, as a consulting firm that describes itself as a full-service Web Analytics Consultancy, we’re in a similar boat. A lot of the work we’re doing these days isn’t classic Web analytics. On the other hand, it isn’t just analytics either. There is a whole universe of analytics that we never touch and never hope to: things like Financial Analysis, Fraud Detection, Manufacturing, and Risk Management are all robust analytic disciplines. In past lives, I’ve even had exposure to those things. But at Semphonic, we don’t touch any of them. So what’s the boundary?
At the Symposium, I described what we do at Semphonic as Customer Analytics with an online focus. I think that’s already more accurate than Web analytics, and the pendulum is swinging sharply in that direction. Nor do I think we’re driving that boat; we’re just rowing with the tide. My whole blog series this year on the convergence of database marketing and Web analytics is a testament to that tide as, surely, is the growing amount of work we’re doing with Customer Research/VoC and Social Media analysis. All three converge within a single discipline that could fairly be described as Customer Analytics.
To me, the two most important and exciting developments in analytics (the second panel question) are both related to this trend. The first, a theme I remarked on as the prevailing theme at this year’s X Change, is the actual availability of Customer-Level data for Customer Analytics.
You can’t do Customer Analytics for real without Customer-Level data. This isn’t about BI vs. Web analytics. Lots of BI systems (including all traditional OLAP analysis) don’t let you analyze customer-level data. The behavioral analysis methods used for customer-analytics simply require a different granularity of data. Nor is it about Web Analytics tools vs. traditional data warehousing. Tools like Webtrends’ VDM deliver Customer Analytics and some traditional warehousing solutions don’t. What we’re seeing, however, is a tremendous push toward solutions that deliver the real goods for Customer Analytics.
I’ve notice that tool vendors tend, almost, to be too self-effacing about these sorts of things; almost too willing to re-iterate the view that it’s people that matter. Of course. But tools are the enabler (or preventers) of everything we do. Not even Michael Phelps could swim in a suit-of-armor. The rapid availability of systems with the horsepower, the data models, and the tools for studying and acting on visitor-level data is the single biggest change in the Web analytics landscape in the last half-decade. It’s a change every bit as big and rather more important than the move from log-files to SaaS/tagging that triggered the last revolution in our industry.
My other “exciting development” is quite different though equally related to the theme of Customer Analytics. I’ve come to believe in the last year (and this was one of my Think Tank Classes at X Change), that the single biggest information opportunity in analytics is not behavioral. It’s the centralization, classification, and standardization of the various customer-focused attitudinal, VoC and customer feedback systems that exist in today’s enterprise. For decades now, we’ve all understood that the centralization of structured data in the warehouse provides signal benefits to the organization. It allows for greatly streamlined processing, much higher levels of data quality, and for the use and dissemination of a single view of the data throughout the organization. All of these are critical benefits.
Sadly, non-structured data collected by the organization has never received similar treatment (for some good reasons). However, driven by Social Media listening and the necessity for handling, classifying and reporting on customer verbatims, I think there’s a growing awareness that the tools necessary to do this work across every text and voice collection point are becoming available.
I’m a big fan of this slide from Clarabridge (one of the leading vendors in the NLP analysis space) that shows the problem:
And I’m a big fan of their solution, which is one of the best for solving the problem. Centralizing Call-Center, Social Media, Opinion Research, and Customer Feedback data is about far more than Enterprise Feedback Management. It’s about creating a centralized classification scheme within which ALL customer feedback can be segmented and analyzed in a consistent fashion.
I think the fruits of this in driving rich and consistent enterprise awareness of customer attitudes and concerns are enormous. For any organization committed to embracing a customer-driven strategy, this ought to be the single highest business intelligence priority on their calendar.
Interestingly, these two themes tied together in response to the third big panel question – one that drew a fascinating range of response; namely, what happened to Netflix – a company that has a great reputation in analytics – but seemed to botch a huge customer analytics issue?
I thought every answer given in response to this question was fascinating – including the idea that from a broader perspective Netflix may have nailed the basic customer dynamics and value proposition here, and that they might be better off shedding a small percentage of unprofitable customers.
Perhaps. But I do think that no matter how cogent the business analysis, Netflix badly misunderstood the customer attitudes that might tax this decision and they’ve struggling to control the conversation from day 1. If you think about the two big exciting trends I described, you’ll see that while they are both about Customer Analytics they approach it from two fundamentally different directions. Many companies that are quite good at one direction are rather poor at the other. Netflix may be doing a fine job of customer behavioral analysis without really understanding or embracing the more traditional disciplines around customer attitudinal research.
There are also, of course, many ways to misread the data even when you’re paying close attention. It’s important to understand not simply the majority view, but the extent of the minority interest. In politics, I learned that while the majority of voters might favor a candidate’s position, on hot-button issues that majority may care much, much less than a virulent minority. Often times, the only people who will really vote on the issue are in a distinctly minority opinion. Always selecting the majority opinion is nearly guaranteed to be a losing strategy.
It’s critical when assessing customer attitudes to measure not just the overall break-outs, but the depth of each segments opinions and emotions. Customers who will “vote with their feet” may be much less numerous, but around any given issue, they are the ones that matter most.
I mentioned a similar case on the behavioral side, one which greatly shaped my thinking about behavioral analytics. We used to work a great deal at AOL in customer acquisition. During that time, I several times suggested that AOL should make cancellation of their service much easier and even advertise that fact in customer remarketing. In every such meeting, the team there were able to produce studies which proved that AOL would lose money if they changed this tactic of actively discouraging cancellation. True data-driven decision-making. And perhaps it wasn’t even the wrong decision. Given dial-up ISPs future, perhaps milking your customer-base was really the best organizational solution. But the fact is that by constantly optimizing for the short run, they were eating away at the true capital their business carried. When theier aggressive no-cancellation practice finally blew up, it drove a sudden and catastrophic collapse of an already vulnerable business.
These kind of “short-traps” where you optimize to immediate revenue at the expense of customer capital can easily be the result of data-driven decision-making and are one of the reasons why it’s so important to maintain an appropriate balance in your research and analytics efforts.
Finally, I wanted to mention one question we got from the audience that I think touches upon another critical aspect of data-driven decision-making. Essentially, the question was about the importance of real-time. Just how/when is real-time important?
I think of real-time in two ways. First, in classic Web analytics, it’s our business to inform decision-making. That works at the speed and level of organizational decision-making. It’s almost never in real-time. High-volume editorial and certain holiday retail optimizations are the only two instances where real-time decision-making can and does happen. In both those cases, it’s essential that our analytics be capable of real-time. For any kind of strategic decision-making and for the vast majority of tactical decisions, real-time simply doesn’t matter.
However, that’s not the whole story. Informing organizational decision-making is just one piece of analytics. Increasingly, we use analytics to drive customer marketing and messaging. This isn’t about informing decision-making; it’s about making automated decisions. Automated decision-making needs, quite often, to be real-time. Nor is this solely the domain of black-box optimizations. Rule-based real-time decision-making still needs real-time data to deliver good results. So if you’re relying on something like the Omniture data feed to provision your analytics, you’re out of luck.
Real-time processing/decisioning capabilities will be a growing focus point for organizations. If you’re moving to customer-level data, one of the biggest decisions you’ll have to make is whether to try and support real-time. It’s a big difference and it’ll impact every technology decision you make. Right now, there aren’t many systems that can legitimately support real-time (Celebrus comes to mind). In the future, I expect this to become a flash-point; a fundamental dividing line in the way enterprises approach and drive value from analytics.
What’s the danger here? You may think of this as a future capability, but the decisions you’re making right now about data collection, warehousing and analysis will profoundly impact how easy/difficult it is for you to add real-time analytics and automated decision-making to the mix.
And speaking of the future, I have another WAA panel back on the East Coast (Philadelphia) in early November. I’m going to be moderating a panel (something I’ve never done before) on cross-industry learnings in Customer Analytics.
This is a chance, as it were, for me to explore the panel format which, as regular readers know, has never been my favorite. David McBride is hosting the event and he and I share pretty similar feelings about the format. So I have a green light to try something a little different. I’m not going to do it in the traditional fashion (at least I’m going to try not to). My goal is to have less of a round-robin format and more to push on particular answers to force panel members to engage with each other and to get away from pre-canned answers. Panelists beware!
Honestly, I really don’t know if it will work. I believe Panels usually fail because, like political debates, the interests of the participants are opposed to the interests of the conversation. That’s not easy to change or fix. If you’re in the area, I hope you’ll come out and see it. I’m sure I’ll be very curious to hear what people think!
