As part of the upcoming Digital Analytics Hub Conference in Berlin (1st week of June), I did an interview on the sessions I’ll be doing there and thoughts on some of the broader directions in digital analytics including a surprising amount of discussion around analytics methodology. Analytics method has never gotten the attention it deserves in digital and with the growth in big data systems, the absence of robust analytic methods to analyze digital data has become an issue of paramount importance.
What are the key challenges your clients are currently facing? And how have these challenges evolved over the past year or so?
The digital analytics market has matured a lot in the last couple of years. I don’t see as many clients struggling with infrastructure or basic reporting any more. That’s a good thing. I think the biggest challenges are:
- Transitioning from reporting focused organizations to analytic ones
- Integrating into a broader, customer-focused analytics world
- Handling the influx of technology as organizations go from SaaS to more native analytics applications whether internal or cloud-based
- Driving real-world impact with analytics via personalization and cycled testing
You are running two workshops on the Monday. Can you tell us more about them [Analytics Toolkit and Customer Analytics]?
The Analytics Toolkit session is one I put together a while back that integrates six or seven different workshops on analytic method into a single session. I don’t think our community spends nearly enough time thinking or talking about analytics method. Too often, the presumption is that we sit in front of data and insights emerge through some magical journey of exploration and intuition. I don’t see it that way. The toolkit session walks through Functionalism, Use-Cases, Topology Analytics, Funnel Analysis, Digital Segmentation, and a couple more techniques for studying digital behaviour. It’s one of my favourite sessions. The Customer Analytics session actually has some overlap (buyer beware) but focused on customer journey and experience mapping. Functionalism and Use-Case analysis both get a treatment here but I also spend a lot of time diving down into Voice of Customer analytics and how it integrates into behavioural analysis.
Voice of Customer research is a recurring theme in both your workshops and discussions. Why should digital analysts be passionate about VoC?
In some ways I’ve come full-circle on this because I didn’t always think it was that necessary. Many years ago I did quite a bit of survey research in the political realm, but I gradually became an advocated for behavioural techniques. When survey research first got popular on the Web I was sceptical, but as we actually got our hands on it and were able to use it, I changed my views. Online survey research is incredibly cost-effective and if done well (which it almost never is) can answer questions about customer choice and decision-making that simply aren’t retrievable from behavioural data. What’s more, I’ve come to believe that because most organizations do VoC so poorly, it’s the biggest and most easily capitalized opportunity in digital analytics.
In the description of your “Statistical Analysis of Journey Data” you mention pattern-matching techniques, momentum analysis and simulation methods. Most digital analysts would have never heard of these terms. Can you briefly describe them? How should digital analyst go about developing skills in these techniques in your opinion?
Digital data turns out to be rather more challenging that many of the data sets I’ve worked with in the past. I describe it as a paradigm case of big data – not because there’s so much of it (though there is), but because useful analysis of digital data requires our methods to account for structure, sequence, time-between events, and the pattern of events. None of those are easily accomplished in either Digital Analytics tools OR traditional statistical analysis tools. Pattern-matching techniques are essential in digital because paths are so varied. To cull meaning from them, we need to understand pattern not path. Pattern is also critical when you move up one level and try to understand how a digital touch fits into a larger customer journey. Momentum analytics is something I was deeply involved with back in the early ‘90s when I was writing analytic software for commodity traders. There are a set of analytics techniques that help you understand how something is trending (in commodity analysis it’s usually prices but in digital it’s more likely to be things like traffic or conversion or error rates) and when trends are significant. These techniques are often paired with visualizations that make it much easier to understand the data and which I think are quite useful in digital. The application of simulation (building a virtual, runnable model of a system) is a technique I’ve long been fascinated with. It’s challenging, but it allows an analyst to combine multiple analytic methods into a single system and then test that system against real-world historical data. It’s a way under-utilized technique (including in our practice) and I think it’s quite applicable to a certain class of digital problem – particularly digital marketing systems.
Of the other discussions at the DA Hub which are you hoping to attend?
Truth is, I can’t remember what I requested – and it’s so hard to choose. Reading the sessions is like walking into an Apple Store. I want everything. Looking at the list again, here’s a few I hope I signed up for: ‘Listening Better’ w. Alex Emberey, ‘Web Analytics Personas’ w. Xavier Colomes, ‘The Wretched Executive Dashboard’ w. Kyle Keller, ‘Do you want purple fries with that’ w. Tom Betts (yes!), and ‘Cohort Analysis’ w. Tim-Fabien Pohlmann which happens to compete with one of my Huddles so I know I won’t be attending but I’d sure like to.
Readers of my blog can probably see why all of these topics might fascinate me. Social Media Listening is not only a new frontier in customer analytics, it’s a place where techniques and technologies are just beginning to gell into an effective whole. For Personas I read segementation and there’s surely no topic I’ve thought or written about more often than digital segmentation. Of course, a close #2 on that list of topics would probably include the miserable state of our reporting and how to fix it – a topic that I’ve gone through multiple evolutions on and don’t expect to ever stop trying to perfect. “Purple fries” speaks, of course, to Personalization which is my current series topic and which I believe is the ultimate focus of digital analytics. And, finally, cohort analysis is one of the under-utilized but critical techniques in digital analytics. It’s a great topic and speaks to the focus on methodology I’ve been talking about. Cohort analysis is critical, in particular, to community analytics but it has broad applications to digital.
Finally, a question of particular personal interest – as opposed to statistical and other traditional analytical disciplines I sense a general lack of consistency and rigour in digital analytics. Do you feel that our industry should develop common and agreed methodologies or is the set of problems we look to resolve is too diverse and evolving to be confined by one strict methodology?
Well, if I’d known all the questions first I might have saved my answer to the question about my Workshops for here. Yes – I obviously agree. I don’t think there’s going to be one digital analytics method. There may not even be a single method per problem. But I think it’s clear that if you’re going to tackle problems like attribution, optimal pathing, content association, or funnel optimization (just as examples), there are going to be specific methods that have proven to work better for each of these. If you don’t know the methods, chances are you’re not going to do a good analysis. Maturity will come when we’ve learned proven and appropriate methods for many of the key digital analytics problems. We sure aren’t there yet, but I think a big part of the reason for that is that lots of people in the industry don’t even recognize the need. The analytics as “magic” worldview is still surprisingly prevalent.
If you’re based in the EU, check out the Hub. It should be a great experience and a chance to delve deeply into questions of methodology (and more) that are at the heart of good analytics practice. See you there!
And speaking of conferences, I’m going to be out at the DAA Symposium in Austin in a couple of weeks and I’ll be leading a session on Personalization. The sessions are meant to be collaborative discussions with a minimum of presentation – somewhat X Change like – so it should be interesting. If you’re in Texas and can make it, it should be very worthwhile. Kelly Wortham is leading a breakout on testing as well – so you have your choice of Personalization or Testing. The hardest choices are good ones.
