My predictions and hopes for digital analytics in 2013

Well, it’s that time of year again! But this time, we’re rocking with DIGITAL analytics, not just web. A lot of new stuff, and a lot of good, solid fundamentals applied to new things to be conquered in 2013. I hope 2012 was a good year for you (no doubt!), but I think 2013 is going to be a HELL of a lot better.

First, let’s recap the 2012 predictions from this post:

1. Analysts need to stop talking about savings. And boy did we ever. I knew you badasses were up to the task. What we do is not about shrinking businesses or not participating. It’s about growing them. Yes, it’s a perfectly valid idea that in order to reallocate budget, you have to trim waste, but in my mind, that’s not a strong sell. I think that there’s something [sadly, yet] deeply satisfying about looking at something hugely inefficient, calling dumbasses out on their poor work and waste, and recouping the losses. But not nearly as rewarding as turning losses into gains over and over again. This is a cycle that will not stop. You can’t / won’t clean house just once.

It’s awesome to see conversations about superior investment, multi-channel, and…ick…big data blossoming, all in the context of efficiency, shared learning, more complete understanding, networked thinking and upside, not just where we should trim the hedges. Lovely stuff indeed.

2. Get businesses to stop chopping our jobs down to bite-sized questions. When I wrote this last year, I was seeing a lot of businesses only asking for what they knew was available in reports, rather than sharing their true needs and challenges. Again, I have to say HOLY SHNIKEYS our industry murdered this one.

I’m now hearing so many more businesses coming to us, our competitors, in-house analysts and even evil vendors with real business questions. Analytics tools are now being bought on the basis of being able to pick apart complex, segmented, multi-session/channel behavior, not on the basis of pre-baked reports. Tag management is being widely adopted to solve efficiency and process problems that plague operations. Testing platforms like Optimizely are becoming more WYSIWYG than JS. We are catapulting into an era where the tools are becoming easier to use in the course of business, where stakeholders can take more control, and where analysts are less technical in nature and more business-oriented.

Analysts today are asked the big questions, and when you are, you deliver. Awesome to see.

So, I’m pretty much pumped. 2012 was a huge year in the understood and captured value of our industry. So, what’s up now?

Well, there are a lot of buzzwords flying around these days, all about big data, R (I guess this is technically more of a “buzzletter”), data science, multi-channel, attribution, and more. And for the most part, people are really thinking these problems through. Businesses seem to be cooling their jets on finding the next magic bullet and they’re getting serious about putting significant intellect and funding behind tackling big, complex problems. This is awesome for us, and will introduce a lot of choice into how career paths look moving forward. Years ago, you could either be a marketer-type or a tech-type. Now, there are so many things it would make your head spin.

So, how does that play into predictions? Well, that’s tough. Now that we’re out of our infancy and even adolescence, the market has fragmented. Not long ago, web analytics happening at a small, savvy business wasn’t all that different looking than web analytics happening at a Fortune 500 (you can fight me on that one, but I have to warn you of my stubbornness). Today, they are radically different. Enterprise analytics is separating from the “pack,” if there even is a pack, and that separation seems to be widening by the second. Enterprise is committing to huge investments in both efficiency and sophistication, and the multi-channel strategies in place are simply something most smaller businesses can’t afford.

But I do think that there are some shared ideas or enterprise themes that other businesses can borrow. So I’ll do my best to make this make sense. If it doesn’t, hit me with a rubber chicken next time you see me.

2013 in Digital Analytics

Prediction 1: You can’t buy maturity any more

Over the course of the last 5 years, there have been a number of analytics “maturity models” used to score businesses. These have been awesome, as they typically highlight strengths and weaknesses, and help a business identify where they need to put their attention and their money. Historically, though, these models have put a somewhat heavy emphasis on both the sophistication of tools and the types of people your organization has hired. Also to note, the models have looked at your KPIs, goals, etc.

The problem with these models of maturity is they measure potential rather than productivity. Newer models are focused on the altitude, speed, and direction of your shuttle, rather than the quality of the launch pad. You can buy your way into potential. You have to fly your way to actuals.

I would strongly recommend that every large organization assess their maturity in a qualitative and quantitative manner. We do this with every single one of our clients, whether that client uses our consulting or if they use Satellite, as both product and consulting have enormous implications on the way your business will work.

When you assess your organization’s maturity, you need to look at actuals. Don’t ask, “Do I have a testing tool with segmentation capabilities?” Instead, ask, “How many segment-driven tests have yielded conclusive findings that were adopted widely in the organization, in the last quarter?” Also, you need to identify whether the maturity of your organization occurs in rogue pockets or if it’s widely accepted and embedded into the culture, goals, process, and executive mindset. I can’t tell you how many great analytics, testing or personalization operations or people I’ve found inside of an organization that deploys their web site twice a year and doesn’t know what these peoples’ names are. Rocket fuel inside a Pontiac Aztec.

Prediction 2: Widespread use of consultants

Building on #1, I see people really getting help this year. Not just tactical help with implementations and not just strategic help with seminars or executive retreats. Real soup to nuts help. Just look at what Eric Peterson is doing at Web Analytics Demystified: that organization is growing into something that won’t just teach you what swimming is, but will let you swim on their backs, have them swim next to you, swim for you, or whatever it takes to get you there. Results. Actuals.

The truth is this: there are a lot more open slots for great analytics leaders than there are great leaders or analysts. Every company wants to hire in-house talent to absolutely rock the house with analytics. And while there are tons of total badasses out there, there just aren’t enough.

So, here’s an idea: get a absolute killer rock star consultant FIRST. Let them tell you who to hire. Let them manage the transition. But don’t even start if you aren’t ready to have rockets strapped to your ass and lit. You’re wasting everyone’s time if you aren’t ready to roll. Analytics consulting is not a daycare. It is more like crossfit in space at 10,000 mph with pods that will automatically shoot morons and progress preventers into deep space.

Prediction #3: Small wins will give big data big momentum

You like to say, “big data,” huh? Think that makes you tough?! Well, in 2013, it will.

Here’s what I think will happen. Big data is currently the closest thing we have to a “magic bullet” obsession in our industry. Hopefully we’ll get over that pretty quickly so we can get to work on it. I think the first step in tackling this mountain is to break it down into tiny little wins, and each of them will be awesome. But let’s set expectations ahead of time: just because it’ll be awesome does NOT mean each win will magically rain money down on our heads. So get over that and you’ll be ready to take your first steps.

The first steps will be in simple cross-channel and tangential data connections. There are some incredibly cool projects happening around the country right now when it comes to this. Here are the questions I’m excited to start answering:

  • How does our conversion vary by happy vs. upset customers (social)?
  • How does our conversion vary by geography by weather and other external events?
  • What are correlations to capital markets, pricing indices, legislation, or news?

Now, for many businesses, these types of things may not be relevant. But at huge enterprise scale, I think that all of the modeling and R work being done to assess forces within our small ecosystem are probably brewing a fresh, giant pot of false positives and negatives. It’s time for us to think more like the BI side on this front, looking at external market forces equally, if not more. Wal Mart stocks strawberry Pop Tarts on the basis of weather patterns and has for decades. Our jobs are thousands of times easier when dealing with complex or incongruous data marriage than Wal Mart’s was when they explored this for the first time, so let’s take advantage of that. But, of course, that’s not a small win, necessarily, so I digress. Just saying the precedence and success stories are there…

Large brands are currently in the process of making big data small by breaking it into pieces. I’m hearing great stories about user-level (anonymous, of course) understanding of sentiment vs. site behavior. This is taking a really potentially lame social media metric and marrying it with really not-lame business metrics on our web sites. I’m also hearing about cool basket analyses, multi-session trends (for predictive suggestions), and some smart attribution…

Prediction #4: Attribution will be appropriately…uh…attributed

First click. Last click. Weighted. Time decay. Brain decay…

The problem with our current approach to attribution isn’t just related to the sequence of the click, but also the relevance of the conversion type. Conversion type? What the hell are you talking about, Evan?

I’m talking about the fact that today, the vast majority of marketing channel efforts are measured against one thing: conversion. The way we then tie credit back to earlier interactions is through attribution. That keyword, email, or display creative “opened” this relationship, while another one “closed” that customer. Poppycock.

Here’s where we start to mend this fence, though: marketing needs to trust analytics to handle this, and the business needs to trust marketing to “soften” some of their metrics. Let me explain…

While there may be a “conversion” on your web site, more often than not, that conversion is a business use case, not a user’s use case. The majority of your users are there to research, compare, consider, hunt, etc. We will start to solve attribution when we credit appropriate to the use case for the user, not just the business.

One cool thing we’re doing with Satellite is assigning meta to your different site actions and audiences. Certain interactions like image gallery views, video views, product list filtering or ordering, site search, etc. would be considered “high funnel” interactions, while things like newsletter subscriptions, whitepaper downloads, or sales would be considered “low funnel.” Now, you can have as gradiated a funnel as you like, but simple high/low is a good start. Now, you take your search campaigns, emails, tweets, wall posts, display creative, etc. and you figure out what each of those is supposed to do. If a user searches for reviews, ratings, product pictures, or comparisons, don’t measure success to financial conversion. Measure it instead to those high-funnel interactions that make sense for that user’s use case. For a 1-day sale creative, go ahead and measure to monetary conversion.

What you’ll start to see is that when breaking your media and your conversions down and pairing them appropriately, your effective rate of converting users based on their use case will be extraordinarily high, restoring faith in these “high funnel” media investments. You’ll then also be able to model out consumer lifecycle: users who search high-funnel and have one or more high-funnel conversions in that session are x% more likely to return and purchase, and the “middle-50″ value of that purchase will be between y and z dollars. That’s a lot more transparency than some simple, “This was the third click out of 7 so we attribute dollars based on the assumption that all mammals are dogs, which has the same empirical value as current attribution models.”

Predictions 5+

I had a few more of these ready, but this post is already ridiculously long. I’ll turn it over to you, instead. What are the big themes you think we’re poised to tackle this year? The fan is widening, so it’s getting tougher to write this type of a post! And that’s a sign of the success you’ve created in this industry. Let’s make 2013 epic.

No Trackbacks


  1. While I agree with your premise of trying to tie your marketing efforts to the visitor’s use case, should we really be calling this “attribution” as the word is commonly known in the industry?

    It’s bad enough that there are arguments about Markov Chains and Bayesian statistics when talking about marketing attribution (and these arguments aren’t leading to answers, but more questions), but to expand the definition further to micro-conversion seems like a way to introduce more paralysis into analytics, rather than enlightenment. Then, add in the “the data is wrong” naysayers and we’re left with the last-click model that everyone knows is wrong but no one is willing to try something else.

    For me, I’m more inclined to keep use case conversion away from marketing channel success. Without having data to prove the hypothesis, I don’t believe we add anything substantive to the discussion by trying to infer multiple marketing touchpoints affect micro-conversions occurring in today’s visit. Is it not the job of the landing page from organic search to serve the user looking for “cute cat videos”, rather than whether I got an email 3 weeks ago that talked about dogs (and didn’t get a click-through)? I’d say yes, which puts the onus towards on-site optimization rather than on the chain of marketing collateral you’ve seen over time.

    Perhaps what I’m arguing here is that last-click is the appropriate model for micro-conversions, and that the “attribution” is the series of micro-conversions is the path we’re trying to attribute the conversion to? I don’t know. In any case, I don’t think this is “marketing attribution” from a semantics standpoint, unless we’re trying to completely FUBAR all success we’ve had getting CEO’s to actually believe any of us know what we’re doing.

    Posted January 15, 2013 at 1:49 pm | Permalink
  2. Hmm, you make a really good point about getting stuck and never making a decision we can execute on. I’d say we need to have an execution track but also a maturation track. We need to stop executing on crappy frameworks asap, but sometimes crappy frameworks are better than no frameworks. For me, it’s not a matter of whether models are wrong. It’s a matter of realizing that what we’re modeling is wrong.

    I definitely feel that evolving something does not risk our reputation or FsUBAR (since FUBARs would be grammatically incorrect) our success. That mindset is possibly a huge problem. No CEO on attribution 1.0 wants to be left behind when there is an attribution 3.0, not to mention the fact that no CEO has a clue what this is in the first place. They are not in those weeds.

    Maybe if we don’t pair planes that are in-flight with new plane design so tightly, it makes more sense. No, you should not redesign a plane that is in midair. That is fairly not good. But having learned from planes that are in air, let’s design new, better planes, then bring them into rotation, eventually phasing the old planes out.

    Posted January 20, 2013 at 9:10 pm | Permalink
  3. Good post, Evan. I thought I would never finish it :) … and then came those planes.. Always learning here.

    Quick thoughts:

    1. I would say that efforts on data analysis and reporting are finally proportional to an entity’s data collection capabilities. So yes, the gap is widening between enterprise analytics and DIY hobby analytics.

    2. Good points on maturity. I guess it all boils down to internal communications and corporate culture. Just look at online travel or direct insurers: they are far ahead in terms of having a process in place for data-driven optimization.

    3. Big Data: Perhaps the big news here is Web Analysts seem particularly well positioned to go and grab other chunks of non-digital data, as we are already sitting in the right place (Marketing) and used to unstructured, real-time data.

    4. Attribution: For a second I was teletransported to the Webtrends Score years reading you here (terrible…the outer space jargon is getting in my head -evil branding move on your part ;).

    Anyhow, it makes sense and I would take it even further with complete disregard for conversion-bound attribution: How in the world are we expected to apply a truly cross-channel model when there are channels that simply do not speak the same language? (ie we cannot tie their cost/impact to the one cookie ruling our business). And the answer may perfectly be somewhere else in the funnel.


    Posted January 22, 2013 at 5:35 pm | Permalink