Today I am going to talk about something that plagues companies and consultants everywhere--half baked analysis. It's something we've all done at some point, and something a lot of us still do on a regular basis. It's unfortunate because as online marketers we all understand the power of good data mining, but time and time again we revert to generic inquiry, at best, and default report templates.

Disclaimer: Origionally I attempted to write about the five steps I follow for solid data analysis in one post, but as I approached my 6th page of content, I realized it may be best to break up into a series.

Alas, this will be the first of three posts, tackling a five-step process toward good data analysis. The three topics are:

1. Asking the Right Questions
2. Identifying What is Going Wrong
3. Turning Data Into Action

Yup that's right...cancel that afternoon meeting because you my friend are going to be stoked about data analysis in 3...2..1...

Rethinking the Questions

A few weeks ago at our SEOmoz PRO Seminar I spoke on "Analyzing What Matters & Ignoring the Rest" and I challenged the attendees to rethink the questions that guide their data research. Too often we get caught up in asking questions that simply put-- don't really matter. Let me explain. It will always be important to know things like "How much has traffic increased" and "What referrers are performing better this month," but this sort of inquiry does not qualify as marketing analysis.

Sure it's valuable to report that to your clients or boss, but as an analyst you are tasked with much more. You are tasked with finding things others can't. You are expected to dive into the data head first and find issues before they become huge problems. You are also responsible for finding opportunities a.k.a. the "game changer" for your company...that is your job. If you don't like the way that sounds, please stop calling yourself an analyst. You are stressing me out.

So what questions should you be asking? Bigger ones to start.

I know they sound uber-top level, but don't roll your eyes just yet. I challenge each of you to write these out and really think about the answers. I think you'll be surprised with what you come (or can't come) up with.  I'm going to apply this to SEOmoz as an example.

An outsider would look at our site and say we are -

1. Trying to sell PRO memberships
2. An increase or decrease in completed goals would show us if we are being successful
3. Losing traffic to our sign-up page, and a lower traffic count would be detrimental to our success

Well that is great, but honestly SEOmoz can't succeed solely on increasing PRO memberships. The truth is, there is a lot more to it than that. We have a recognized brand with expectations on it, and a community of over 200,000 people that come to us for the latest SEO information on the web. We can't afford to lose ground on either of those two. These are defining qualities of SEOmoz, and strong advantages over our competitors. So my three questions would leave me more complex answers, something like this:

1. Increase organic traffic on "Learn SEO" type queries, increase branded term searches, increase YOUmoz member engagement, and increase signups
2. More referrals from links to our resources, more traffic from people researching SEO, more YOUmoz submissions, more comments, improved engagement metrics on site, higher sign up attempts, higher signup completions, etc.
3. Decline in branded term searches, decline in organic traffic to resource pages, decline in time on site for YOUmoz members, etc.

So now what? You are left with a handful of metrics to investigate. Those metrics should be the base of your analysis efforts. I urge all of you to revisit the reasons why you analyze what you analyze, you'll be surprised to learn that you don't really have a good reason most of the time. After you have your new questions nailed down and you know what metrics you want to analyze,  it's time to jump in the data.

Start Macro and Go Micro

This is when I highly suggest you fill your coffee cup, or grab another Red Bull. I also support locking your office door, or putting up a "Do Not Disturb, I am Data Mining You Silly Non-Analyst" sign up on your cubicle. Okay anyway...so the main roadmap to solid analysis includes five steps and they are:

*Please note that Analyze, Value, and Action will be covered in upcoming posts in this series.

What Do We Mean by Macro Analysis?

Macro analysis means you have a solid understanding of the different sections of your site, the different user types that navigate it, and the top-level metrics. You should know these like the back of your hand. In addition to knowing these actual numbers you should know their rate of change (how often does that data point change), the depth of change (how extreme are those changes--big jumps? small steps?), and the way they interact (is there a consistent relationship between two metrics--one goes up/down, the other will too). If this sounds like a lot to continuously track, you are right. Good analysis is a lot of work. Thankfully SEOmoz pays me in cupcakes, and Champagne Wednesdays, I highly suggest negotiating for these perks ;)

At SEOmoz we track our top sections by week, so we can easily identify shifts in the data, and it looks something like this:

(A portion of our weekly analysis for full site stats)

You can see we aren't just looking at our homepage, we are looking at our subdomains, our highest trafficked sections. We also are going beyond visitors, we are pulling top-level stats like pages/visit, time on site, bounce rates, etc. This graph goes around to the entire company once a week. This macro level view helps all of us understand the momentum of our site's growth. It helps us easily isolate problem areas so we can address them before they grow into huge "Oh sh*t" moments. Trust me when I say, if you aren't tracking your data at this macro level, you should start today.

What Do We Mean by Micro Analysis?
This part of the puzzle is the one that most people skip over. Micro analysis means you don't just have a sense how your blog's traffic is doing you know how many comments you get on it, how long they spend on it, how deep they go into your site after reading a post, and how many of your blog visitors end up converting for you. In short, micro analysis means you look at all those secondary data points that you can actually manipulate.

While it's great to go into work on a Monday and say I want to increase traffic to my blog by 20%, it is a big feat to accomplish. Not only will it take a lot of time conceptualizing, writing and sharing that content, it will also, most likely, be less lucrative than if you took the existing traffic and increased its conversion rate by 5%. That sort of move is done by honing in on data at a micro analysis level.

Specifically this is where things like event tracking in Google Analytics and deeper dives into your preferred analytics package come in handy. Everyone has their own approach for micro analysis, but I think a good place to start is see where successful events (downloads, subscriptions, sign-ups, conversions, etc.) are taking place and see if you can come up with common demoninators. If you see that successful pages all have one or more thing in common, you can start testing these on other sections to increase conversions across your whole site. Here is an example of what we pull for SEOmoz:

(A portion of our micro tool usage analysis report)

We can see which tools are performing the best, and analyze those pages to see if we can isolate out page tweaks to roll out across all tool pages. It seems simple, but way too often analysts look into analytics to see how they are doing, and fail to put in the time required to uncover what they could be doing for increased success. You should know, for every single section and user type on your site, what makes it "successful." You need to be tracking these "successes" as closely as you would your visitor count.

Well this post got a little long, but I really wanted to give you guys some real examples on how I approach data analysis both at the macro and micro level. Hopefully, you can take some of this and apply it right away. I know we all have our own unique approach to analysis, and I'd love to hear yours in the comments below!

Next post I will be talking about the "analyze" step of a solid analysis strategy. That post will hone in on quick ways to figure out what is going wrong. I will talk about some GA features that you can use to make your analysis more effective and less time consuming. So stay tuned!