How To Overhaul Your Analytics To Get True & Accurate Data Again

At every in-house SEO job I've ever had, I've reached a point where it becomes clear that we need to do an analytics overhaul.

Over time, the quality and reliability of any analytics program is going to degrade. Some common reasons this happens:

• The people who originally set it up have since moved on, and their naming conventions don't make sense to anyone else.
• The goals that have been set up no longer fully reflect the websites objectives.
• The data doesn't match up well enough between different tools.
• Tracking has been improperly implemented, or not implemented at all, on certain page types.

Whatever the reason, people aren't confident that the data is accurate. Rather than fix the problem, they've been soldiering on, doing the best they can despite their untrustworthy or incomplete data. It's time to make the data better.

Step 1: Talk With The Data Users

Meet one-on-one with individual stakeholders or with individual teams who consume analytics data. Talk about where the data cant be trusted, and why.

Find out what accurate data would look like.

It's very rare that data from a web analytics tool like Google Analytics or Omniture SiteCatalyst will perfectly match up with a CRM or accounting tool. Find out what an acceptable margin of error is.

These heart-to-hearts are a great way to gain additional context into how the data became unreliable, so listen closely.

Step 2: Ask What Data Points They Need

Now that you know the ways the data cant be trusted, it's time to discover the ways the data isn't serving the teams current needs.

Meet with any team that might possibly consume analytics data. That might be the Marketing team, the Executive team, the Help team, the Sales team, the Dev team, the IT team - whoever they are, if they have key metrics that are measured by the analytics tool, you should talk to them.

Find out: how do they know they're being successful? Obviously for most companies, the bottom line of sell things and make money is going to be pretty important, but what else is important?

Avinash Kaushik has an amazing blog post on defining business objectives and metrics that I recommend  your teams read and discuss prior to the meeting. From these conversations, you'll be able to pull together a full list of all the data points each team needs in order to have a full picture of their success or failure. I'm not saying you'll be able to give them all everything they want, but at least you'll know what they want.

Step 3: What Can Be Delivered Where, How

It's time to compare the current state of affairs with the ideal one. Wherever a data point is wanted but not available, figure out what it would take to get it (try, whenever possible, to reduce what it will take to a series of specific tasks or actions). Do the same with the data that's currently untrustworthy. You'll wind up with three main groups of tasks:

1. Things that can be accomplished in the analytics tool, like new goals and segments
2. Things that require dev work to implement, like on-click tracking and code updates
3. Things that cant easily be fixed and will need either a larger dev project or another tool to adequately measure

Step 4: Making Magic Happen

Go ahead and table that third list for now. What you're left with is two lists of specific, achievable tasks.

Now you have a Plan.

The tasks that can be performed in the tool are good to go; you can do those. For the tasks that require coding, sit down with the dev team and get them scheduled out based on the dev's existing workload.

Step 5: Show & Tell

Over the course of this exercise you've spent a lot of time with the people who consume your analytics data. Don't abandon them now. Spend some time training them how to use the tool. Help them get their custom segments and dashboards set up so that they can easily and happily use the data you've so painstakingly crafted for them (for more on actionable dashboarding, see my post This is How You Easily Setup Actionable Google Analytics Dashboards ). The time you spend training them now will pay off down the line, when everyone is able to pull their own data without having to bug you for it.

So there you go. In just 5 steps your data is more accurate, trustworthy and useful to internal teams. Remember this process: you will have to repeat it.

If you liked this post, you might also enjoy How To Setup Query Attribution Modelling In Google Analytics [PPC]