Whether you’re in SEM, SEO, paid social, display, email marketing, etc., not all metrics are created equal, and bad data really is a thing. Being able to make sense of the noise is where you’ll provide the most value to your team and your client.
Aggregate data: good, to a point
Data in aggregate is really useful when you want to evaluate directionality: in what direction is traffic trending, how is tablet/mobile growth related to overall traffic growth? Where is revenue trending? ALL are important questions to answer, but they’re not really hard questions to answer. See the charts below for examples of aggregate data:
Aggregate data, however, also masks the specific trends driving performance, leading to (at best) ineffective or (at worst) counter-productive analysis. Few trends apply to all users/pages/ads, so you’ll need segmentation to narrow your focus to find leverage/actionable items.
Segmentation: useful, but avoid excess
Through segmentation, you can dig into your data set to find insights and actionable items.
Below we have the same data set as above, but the dimension of time is added. Now we can better see what our poorer-performing months are. While still not ideal, from an executive standpoint, we can see that we weren’t underperforming all year, as the other graphs may have suggested.
By simply adding a dimension (in this case, months), we can break out the data and see when performance might have dropped and if there are outliers in our data set (in this case, $/Visit is skewed in March of last year).
There are times when the solution to a problem simply introduces more problems.
Can you tell what’s going on these two graphs in 10 seconds or less?
Segmentation is a tricky game. For every metric you segment, keep in mind that it will multiply your grand total of metrics by the number of times you segmented your data set.
Remember, representing data in a meaningful way is how you can find your sweet spot. For example, the graphs above have so much information that it’s hard for us to digest. There are things about the way the data is set up that can improve visualization, but the truth of the matter is there’s simply too much data. Any data monkey can pull numbers and create basic bar graphs, line graphs, etc. Data scientists, on the other hand, don’t just look at numbers; they contextualize, they visualize, and they make decisions based on information presented in front of them.
For ease of reporting, sometimes we combine metrics that needn’t be combined. By separating metrics that aren’t necessary related to each other, we can create simpler representations that speak to our audiences better.
If you want to add some flair and contextualize these graphs, you can.
-For organic traffic, we sometimes compare total number of rankings to traffic or compare against conversions.
-For paid traffic, adding spend would be a great way to show the relationship between visits to spend (which is different than clicks to spend); adding conversions would also help contextualize the graph.
Adding conversions can not only tell you how well each month performed, but speak to the quality of the traffic that month.
So…how do we separate insightful data from available data?
-Data is only valuable when you have the right combination of dimensions and metrics.
-That data becomes actionable once it is segmented and graphed correctly.
Questions to ask yourself:
–Does this data teach me anything?
–Can I do something useful with this new information?
–Will that impact performance? If so, how?
Who is the data for?
Probably one of the most important bits when talking about data pulls is curating that information based on audience.
-Your reports should speak to their intended audience and should be available as standalone documents. If you have to walk your client through it every month, it’s possible that your report is too confusing.
-Sometimes, when we report on data that is too granular or data that is too general, we call attention to things we really shouldn’t.
-Finding the perfect balance of information and reporting is key when presenting data—especially when that data impacts the bottom line.
Finally, never assume that your audience will understand every metric you present. Assumptions are always a grey area you want to avoid. If you ever look at a report and can’t explain what you’re looking at in 2 sentences or less, it may be time to investigate why you’re reporting all of that information.