Let's also acknowledge data are not infallible, utterly objective and a pure distillation of 'the truth' whatever that even means in a hyper-nano economy.
Finally, we have to get over the belief that somewhere there exists something called 'the answer.' And that all we have to do is find it.
If these propositions leads us to a point at which we ask, 'then what am I doing,' we are making progress.
Because there is no goal line, which means there is no one way to get there, which means there is no truth or answer that will survive the next five seconds, let alone five minutes or five years.
The reality is that we are in the realm of the uncertain and imprecise, searching for flickering indicators like a ship in the fog. Eventually, it seems likely that this will all become clearer. But eventually is not now. We can use these data to offer guidance to ourselves and others, but we are learning as we go. The journey, when it comes to web analytics, is quite literally the destination. So let's learn from it. JL
Yaniv Navot reports in Dynamic Yield:
Countless stories emerge from the same data, from time-sensitive themes, predictions, behavior and correlation.You can visualize data in endless forms, but it all boils down to one critical component: The visualization must transcend the raw data to communicate actionable insights.
Distilling valuable, actionable insights from web analytics applications require sheer hard work and in-depth understanding. The raw data holds a treasure trove of underlying insights just waiting to be uncovered – IF you know when, where, and how to look at the data. Misinterpret the numbers, and they can become a doubled-edged sword. In fact, jumping to conclusions can undermine any insights you think you’ve gleaned. Understanding attribution, data modeling, as well as the boundaries, terminology and context – are all critically important to produce a powerful analysis.
“Data is Just a Clue to the End Truth.” / Josh Smith (@joshsmithnyc)
One of the biggest and most common analytics mistakes one can make is to confuse the data with “the truth”, neglecting basic quality assurance procedures. Always question how the data is collected and perform a comprehensive audit that can reveal major tracking issues with how the data is collected. This is vital to ensure a high level of confidence in the data you’re analyzing. Any decision you base on flawed data could have catastrophic consequences on your business. As an example, some of your analytics data may include artificial visits generated by bots and spiders, as well as visits from your own company office or QA environments. Without proper analysis, you may come to the wrong conclusions about the alleged behavior of your site’s visitors.
Uncovering Meaningful Stats
To gain sustainable analytics insights from your data, you need to ask the right questions within the right context. Focus on meaningful metrics, and establish the right objectives and measurement KPIs. Consider the following question: “How many visitors have used my site search box this month?” The answer won’t get you very far. A good reframing suggestion might be: “How frequently do visitors use my site search box and what are they looking for?”, or even better: “What was the influence of the search box on revenue?”
Sometimes, asking the right questions simply isn’t enough. Collect the right metrics, but also tweak them for best results. For example, Simo Ahava, a Google Analytics Developer Expert, suggests a method of making page view data more meaningful by recording impressions that can be attributed to an active user and filtering the rest out. Meaning, pages that are not open in the active browser tab, or that have just been minimized, will not be tracked. This is a simple yet very effective process for tracking a more meaningful metric. A common mistake is to succumb to the urge (or not making a compelling case against the requirement) to report on specific data points outside of their relative context. In other words, reporting for reporting’s sake rather than identifying trends, which should be your goal. You’ll discover a much clearer picture emerges when you compare data sets over time periods, shedding light on trends and data relationships. Trend analysis is a very effective and powerful capability you want to nurture. It is often the only lens that illustrates context and spots patterns, and accurately forecasts results (e.g. seasonality, geography, or current events).
When comparing trends, avoid any arbitrary correlation between data sets. Don’t compare two distinct data sets in your web analytics reports and expect them to behave in a similar manner. Refrain from making any rush analytical decisions without forming a qualitative opinion first. Understand the context and nature of each data set. Your analytics data relies heavily on mathematics and statistics, which can both help and hinder your analysis. Calculating statistical significance can lead to better marketing decisions. In other words, understanding the likelihood that your data has not occurred by chance will improve your confidence in the data.
Another common mistake is to automatically compare one year’s results to those of the previous year. Because so much has changed during that time, comparing the two periods without understanding the underlying cause of the differences is essentially like comparing apples to oranges. For example, different tools use different metrics (and different tracking methods). If you want to compare two different data sets, make sure you’re comparing trends (as opposed to absolute numbers). Also, make sure you’re comparing two closely related metrics. For example, while ad clicks are directly tied to page views, the same visitor will often click on an ad more than once. Comparing clicks to site visits is therefore liable to produce misleading results.
Storytelling Means Everything to Your Data
Much like journalists, web analysts are essentially well-disguised storytellers. They structure data along memorable human storylines, creating narratives, making sense of the numbers, separating signals from noise, and revealing narrative visualization. Countless stories may emerge from the same data, ranging from time-sensitive themes, predictions, behavior and correlation. Carving out the story buried in the data is obviously a challenge, but also a skill well worth honing. A good story will breathe life into your data with incredible results.
Segmentation helps you uncover those stories by isolating different audience personas and relating them to the figures based on common behaviors, sources, demographics, interests and more. Segmenting the data makes it easier to elicit valuable insights. Without segmentation, you’re just looking at aggregated, out-of-context data. Approach segments as a spotlight that allows you to identify interesting stories behind objects obscured in darkness.
You can visualize data in endless forms, but it all boils down to one critical component: The visualization must transcend the raw data to communicate actionable insights. Whether you’re looking to visualize relationships, distributions, comparisons or compositions – choose the chart that highlights the deeper meaning. In his article titled “Telling a story with data“, Professor Thomas H. Davenport describes the importance of communicating results: “Bar and pie charts only scratch the surface of what you can do with visual display. There are scatterplots, matrix plots, heat maps, line graphs, bubble charts, tree maps, and many other options.”
A great case in point is this innovative chart from The New York Times, which presents a fascinating analysis of how the Great Recession has reshaped the US economy over the past 10 years:
Data visualization is about deciding what you want to communicate to your audience, and how your message is best presented. The reader must be able to recognize patterns and trends effortlessly, making snap decisions based on the visual. There’s an emotional layer involved. Charts are, as a rule, more direct than numbers. That is why visual communication is so useful for affecting behavior and engaging readers. To get started, experiment with several visualization tools to help you tell stories with data, such as Tableau, Google Fusion Tables, Visual.ly and more.
Getting Web Analytics Right
Going back to the data itself, choosing what to present is a major issue. Vanity metrics, as opposed to actionable metrics, are defined by Eric Ries in his 2011 book The Lean Startup. They might look good on a presentation, but they sugar-coat reality at best, and offer erroneous insights at worst. Vanity metrics are available, trackable and can easily indicate improvement. Nevertheless, they are devoid of context and therefore misleading. In contrast, actionable metrics are more difficult to pinpoint, but offer a clear focus that is aligned with your goals. For example, if growth is your primary objective, you might be inclined to report on a steady increase in new user acquisition. But if your churn rate is too high, that acquisition is meaningless without considering visit duration as well.
You can easily misinterpret web analytics data using a variety of metrics that may seem similar on the surface. For example, do you know the difference between the Google Analytics definition of “Absolute Unique Visitors” vs. “New and Returning”? (Read answer) How is Time on Page calculated? How does ‘Web Trends’ calculate the Average Time Viewed measure for pages? (Read answer)
No comments:
Post a Comment