Hello again. Whenever we start to think about data, there are common issues to deal with. Data analysis is inherently challenging due to the amount of options available to understand what the data means. In this lesson, we'll look at common mistakes and concerns and how you can overcome those. We'll also touch on how to come up with some of the most useful metrics for your own business objectives. In this lesson, you'll learn about the common mistakes of data analysis, and determine how to come up with useful metrics for your own business. Preparation is vital in data analysis. Abraham Lincoln said, give me six hours to chop down a tree and I will spend the first four sharpening the axe. The wisdom of this maxim holds true for data analysis. Data analysis presents common issues and errors. If we start by preparing our tools to deal with these, our work will be easier and more effective. Having the right mindset when approaching data makes a significant difference. I'd say that in data analysis, 90% of the analysis takes 90% of the time, and the last 10% may take another 90% of the time. Meaning you can get far along with data analysis but sometimes that last little element stumps you, making you wonder if you're approaching the problem in the right way. Therefore, I want to identify some common mistakes in data analysis to help you avoid getting stumped. When it comes to common analysis mistakes, the most elementary is assuming that correlation means causation. How often do we think that A = B, or A led to B, when actually, A just happened in conjunction with B? Just because A occurred around the same timeframe or happened before B doesn't necessarily mean A led to B. There's also a data bias. This includes making decisions based on a small sample size, and/or analyzing too much data. This is the classic analysis paralysis. Don't over analyze by trying to squeeze everything out of the numbers when maybe half the data will provide the information a particular report or stakeholder needs. The final mistake is working with biased data but not realizing it is skewed. This typically happens with people new to data analysis. They don't understand that seasonal trends, particularly those occurring around holidays or summer, impact spikes or drops. Concerns arise around how the data is collected or shared. Topical themes related to how you gather data and report it are very common. For instance, how is the data being collected? What is the right time period in which to analyze trends? Therefore, how do you collect the data you need? If you have access to several different data sets, perhaps an analytics package, something about ranking, and another related to search page performance, how do you determine the right time period? How should you combine data into a context that can be shared and what segment should you use? What are the slices of a website? How do you compare performance for a home page versus a landing page, r performance of your main content versus secondary content, or against different languages? Therefore, collecting and sharing data can be problematic. I want to dive into issues related to data bias that are prevalent in data analysis. Page views do not equal sales. In many cases, traffic is a vanity metric. Typically, it's not the most useful metric to consider. Vanity metrics, like traffic or ranking, can make you feel good without offering clear guidance for what to do. Sample bias leans toward extremes of negative or positive. Are you simply looking at the outliers because they attract the attention of those assessing the data, but not thinking that maybe it's a sample set that incorrectly indicates what you're trying to understand? There's also variance. Are you off by 1% or 10%? By one point or five points? Diving into the variance of the data helps provide perspective and context to the size and importance of an issue. Ultimately, you want to develop a rigor about becoming a data-driven marketer. Then you won't make decisions from a gut feeling, but use your intuition based on the data you have, while staying aware of biases that are always in place. Then you'll make the best assertion possible based on the the data you have. Common issues also occur with data visualization. This is how you depict data to make it meaningful and useful to others. Here's what I recommend you not do when visualizing data. Stay away from 3D graphs, dark backgrounds, and having baselines that don't resolve to zero. In other words, start with 100 or 1000 or whatever your baseline number is, but don't let that skew what you're reporting out on. Don't overcrowd charts with grid lines in the background. These are common errors frequently seen. If you use Excel beware because it makes it easy to fall into these errors. Aim for clarity. How do you think through bar charts, line graphs, and bulleted graphs? Spark lines are very useful, they give an indication via lines and different target points across the line. You can also use waterfall charts that show a red or green indication of success. These are some ideas to think about when it comes to data visualization. The more simple and clear you can make it, the better. Let's focus on a few key points to remember. Number one, keep it simple, the simpler, the better. Number two, report on current trends. Number three, provide context for the changes you're making to the site. Seek ways to provide additional insights. These rules are especially good for executive summaries where you're providing the highest level VIP what they need to access. Follow these rules when presenting data. It's ideal to first provide context for what was done. Then show results, either in a separate slide or side-by-side. You may want to present a commentary box of two or three bullets explaining what the data shows. Have one take-away per slide, don't throw all the data you have onto one slide or one chart. Keep it simple, nobody will be as close to the data as you, so present that data as cleanly as possible. Never assume people know what you do about the data you're presenting. They're coming from a different perspective. Explain what they'll want to know. Tell them if the trend is good or bad. Include a text box describing the trends or an image of Google Analytics, adding annotations or call outs to highlight the important parts. Less is more. Resist the temptation to flood your audience with data. Hit the key points while providing additional details in the appendix. Prepare it ahead of time, add it to the end of a slide deck, or a secondary Excel tab you can reference if it comes up in conversation.
