The data will never ever be perfect! There, I’ve said it. Someone had to. If you are having a rough time showing a report to your boss because the data is still imperfect, just do it quickly - tear it off like a patch :) It’s better to make imperfect decisions today, than to make perfect decisions tomorrow when your company is already dead (well, maybe not dead, but a little dramatization will help you get my point). You are just going to miss the opportunity window if you are waiting for perfection. It’s as simple as that.
I know from my own experience how this notion is hard to accept. When I write an article, I always get that itching feeling that it’s not mathematically, statistically, morally (or whatever) 100% accurate. But then my cofounder comes to me and asks where’s the cool article I promised him weeks ago. And I tell him I can’t finish the content piece because it’s still imperfect. Then he has that scary look on his face like he wants to punch me really hard :) Just let it go! Useful data does not equal 100% perfect.
I mean, just look at Google Analytics, for example. It shows you data not for a 100% session but somewhere between 80% and 90%. And it overrides sources of users. Studying how it stores data inside just made my jaw drop. Maybe sampling and source overriding makes data in GA somehow imperfect, but it is still statistically significant and valid. Data imperfection doesn’t always equal data invalidity. Even Mighty Google Analytics isn’t perfect. So next time you feel your right eye twitching because of data imperfection - just let it go :)
Of course, there’s some hard limits in analytics that you shouldn’t just let go, but most of the data imperfections can be overlooked for the sake of the timely decision. Strive for the best data you can get, but don’t spend your whole life waiting for perfection; work with what you have now.
Don’t overanalyze it.
When you look at some number, always think about the context. Like if you have 50 Comments with 100 Fans - congrats you rule, and if you have 50 Comments with 1,000,000 Fans - man, you are in trouble.
Like that time when I analyzed the Fifty Shades of Grey Facebook page, they had had 6 million Fans and 4,000 comment to each post - looks pretty awesome, huh? But when I looked at these comments, 99% of them were spam. Now you can imagine how many of these Fans are zombies and cut all their stats at least in half.
What does seeing the plain Number of Likes tell you? Nothing. I have 30 Likes. How many posts do you have? And how many Followers? And how are Competitors doing with the same number of posts and followers? So many factors should be taken into account, because they will change the picture drastically.
You see where I’m heading - the context changes the picture.
Don’t overlook it.
Your site, social media page, or brand are like a dark room - you have no idea what’s going on inside, how customers interact with your product, what they think about your content and so on. That is, until you turn on the flashlight of analytics. Suddenly, you can see that customers hated your posts about super bowl and your inspirational proverbs but totally loved your silly videos about cats; that they had troubles subscribing to your newsletter on a site and have no idea how to navigate the pricing page.
But that’s only the part of the deal. Don’t only report what happened; report what to do next. When you are burying your boss under a load of figures, it’s like this dark room for him again, give him the flashlight - tell him what to do next based on this data. Recommendations are the most important part of the report.
Even if no one will ever see your hours of digging in rough data but will see a simple and actionable sentence of recommendation: “We need to invest more in a silly cat videos - they help us sell more donuts, let’s hire a cat-video guru” - it is still worth it. If you don’t show recommended actions in your report, it’s like making them do all the work you’ve done again. You’ve spent hours trying to figure out what happened and what should we do next, and then you are loading a track of figures onto your colleagues and wait until they analyze them again in their heads to figure out what to do next. To avoid such pitfalls strongly reccomend you reading Avinash Kaushik's article on the topic The Difference Between Web Reporting And Web Analysis for awesome examples of reports with recommendations.
Report Without Recommended Actions = Unfinished Analysis.
Go beyond figures and graphs to actions and recommendations.