How to Avoid "Analysis Paralysis" When You're Drowning in Data

Sometimes I wonder what marketing analytics would have looked like during the Mad Men era of the 1960's...

You've got three (3!) channels to work with -- TV, Print, Radio -- virtually no tracking, and a pretty simple objective: throw your message out to the world, reach as many ears/eyes as possible, and cross your fingers that the awareness eventually translates into foot traffic and sales.

What a blissfully simple time it must have been.

Then, of course, came the rise of digital. With it, new platforms and ad technology created a brand new and absolutely massive media ecosystem, and the development of sophisticated tracking and measurement tools created billions of data points out of thin air.

Before digital, the customer journey was something that you could actually wrap your head around: Bill sees TV ad. Bill visits store. Bill buys toothbrush.

These days, media and advertising have permeated virtually every waking moment of our lives, to the point where ads are so damn pervasive that you couldn't avoid them if you tried. Today's customer journey isn't really a journey at all, but rather a tangled web of so many touch points that trying to make sense of them -- even with the most sophisticated tools -- is often nearly impossible.

Consider the following case:

Bill scrolls through Facebook on his iPhone and sees a post from his buddy George, raving about an incredible new toothbrush. Eager to step up his hygiene game, Bill heads to Google, clicks through a paid ad for epictoothbrushes.com and views a few products. Later that week, Bill opens up his NY Times app and is retargeted with a display banner featuring the exact toothbrush he had been checking out earlier that week. Convinced that it must be fate, Bill returns home, grabs his laptop, heads directly to the site and makes a purchase.

The million-dollar question: Why did Bill buy the toothbrush?

It sounds simple enough, until you put yourself in the shoes of the poor analyst trying to interpret the data. What he sees is one mobile paid search click that failed to drive a conversion, one mobile remarketing impression that also failed to drive a conversion, and a single, seemingly unrelated sale driven by a direct visit to the site. He has zero visibility into the fact that George's recommendation was the first link in the chain, and has no means of connecting Bill's mobile activity to the direct visit that ultimately led to a purchase.

But wait, it gets worse!

Even if our analyst could see the entire picture, the question still begs to be answered. Does he attribute the sale to George, since it was his recommendation that created awareness in the first place? Does he give credit to paid search, which drove Bill to epictoothbrushes.com instead of another retailer? What about the remarketing ad, which served as a crucial reminder and the last media touch point prior to the purchase?

You've probably realized by now that there really is no "right" answer, which is often the case when it comes to most advanced analytics exercises: segmentation, media mix, forecasting, lifetime value, etc. Even with a personal squad of NASA scientists armed with the most cutting-edge software on the planet, you will always face some degree of error -- that's just the nature of the game.

So there you are, with a meager little slice of the data you need, trying to answer a question that has no real solution. What do you do?

Option A is to succumb to what I like to call "analysis paralysis", which typically involves curling into a tiny ball and remaining motionless until the problem solves itself (not recommended).

Option B is to take action. To experiment. To do something. You may not have the tools, resources, or aforementioned squad of scientists to post up nobel-quality work, but remember that you must crawl before you can run. And if you aren't willing to crawl, you're paralyzed.

Here are a few tips to help get you on the right track:

1) Stop chasing perfection. This isn't basic algebra, and we don't often have the luxury of working with neat little black and white problems (i.e. "solve for X"). Analytics is a thousand shades of gray; it's usually not a matter of right vs. wrong, but rather wrong vs. less wrong. Perfectionism can be a virtue in many ways, but a strong analyst knows when to trade precision for progress.

2) Focus on slaying one beast at a time. There's a reason why advanced analytics can sometimes feel overwhelming, and it's not because you aren't sharp enough to hold your own. It's because these are massively complex and thorny problems that require a tremendous amount of skill, intuition, and creativity to manage (notice how I didn't say "solve"). Don't try to fight an entire army at once; choose your battles and focus on one challenge at a time.

3) Think beyond the analysis. Which is more important, a multichannel attribution model or a customer segmentation analysis? Trick question, fool! The answer depends on how -- and in some cases if -- the analysis will drive real, tangible growth. What good is a customer segmentation if you don't have the means of targeting those audiences? Why build out an attribution model if you have no flexibility to adjust your media mix? Don't lose sight of the bigger picture -- prioritize your efforts with the end user in mind.

4) Test. Learn. Repeat. Even if you don't know a statistical model from a ham sandwich, that doesn't mean you're out of options. Get creative! Split out test and control groups, run A/B or multivariate tests, or just bang on things in a somewhat systematic way and see what happens. As long as you're documenting what you're doing and learning something along the way, you're making progress.

As the saying goes: "The only thing more dangerous than trying and failing is not trying at all".   

So what about you? What are your tips for avoiding "analysis paralysis" when you feel like you're drowning in data?